Neural-mechanical coupling of breathing in REM sleep
C. A. SMITH, K. S. HENDERSON, L. XI, C.-M. CHOW, P. R. EASTWOOD, AND J. A. DEMPSEY
The John Rankin Laboratory of Pulmonary Medicine, Department of Preventive Medicine,
University of Wisconsin School of Medicine, Madison, Wisconsin 53705-2368
dogs; rapid-eye-movement sleep; non-rapid-eye-movement
sleep; obstructive apnea; sleep-disordered breathing
in rapid-eye-movement
(REM) sleep, especially ‘‘phasic’’ REM sleep, is substantially different from that in non-REM (NREM) sleep
(21, 26). During eupnea, frequency is generally higher
and more variable and tidal volume (VT ) is reduced in
REM sleep; during airway occlusion or with increased
arterial PCO2 (PaCO2), the more negative tracheal pressure (Ptr) values and/or increases in flow rate tend to be
more erratic and blunted in contrast to the predictable,
progressive responses to increasing chemoreceptor
stimuli observed in NREM sleep (5, 27, 30). After
release of airway occlusion in NREM sleep, marked
overshoots of flow rate and VT occur and in turn often
lead to central apnea. In contrast, in phasic REM sleep
the VT responses following occlusion are generally
much smaller and more variable, and central apneas
are very rare (2, 29).
THE REGULATION OF BREATHING
http://www.jap.org
The goal of the present study was to determine the
cause(s) of the decreased ventilatory responses during
and after airway occlusion in REM sleep. The three
most likely causes of these effects of REM sleep on the
ventilatory responses are 1) increased distortion of the
chest wall during inspiratory efforts secondary to the
atonia of rib cage and accessory respiratory muscles,
thus compromising neural-to-mechanical coupling [i.e.,
a given inspiratory neural/EMG input produces less
mechanical (pressure and flow) output] (20, 21); 2)
decreased responsiveness of chemoreceptors and/or of
medullary integration of chemoreceptor input to the
asphyxic stimuli developed during the occlusion (27);
and 3) phasic REM events that cause fractionations of
neural input to inspiratory muscles (16).
The present study analyzed the flow rates or Ptr and
diaphragm electromyogram (EMG) obtained in NREM
and REM sleep in three dogs during eupnea, during
airway occlusion, and after occlusion. The findings
support a major role for a decreased and highly variable
neural respiratory motor output, as reflected in the
diaphragm EMG in REM sleep, in accounting for the
limited and erratic pressure and ventilatory responses
obtained during eupnea and especially during and after
airway occlusion. Chest wall distortion may also contribute to the reduced ventilatory responses following
airway occlusion where ventilatory drive is highest.
METHODS
General. The present study consists of an analysis of
measurements of diaphragm EMG and of mechanical ventilatory output in three dogs studied during NREM and REM
sleep. Data on ventilatory output and breath timing obtained
in these same experimental trials have been previously
reported in studies concerned with causes of sleep apnea (5,
29), and the methods and protocol for the studies reported
here have been presented in detail in those studies. Briefly,
under general anesthesia, three trained, female dogs were
surgically prepared with chronic tracheostomies, and bipolar,
multistrand, Teflon-coated stainless steel EMG electrodes
were implanted into the crural diaphragm. The free ends of
the electrodes were tunneled under the skin and exteriorized
near the scapulae. After healing, the dogs were intubated
with cuffed endotracheal tubes and allowed to sleep in our
canine sleep laboratory. The dogs were free to choose their
own posture, either prone or lateral recumbent, and once
chosen, their posture did not change between NREM and
REM sleep.
Measurements. Diaphragm EMG and mechanical output
variables were acquired breath by breath using a computerbased analysis system developed in our laboratory. Airflow
rate was obtained by means of a pneumotachograph attached
to the endotracheal tube. The pneumotachograph was calibrated daily against five known flow rates. VT was calculated
by digital integration of the airflow signal. Ptr was obtained
by means of a pressure catheter in the endotracheal tube that
was attached to a pressure transducer (Validyne). Integrated
0161-7567/97 $5.00 Copyright r 1997 the American Physiological Society
1923
Downloaded from http://jap.physiology.org/ by 10.220.33.5 on June 14, 2017
Smith, C. A., K. S. Henderson, L. Xi, C.-M. Chow, P. R.
Eastwood, and J. A. Dempsey. Neural-mechanical coupling of breathing in REM sleep. J. Appl. Physiol. 83(6):
1923–1932, 1997.—During rapid-eye-movement (REM) sleep
the ventilatory response to airway occlusion is reduced.
Possible mechanisms are reduced chemosensitivity, mechanical impairment of the chest wall secondary to the atonia of
REM sleep, or phasic REM events that interrupt or fractionate ongoing diaphragm electromyogram (EMG) activity. To
differentiate between these possibilities, we studied three
chronically instrumented dogs before, during, and after 15–
20 s of airway occlusion during non-REM (NREM) and phasic
REM sleep. We found that 1) for a given inspiratory time the
integrated diaphragm EMG (eDi) was similar or reduced in
REM sleep relative to NREM sleep; 2) for a given eDi in
response to airway occlusion and the hyperpnea following
occlusion, the mechanical output (flow or pressure) was
similar or reduced during REM sleep relative to NREM sleep;
3) for comparable durations of airway occlusion the eDi and
integrated inspiratory tracheal pressure tended to be smaller
and more variable in REM than in NREM sleep, and 4)
significant fractionations (caused visible changes in tracheal
pressure) of the diaphragm EMG during airway occlusion in
REM sleep occurred in ,40% of breathing efforts. Thus
reduced and/or erratic mechanical output during and after
airway occlusion in REM sleep in terms of flow rate, tidal
volume, and/or pressure generation is attributable largely to
reduced neural activity of the diaphragm, which in turn is
likely attributable to REM effects, causing reduced chemosensitivity at the level of the peripheral chemoreceptors or, more
likely, at the central integrator. Chest wall distortion secondary to the atonia of REM sleep may contribute to the reduced
mechanical output following airway occlusion when ventilatory drive is highest.
1924
NEURAL-MECHANICAL COUPLING OF BREATHING IN REM SLEEP
Fig. 1. Polygraph record of a typical occlusion
trial in rapid-eye-movement (REM) sleep. Note
irregular slopes of tracheal pressure (Ptr) and
moving time average of diaphragm EMG (MTA Di
EMG). V̇, airflow; au, arbitrary units.
badly over- or underestimate the rate of rise. Similarly,
examining the rate of rise over the early portion of a breath
(e.g., 100 ms) could also be misleading, especially in breaths
with a rectangular waveform where the early rate of rise is
very steep but then reaches a plateau. Measurements of peak
values can also be misleading because of the brief, sharp
transients present in REM sleep that may not be representative of an entire breath. In the present study we have
attempted to avoid these problems by using VT (eupnea and
postocclusion), ePtr (occlusion), and eDi, which make no
assumptions about the shape or duration of the waveform.
Protocol. Our protocol is illustrated in Fig. 2. Once a stable
sleep state had been achieved (NREM or REM), a 60-s
eupneic control period was followed by tracheal occlusion,
which averaged 16.3 6 1.4 (range 11.8–21.1) s; the occlusion
was released before EEG arousal. Ventilation, Ptr, diaphragm
EMG, and EEG/electrooculogram were recorded continuously
throughout each trial, such that 60 s of eupneic control, all
occluded breathing efforts, and $60 s of postocclusive breathing were obtained. We report only on trials in which sleep
state (NREM or REM) was stable during and after the
occlusion.
Analysis and statistics. Regression slopes of ePtr (during
airway occlusion) or VT (during eupnea and after occlusion) as
a function of eDi, which we have termed ‘‘input-output’’ plots,
were compared across NREM and REM sleep only between
zero eDi and the lowest maximum eDi for each of the three
conditions (eupnea, occlusion, and postocclusion) in a given
dog for NREM or REM sleep. We did this because we thought
we could legitimately compare slopes only over comparable
ranges of data, where the data from both sleep states tended
to show linear relationships.
Differences between regression slopes were determined by
means of analysis of covariance (SYSTAT). Differences in
slopes were considered significant if P # 0.05. Differences
between grouped data were determined by independent
t-tests (SYSTAT) and were considered significant if P # 0.05.
RESULTS
Time course. There were 88 successful occlusion
trials (i.e., without change in sleep state) in the three
dogs: 59 in NREM sleep and 29 in REM sleep. The
mean eupneic values for each dog for breath timing and
for diaphragm EMG in NREM and REM sleep are
Downloaded from http://jap.physiology.org/ by 10.220.33.5 on June 14, 2017
inspiratory Ptr (ePtr, i.e., area under the inspiratory pressure waveform) was derived by computer. Raw EMGs were
amplified, rectified, and moving time averaged with a 100-ms
time constant. Integrated diaphragm activity (eDi, i.e., area
under moving-time-average waveform) was derived by computer. The eDi is reported in arbitrary units, normalized to
the daily mean of all eupneic, NREM breaths recorded for a
given dog. Fractionations of the diaphragm EMG were determined manually from the raw diaphragm EMG signal. Fractionations were defined as EMG silence for .50 ms during
inspiration.
Sleep state was determined by means of a five-lead montage of percutaneous Basmajian-type wire electroencephalogram (EEG) electrodes and standard criteria (see Ref. 29 for
details). In the present study we examined only NREM and
phasic REM sleep. Phasic REM sleep was defined as a
desynchronization of EEG, eye movement density .0.25/s
(the average during REM sleep was ,0.35/s), and loss of EMG
activity in the nuchal muscles. Any trials that did not meet
these criteria were excluded from analysis.
Choice of electrical and mechanical variables. There are at
least three potential ways to express neural-mechanical
coupling of breathing in REM sleep. One method is to use the
rate of rise of the moving time average of diaphragm EMG
and its mechanical analogs, VT-to-inspiratory time (TI ) ratio
or rate of fall of Ptr. A second method is to compare peak EMG
activity of the diaphragm and its mechanical analogs, peak
inspiratory flow and peak Ptr. A third method is to compare
eDi with its mechanical analogs, VT and Ptr. We have chosen
to use the latter method in this study of phasic REM sleep in
the dog because of problems associated with determination of
a meaningful rate of rise (or fall) in EMG, pressure, or flow.
These problems are present throughout REM sleep but can be
best illustrated by an example from an occlusion trial (Fig. 1).
Breath 1 (the last eupneic, nonoccluded breath) shows a
reasonably ramplike increase in EMG activity and inspiratory flow; however, the remaining breaths in this trial illustrate various problems. Breaths 2–4 present almost rectangular waveforms (i.e., very steep initial rate of rise followed by a
slowly rising plateau). Breaths 5 and 6 have two distinct
slopes in each breath, and breath 6 has a clear decrease in
activity in the middle of the breath. These different shapes
are reflected in Ptr during the occlusion. Thus determining a
true rate of rise is problematic for electrical and mechanical
variables. Clearly, a simple peak-over-time approach could
NEURAL-MECHANICAL COUPLING OF BREATHING IN REM SLEEP
1925
shown in Table 1. In all dogs relative to eupnea in
NREM sleep, eupneic ventilation during REM sleep
was increased due to increased breathing frequency,
which more than compensated for the trend toward
slight decreases in VT. The eDi was unchanged in two
dogs and significantly decreased in the third.
The duration of airway occlusion was similar between sleep states, averaging 15.4 6 1.3 s in NREM
sleep and 17.2 6 2.8 s in REM sleep (P 5 not significant). End-tidal gas measurements were not meaningful in the postocclusion period because the dogs invariably inspired first, thus diluting the subsequent end-tidal
CO2 measurement. However, in a previous study (5) we
mimicked 20 s of obstructive apnea in anesthetized
dogs and rapidly (every 3 s) sampled arterial blood gas
throughout the obstruction and postobstruction period.
We observed that the mean PaCO2 increased 5.5 6 0.5
Torr and the mean arterial PO2 (PaO2) decreased 27.8 6
3.1 Torr. The time course and variability of response of
VT, ePtr, and eDi are shown in Fig. 3 and Table 1. During airway occlusion in NREM sleep, eDi and ePtr
typically increased progressively with each inspiratory
effort throughout the occlusion period. In contrast,
during REM sleep, eDi and ePtr during the occlusion
period did not consistently increase in a progressive
manner, and their magnitudes tended to be smaller at
the termination of occlusion in REM than in NREM
sleep.
After release of airway occlusion in NREM sleep,
ventilatory overshoots occurred consistently, inasmuch
as VT and eDi were greater than control in most trials.
In REM sleep the ventilatory overshoots and increase
in eDi after release of occlusion were much less consistent and smaller in magnitude.
Relationship of TI and eDi in NREM vs. REM sleep.
Figure 4 shows the magnitude of eDi as a function of TI
for eupnea, airway occlusion, and postocclusion in
NREM and REM sleep. There is considerable overlap of
TI values between NREM and REM sleep. eDi tended
to decrease as a linear function of decreasing TI during
eupnea. Occlusion and postocclusion relationships of
eDi and TI could not easily be described with simple
functions, although here too eDi always tended to
decrease as TI shortened. During eupnea in REM sleep
all dogs had a number of breaths with TI values ,1 s,
whereas in NREM TI values ,1 s were unusual, and
these few were in one dog. Even in this range, eDi
continued to decrease as TI became shorter.
Neural input-mechanical output relationships, NREM
vs. REM sleep. The relationship of neural input (eDi) to
mechanical output (VT ) during eupnea is illustrated in
Downloaded from http://jap.physiology.org/ by 10.220.33.5 on June 14, 2017
Fig. 2. Polygraph records of an occlusion trial in non-REM (NREM, A) and REM (B) sleep in same dog. Note
crescendo of increasingly negative Ptr and increasing crural diaphragm EMG (Cr Di EMG) in NREM followed, on
release of occlusion, by 2 large breaths and beginning of a central apneic period. Occlusion in REM sleep produced
inconsistent increases in negative Ptr and diaphragm EMG. Note also continued REMs throughout occlusion trial.
VT, tidal volume.
1926
NEURAL-MECHANICAL COUPLING OF BREATHING IN REM SLEEP
Table 1. Ventilatory data during eupnea, occlusion, and postocclusion periods
Dog 1
NREM
Dog 2
REM
Dog 3
NREM
REM
NREM
REM
129
107 6 6.1
2.79 6 0.12*
1.50 6 0.04*
16.7 6 0.7*
4.96 6 0.19*
0.32 6 0.01*
172
105.7 6 3.3
3.45 6 0.14
1.35 6 0.03
14.8 6 0.3
3.67 6 0.07
0.27 6 0.01
110
110 6 6.1
3.07 6 0.17
1.30 6 0.04
17.3 6 1.1*
4.33 6 0.18
0.29 6 0.01
50
248.5 6 17.1
2.49 6 0.16*
1.72 6 0.06
16 6 0.9*
9.9 6 0.6*
54
297 6 16.9
2.09 6 0.14
2.3 6 0.08
15.1 6 0.6
8.4 6 0.4
33
190.9 6 22.2*
2.15 6 0.3
2.42 6 0.12
15.1 6 1
7.8 6 0.6
20
341.8 6 35.6
3.2 6 0.31
1.78 6 0.08
13.1 6 0.8
8.7 6 0.76
0.67 6 0.04*
22
222.3 6 18.7
2.57 6 0.68
1.25 6 0.08
20.3 6 1.6
9.51 6 0.61
0.5 6 0.03
12
203.2 6 24.8
2.60 6 0.29
1.35 6 0.08
16.0 6 1.1
7.26 6 0.56*
0.46 6 0.03
Eupnea
n
eDi EMG, AU
TE, s
TI, s
f, breaths/min
V̇I, l/min
VT, liter
352
90.2 6 0.9
3.2 6 0.04
1.58 6 0.31
13.1 6 0.2
3.79 6 0.07
0.32 6 0.003
225
68.8 6 2.7*
1.80 6 0.08*
1.23 6 0.03*
27.3 6 1.3
5.58 6 0.19*
0.24 6 0.006*
152
120.9 6 2.8
4.19 6 0.08
1.9 6 0.02
10.1 6 0.1
3.84 6 0.06
0.38 6 0.004
Occlusion
n
eDi EMG, AU
TE, s
TI, s
f, breaths/min
ePtr, AU
37
152.9 6 17.3*
1.17 6 0.14*
2.33 6 0.13*
19 6 1.1*
6.3 6 0.6
34
233.7 6 15.9
3.52 6 0.67
1.34 6 0.04
16.5 6 1.1
9.08 6 0.9
0.62 6 0.02
22
230.4 6 24.5
2.03 6 0.33
1.35 6 0.06
20.1 6 1.3*
10.6 6 1.23
0.54 6 0.03*
47
287.9 6 16.9
3.78 6 0.14
1.85 6 0.04
11.1 6 0.4
11.8 6 0.7
Postocclusion
n
eDi EMG, AU
TE, s
TI, s
f, breaths/min
V̇I, l/min
VT, liter
32
322.2 6 31.3
3.16 6 0.25
1.73 6 0.06
13.1 6 0.6
10.49 6 0.72
0.81 6 0.05
Values are means 6 SE; n, no. of observations. REM, rapid-eye-movement sleep; NREM, non-REM sleep; eDi EMG, area of integrated
moving-time-averaged diaphragm EMG; AU, arbitrary units; TE and TI, expiratory and inspiratory time; f, breathing frequency; V̇I, minute
ventilation; VT, tidal volume; ePtr, integrated inspiratory tracheal pressure. * Significantly different from NREM, P # 0.05.
Fig. 5, and mean values are listed in Table 2. In all
cases the slopes of the relationship between VT and eDi
were significantly different from zero. Also in two dogs
there was a small but significant increase in slope in
REM vs. NREM sleep, and in the third dog there was a
small but significant decrease in slope. On average, at a
representative eDi of 100 units, there was an ,6%
decrease in VT during REM vs. NREM sleep.
The neural input-mechanical output relationships
during occlusion and after occlusion are illustrated in
Figs. 6 and 7, and mean slope data are presented in
Table 2. The slope of eDi vs. ePtr (occlusion) or VT
(postocclusion) was significantly greater than zero in
all conditions in all dogs. During airway occlusion the
slopes of the neural input-mechanical output relationships in REM were equal to or slightly less than those
in NREM sleep. On average, at a representative value
for eDi of 400 units, ePtr was only ,2% less in REM
than in NREM sleep. In contrast, after airway occlusion the slopes of the neural input-mechanical output
relationships in REM sleep were significantly less than
those observed in NREM sleep. Thus, on average, for an
eDi of 400 units the VT would be ,26% smaller in REM
than in NREM sleep.
Fractionations of EMG and mechanical outputs in
REM sleep. Examples of diaphragm EMGs with and
without fractionations are shown in Fig. 8. Fractionations longer than ,50 ms were detected in 41% of the
occluded breaths. Some of the fractionated breaths had
more than one fractionation; thus, on average, there
were 1.6 fractionations per fractionated breath during
airway occlusion. Sixty-three percent of the fractionations were associated with positive-going spikes in
Ptr, i.e., an interruption in the ramp of negative-going
Ptr. This mechanical effect of EMG fractionation was
usually associated with the longer (more than ,100ms) fractionations.
DISCUSSION
In summary, we have found that 1) for a given TI the
eDi was similar or reduced in REM sleep relative to
NREM sleep; 2) for a given activation of the diaphragm
EMG in response to airway occlusion and the hyperpnea following occlusion, the mechanical output (flow
or pressure) is similar or slightly reduced in REM vs.
NREM sleep; in eupnea the results are equivocal; 3) for
comparable durations of airway occlusion the eDi and
ePtr tended to be smaller and more erratic in REM
than in NREM sleep; and 4) fractionations of the
diaphragm EMG during airway occlusion in REM sleep
occurred in ,40% of breathing efforts and in some
instances may have contributed to the diminished and
erratic ventilatory responses observed. Thus reduced
mechanical output during and after airway occlusion in
REM sleep in terms of flow rate, VT, and/or pressure
generation appears to be attributable largely to reduced neural activity of the diaphragm, which in turn is
likely attributable to REM effects, causing reduced
Downloaded from http://jap.physiology.org/ by 10.220.33.5 on June 14, 2017
53
240.3 6 16.4
2.54 6 0.13
2.95 6 0.08
11.4 6 0.4
6.8 6 0.3
NEURAL-MECHANICAL COUPLING OF BREATHING IN REM SLEEP
1927
chemosensitivity at the chemoreceptor level or, more
likely, at the level of central integration.
Critique of assumptions. Interpretation of inputoutput relationships during and after airway occlusion
relies on the assumption that similar occlusion times
would produce similar changes in PaCO2 and PaO2 (and
saturation) in REM and NREM sleep. There are several
potential limitations to this assumption. 1) The initial
PaCO2 and PaO2 will of course affect the blood-gas status
that is ultimately achieved during occlusion. The
initial end-tidal CO2 values were not different between
NREM and REM sleep, so this was probably not a
factor in this study. 2) We showed previously (5; see
RESULTS ) that significant asphyxic stimuli build up over
the occlusion period in the anesthetized dog. The
degree of asphyxia should be at least as great in dogs
that are merely sleeping. 3) Metabolic rate may not be
equal between NREM and REM sleep. We have assumed that there are minimal differences in metabolic
rate between REM and NREM sleep, inasmuch as
metabolic rate was not measured in this study. However, we are aware of three studies in the human
literature that suggest no change in metabolic rate
between NREM and REM sleep or an increase in REM
of ,5%, so this assumption seems to be a reasonable
one (3, 22, 26). 4) Functional residual capacity (FRC)
Downloaded from http://jap.physiology.org/ by 10.220.33.5 on June 14, 2017
Fig. 3. Time course of response to airway occlusion trials in all dogs during
NREM (A) and REM sleep (B). Of 60-s
eupneic control period, 20 s are shown.
All efforts during occlusion are shown;
time 0, 1st occluded breath. First 2
breaths after release of occlusion are
shown; time 0, 1st postocclusion breath.
In NREM sleep, note progressive increase in integrated Ptr (ePtr) and
integrated diaphragm EMG activity
(eDi) over time of occlusion. Also note
VT and eDi achieved after occlusion.
REM sleep shows more variable eupneic breaths, a reduced and disorganized response of ePtr and eDi during
occlusion, and generally smaller VT and
eDi after occlusion. Linear regression
of ePtr and eDi as a function of time
during occlusion was positive and significant in all dogs in NREM and REM
sleep.
1928
NEURAL-MECHANICAL COUPLING OF BREATHING IN REM SLEEP
may decrease in REM relative to NREM sleep. We did
not measure FRC in this study, but the atonia of REM
probably leads to a small decrease in FRC (12, 18). This
in turn would tend to cause a more rapid development
of asphyxic stimuli than during NREM sleep. So there
is the potential for slightly higher chemoreceptor stimuli
for a given length of occlusion in REM than in NREM
sleep. We believe that these effects are likely to be
small. Moreover, our approach of looking at the entire
response breath by breath by comparing neural input
with mechanical output throughout airway occlusion
should compensate for small discrepancies in stimulus
levels unless one state or the other had quite alinear
response characteristics, which does not seem to be the
case.
REM effects on diaphragm EMG and breath timing.
Orem et al. (19–21) showed in unanesthetized, chronically instrumented cats that the rate of rise of activity
of medullary augmenting inspiratory cells, the rate of
rise of diaphragm EMG, and the mean EMG were
markedly increased during eupneic breathing in REM
sleep. In their studies, at any given TI, virtually all
Fig. 5. Input-output relationships during eupnea in NREM and REM sleep. A
and B: scatter plots of all breaths during
NREM and REM sleep, respectively, in all
dogs. Note generally similar ranges of
both variables whatever sleep state, although REM sleep manifests greater variability and includes some extremely shortened TI values. C: regression lines for all 3
dogs in NREM (solid lines) and REM
sleep (dashed lines). Regression lines were
generated only from data between zero
eDi and lowest maximum eDi value in
NREM or REM sleep.
breaths showed a greater rate of rise of diaphragm
EMG during REM than during NREM sleep. At the
very short TI values that are commonly observed in
REM sleep, the rate of rise of diaphragm EMG was
even greater. Despite these increases in neural inputs,
the generation of negative Ptr during inspiration was
reduced. Given these observations, Orem and colleagues proposed that atonia of the chest wall must
have occurred in REM sleep, which in turn produced
much less efficient coupling of EMG activity to mechanical output. This supports Bryan’s (4) suggestion that, in
the human infant at least, the increased rate of rise of
diaphragm EMG is a means of compensating for this
inefficient coupling. This is a logical suggestion, inasmuch as it is well known that although the diaphragm
is largely spared, activation of respiratory chest wall
muscles is reduced or absent during REM sleep (9, 24),
which could contribute to increased chest wall distortion.
Our findings in the dog using eDi would appear to
disagree with the findings of Orem et al. (19–21) in
Downloaded from http://jap.physiology.org/ by 10.220.33.5 on June 14, 2017
Fig. 4. eDi as a function of inspiratory
time (TI ) during eupneic breathing in
REM (A) and NREM sleep (B). All trials
in all dogs are shown. Note considerable overlap of TI values between
NREM and REM sleep and prevalence
of breaths in REM sleep that have TI
values ,1 s. Despite these short TI
values, eDi continued to falll as TI
decreased. Mean eDi for each dog for
all breaths with TI between 1 and 2 s (a
range common to all conditions) combined revealed no significant differences between NREM and REM sleep
for eupnea (108 6 21 vs. 102 6 18),
occlusion (161 6 84 vs. 205 6 73), or
postocclusion (346 6 165 vs. 300 6 68)
breathing.
1929
NEURAL-MECHANICAL COUPLING OF BREATHING IN REM SLEEP
Table 2. NREM vs. REM regression slopes
Eupnea (VT 5 m · eDi 1 b)
m
Dog 1
NREM
REM
Dog 2
NREM
REM
Dog 3
NREM
REM
b
Occlusion (ePtr 5 m · eDi 1 b)
r2
n
m
Postocclusion (VT 5 m · eDi 1 b)
b
r2
n
m
b
r2
n
0.00161†
0.00195*†
0.18
0.11
0.65
0.72
352
224
0.0662†
0.0682†
4.45
3.82
0.78
0.67
53
37
0.00126†
0.00087*†
0.32
0.34
0.82
0.72
34
22
0.00101†
0.00134*†
0.259
0.176
0.41
0.54
152
125
0.070†
0.068†
1.58
0.493
0.92
0.88
47
50
0.00132†
0.0007*†
0.386
0.43
0.82
0.37
32
20
0.00173†
0.00132*†
0.09
0.15
0.66
0.45
172
110
0.059†
0.0345*†
0.66
0.31
54
33
0.00151†
0.0006*†
0.168
0.229
0.89
0.48
26
14
2.21
11.84
n, No. of observations; m, slope; b, intercept. * Significantly different from NREM, P # 0.05. † Significantly different from zero, P # 0.05.
25) found no consistent effect of REM sleep on the rate
of rise of diaphragm EMG activity during eupnea.
Without a consensus in the cat, it is clear that more
work in the cat model is required before this issue can
be resolved.
REM effects on neural input-mechanical output relationships during chemoreceptor-driven breathing. The
effects of phasic REM sleep on mechanical output and
on diaphragm EMG and their relationship were most
striking in response to increasing chemostimulation, as
occurred during and immediately after airway occlusion. ePtr (during occlusion) and VT (after occlusion)
were markedly reduced and highly variable in REM vs.
NREM sleep. Furthermore, inasmuch as chemoreceptor stimuli increased with time during the occlusion,
peak Ptr often failed to decrease continuously breath by
breath in REM sleep, unlike the situation in NREM
sleep, where each succeeding breath is typically more
negative than the preceding breath. Most importantly,
these effects of phasic REM sleep on mechanical output
were also present in the eDi, inasmuch as the relationships of mechanical output to EMG of the diaphragm
were significant and equal or slightly reduced in REM
vs. NREM sleep. Accordingly, we attribute these effects
of REM sleep on the reduced amplitude and increased
variability of Ptr during airway occlusion and of inspiratory flow rate and VT after release of occlusion largely
to REM sleep effects on neural respiratory motor
output via decreased chemosensitivity.
REM effects on chest wall stability during chemostimulated breathing appear to play little or no modulatory role in ventilatory responses during airway occlusion in the absence of volume changes. However, chest
Fig. 6. Input-output relationships during occlusion in NREM and REM sleep. Conventions are similar to Fig. 5. Note overlap of
EMG and mechanical output variables in
NREM sleep compared with REM sleep, but
largest magnitudes were observed during
NREM sleep. Over a common range of eDi,
ePtr responses in REM sleep are similar to or
less than those during NREM sleep.
Downloaded from http://jap.physiology.org/ by 10.220.33.5 on June 14, 2017
cats, inasmuch as we found no change or a decrease in
eDi during eupneic breathing in phasic REM vs.
NREM sleep in the dog (Table 1). Moreover, when these
data were normalized by plotting eDi as a function of
TI, we found no difference in the eDi between eupneic
breaths taken in REM and NREM sleep. Very short TI
values (,1 s) were present only in REM sleep, but the
eDi in REM sleep continued to decrease as TI became
shorter (Fig. 4). Thus our data in the sleeping dog do
not support the idea of a compensatory increase in
ventilatory drive during eupnea in REM sleep.
There are several possible explanations for this apparent disagreement between studies. First, our observations were confined to phasic REM sleep, i.e., periods
with consistently high eye movement densities, whereas
Orem and colleagues (19–21) used data from tonic and
phasic REM sleep. At least one study in unanesthetized
cats (16) showed that the rate of rise of diaphragm
EMG is more variable in phasic than in tonic REM
sleep, and the range of values not only overlaps but also
extends above and below values observed in tonic REM
sleep. However, we believe that the large number of
observations obtained in the present study precludes a
systematic error due to variability. Second, we do not
believe that state-related changes in posture were a
factor, because our dogs remained in a constant posture
throughout our observations in REM and NREM sleep
and our diaphragm EMGs did not appear to be contaminated by nondiaphragm EMGs. Third, species differences (cats vs. dogs) may be significant in neural
control of respiratory muscles and/or chest wall compliance. However, in contrast to Orem and colleagues
(19–21), at least two studies in unanesthetized cats (16,
1930
NEURAL-MECHANICAL COUPLING OF BREATHING IN REM SLEEP
Fig. 7. Input-output relationships during postocclusion breathing in NREM and REM sleep.
Conventions are similar to Figs. 5 and 6. Note
overlap of EMG and mechanical output variables in NREM compared with REM sleep,
but largest magnitudes were observed during
NREM sleep. Over a common range of eDi,
responses in REM sleep are similar to or less
than those during NREM sleep.
and show markedly blunted responses relative to those
in NREM sleep? We would predict that this would be
the case, but clearly direct measurement of medullary
neuronal activity during chemostimulation is needed in
REM sleep.
Why is diaphragm EMG reduced and more variable
in REM sleep? We believe that there are four possible
explanations for decreased diaphragm EMG activity in
response to chemostimulation in REM sleep.
First, brief interruptions or fractionations of the
ramp of diaphragm EMG accounted for at least some of
the decreased responsiveness to chemostimulation during airway occlusion (Fig. 8). In cats, Veasey et al. (28)
observed that fractionations were present in 13–27% of
eupneic breaths, and we observed EMG fractionations
in 41% of occluded breaths. We also observed that many
of the longer diaphragm EMG fractionations actually
resulted in a measurable effect on Ptr, flow rate, or VT
developed during inspiration. That not all the EMG
fractionations had obvious mechanical consequences
may not be surprising, inasmuch as many EMG fractionations were very brief; furthermore, Hendricks and
Kline (11) showed considerable regional within-breath
Fig. 8. A: raw diaphragm EMGs and Ptr during occlusion in same dog in NREM sleep, REM sleep with no
measurable fractionations, REM sleep with a fractionation but no measurable Ptr change, and REM sleep with
measurable fractionations in EMG and corresponding positive-going transients in Ptr. Fractionations may be even
more prevalent than data from present study might suggest. B: prominent fractionations during eupnea in REM
sleep in an identically prepared dog currently in use in other studies in our laboratory for which extensive eupneic
data are available during NREM and REM sleep. Note prominent diaphragm EMG fractionations during phasic
REM sleep and clear effects on flow, VT, and Ptr.
Downloaded from http://jap.physiology.org/ by 10.220.33.5 on June 14, 2017
wall distortion secondary to the muscle atonia of REM
sleep appeared to reduce significantly the ventilatory
responses in the postocclusion period (Fig. 7, Table 2).
It is not clear why this reduction was observed only in
the postocclusion period and not during occlusion or in
eupnea. We speculate that the contribution of REM
sleep-induced atonia and associated chest wall distortion is sufficiently small in this species (see below) that
mechanical effects can be detected only where ventilatory drive, airflow, and VT values are high, such as in
the postocclusion period.
Our data obtained during and after airway occlusion
are consistent with the blunted and erratic responses of
ventilatory output to experimentally induced hypoxia
and hypercapnia previously reported in REM sleep in
humans, cats, and dogs (7, 25, 27). On the basis of our
present data in the dog, we would attribute these
blunted ventilatory responses to chemostimulation to a
reduced respiratory neural output. Contrary to the
increased rate of rise of medullary inspiratory neurons
reported during eupnea in REM sleep in the cat (see
above), would medullary inspiratory neuronal activity
reflect what we observed in the dog’s diaphragm EMG
NEURAL-MECHANICAL COUPLING OF BREATHING IN REM SLEEP
less dependent on apnea duration, perhaps contributing to prolongation of occlusive apneas in REM.
The contributions of Dr. Gordon S. Mitchell are gratefully acknowledged.
This study was supported by grants from the National Institutes of
Health. P. R. Eastwood was the recipient of National Health and
Medical Research Council of Australia Fellowship 967312.
Address for reprint requests: C. A. Smith, Dept. of Preventive
Medicine, 504 N. Walnut St., Madison, WI 53705-2368.
Received 3 January 1997; accepted in final form 29 July 1997.
REFERENCES
1. Badr, M. S., J. B. Skatrud, J. A. Dempsey, and R. L. Begle.
Effect of mechanical loading on expiratory and inspiratory
muscle activity during NREM sleep. J. Appl. Physiol. 68: 1195–
1202, 1990.
2. Bradley, T. D., W. T. McNicholas, R. Rutherford, J. Popkin,
N. Zamel, and E. A. Phillipson. Clinical and physiologic
heterogeneity of the central sleep apnea syndrome. Am. Rev.
Respir. Dis. 134: 217–221, 1986.
3. Brebia, D. R., and K. Z. Altshuler. Oxygen consumption rate
and electroencephalographic stage of sleep. Science 17: 1621–
1623, 1965.
4. Bryan, M. H. The work of breathing during sleep in newborns.
Am. Rev. Respir. Dis. 119: 137–138, 1979.
5. Chow, C.-M., L. Xi, C. A. Smith, K. W. Saupe, and J. A.
Dempsey. A volume-dependent apneic threshold during NREM
sleep in the dog. J. Appl. Physiol. 76: 2315–2325, 1994.
6. Dempsey, J. A., and J. B. Skatrud. A sleep-induced apneic
threshold and its consequences. Am. Rev. Respir. Dis. 133:
1163–1170, 1986.
7. Douglas, N. J., D. P. White, J. V. Weil, C. K. Pickett, and
C. W. Zwillich. Hypercapnic ventilatory response in sleeping
adults. Am. Rev. Respir. Dis. 126: 758–762, 1982.
8. Gleeson, K., C. W. Zwillich, and D. P. White. The influence of
increasing ventilatory effort on arousal from sleep. Am. Rev.
Respir. Dis. 142: 295–300, 1990.
9. Grassino, A. E., and M. D. Goldman. Respiratory muscle
coordination. In: Handbook of Physiology. The Respiratory System. Mechanics of Breathing. Bethesda, MD: Am. Physiol. Soc.,
1986, sect. 3, vol. III, pt. 2, chapt. 27, p. 463–480.
10. Harms, C. A., Y.-J. Zeng, C. A. Smith, E. H. Vidruk, and J. A.
Dempsey. Negative pressure-induced deformation of the upper
airway causes central apnea in awake and sleeping dogs. J. Appl.
Physiol. 80: 1528–1539, 1996.
11. Hendricks, J. C., and L. R. Kline. Differential activation
within costal diaphragm during rapid-eye-movement sleep in
cats. J. Appl. Physiol. 70: 1194–1200, 1991.
12. Hudgel, D. W., R. J. Martin, M. Capehart, B. Johnson, and
P. Hill. Contribution of hypoventilation to sleep oxygen desaturation in chronic obstructive pulmonary disease. J. Appl. Physiol.
55: 669–677, 1983.
13. Iber, C., A. Berssenbrugge, J. B. Skatrud, and J. A. Dempsey. Ventilatory adaptations of resistive loading during wakefulness and non-REM sleep. J. Appl. Physiol. 52: 607–614, 1982.
14. Iber, C., S. F. Davies, R. C. Chapman, and M. M. Mahowald.
A possible mechanism for mixed apnea in obstructive sleep
apnea. Chest 89: 800–805, 1986.
15. Issa, F. G., and S. Bitner. Effect of route of breathing on the
ventilatory and arousal response to hypercapnia in awake and
sleeping dogs. J. Physiol. (Lond.) 465: 615–628, 1993.
16. Kline, L. R., J. C. Hendricks, R. O. Davies, and A. I. Pack.
Control of activity of the diaphragm in rapid-eye-movement
sleep. J. Appl. Physiol. 61: 1293–1300, 1986.
17. Lydic, R., H. A. Baghdoyan, R. Wertz, and D. P. White.
Cholinergic reticular mechanisms influence state-dependent ventilatory response to hypercapnia. Am. J. Physiol. 261 (Regulatory
Integrative Comp. Physiol. 30): R738–R746, 1991.
18. Muller, N. L., P. W. Francis, D. Gurwitz, H. Levison, and
A. C. Bryan. Mechanism of hemoglobin desaturation during
rapid-eye-movement sleep in normal subjects and in patients
with cystic fibrosis. Am. J. Respir. Crit. Care Med. 121: 463–469,
1980.
Downloaded from http://jap.physiology.org/ by 10.220.33.5 on June 14, 2017
heterogeneity of fractionations in the costal diaphragm
of cats. Thus many EMG fractionations may occur in
only a small region of the diaphragm and/or may be too
brief to have a significant mechanical effect.
Second, the generally faster and more variable frequency of inspiratory efforts during chemostimulation
in phasic REM sleep suggests that many inspirations
are probably terminated early. This interruption of the
diaphragmatic EMG activity during phasic REM events
could account for the reduced amplitude of EMG activity.
Third, REM sleep or phasic REM events may actually interfere with true chemosensitivity at the level of
the chemoreceptor(s) and/or at the level of central
integration of chemoreceptor sensory inputs. At our
level of measurement of stimulus response, we cannot
distinguish between these proposed effects. Indeed,
even the effects of diaphragm EMG fractionations or TI
truncation on diaphragm EMG (see above) may be
interpreted as an apparent blunting of the ‘‘chemoresponse.’’ The only advance our findings have made in
understanding the nature of this REM-induced blunted
response of ventilatory output and pressure development to the asphyxia of airway occlusion is to show that
it occurs at the level of the neural activation of the
diaphragm. Other respiratory mechanoreceptor reflex
responses from lung stretch and from airway stimulation have also been shown to be blunted in phasic REM
sleep (10, 15, 17, 27, 30). However, we think it unlikely
that these reflexes are involved in the reduced ventilatory responses in REM sleep, because 1) they tend to be
inhibitory, and blunting an inhibitory reflex should, if
anything, enhance ventilation, and 2) the increase of
ventilatory efforts over the duration of occlusion strongly
suggests to us a buildup of chemical stimuli (5).
Fourth, the increase in cerebral blood flow during
REM sleep demonstrated by Parisi et al. (23) in the
sleeping goat would probably have reduced medullary
PCO2. This in turn would have reduced the stimulus at
the level of the medullary chemoreceptors and may
have caused an apparent reduction in chemosensitivity
in REM sleep.
Implications of decreased chemoresponsiveness to
breathing during REM sleep. The depressed chemoresponsiveness of REM sleep may have some bearing on
breathing stability in this state. Central apneas in
NREM sleep are commonly caused by ventilatory overshoots resulting from many types of stimuli, including
transient periods of airway obstruction (2, 5, 14). In the
human, these types of central apnea are thought to be
primarily determined by transient hypocapnia (2, 6,
14), whereas in the dog, hypocapnia and vagally mediated lung stretch (prompted by VT overshoots) cause
ensuing central apneas (5, 29). The reduced chemoresponsiveness to the asphyxia of airway occlusion means
that the subsequent overshoot of ventilation and especially VT are blunted in phasic REM sleep, and this
lessens the probability of ensuing central apnea or
hypopnea (29). The reduced diaphragmatic EMG and
pressure responses during occlusion in REM sleep may
also mean that increasing sensory input from the chest
wall leading to cortical arousal (8) will be reduced and
1931
1932
NEURAL-MECHANICAL COUPLING OF BREATHING IN REM SLEEP
19. Orem, J. Excitatory drive to the respiratory system in REM
sleep. Sleep 19: S154–S156, 1996.
20. Orem, J., and C. A. Anderson. Diaphragmatic activity during
REM sleep in the adult cat. J. Appl. Physiol. 81: 751–760, 1996.
21. Orem, J., A. Netick, and W. C. Dement. Breathing during
sleep and wakefulness in the cat. Respir. Physiol. 30: 265–289,
1977.
22. Palca, J. W., J. M. Walker, and R. J. Berger. Thermoregulation, metabolism, and stages of sleep in cold-exposed men. J.
Appl. Physiol. 61: 940–947, 1986.
23. Parisi, R. A., J. A. Neubauer, M. M. Frank, T. V. Santiago,
and N. H. Edelman. Linkage between brain blood flow and
respiratory drive during rapid-eye-movement sleep. J. Appl.
Physiol. 64: 1457–1465, 1988.
24. Parmeggiani, P. L., and L. Sabattini. Electromyographic
aspects of postural respiratory and thermoregulatory mechanisms in sleeping cats. Electroencephalogr. Clin. Neurophysiol.
33: 1–13, 1972.
25. Remmers, J. E., D. Bartlett, Jr., and M. D. Putnam. Changes
in the respiratory cycle associated with sleep. Respir. Physiol. 28:
227–238, 1976.
26. Ryan, T., S. Mlynczak, T. Erickson, S. F. P. Man, and G. C. W.
Man. Oxygen consumption during sleep: influence of sleep stage
and time of night. Sleep 12: 201–210, 1989.
27. Sullivan, C. E., E. Murphy, L. F. Kozar, and E. A. Phillipson.
Ventilatory responses to CO2 and lung inflation in tonic versus
phasic REM sleep. J. Appl. Physiol. 47: 1304–1310, 1979.
28. Veasey, S. C., J. C. Hendricks, L. R. Kline, and A. I. Pack.
Effects of acute sleep deprivation on control of the diaphragm
during REM sleep in cats. J. Appl. Physiol. 74: 2253–2260, 1993.
29. Xi, L., C.-M. Chow, C. A. Smith, and J. A. Dempsey. Effects of
REM sleep on the ventilatory response to airway occlusion in the
dog. Sleep 17: 674–687, 1994.
30. Xi, L., C. A. Smith, K. W. Saupe, K. S. Henderson, and J. A.
Dempsey. Effects of rapid-eye-movement sleep on the apneic
threshold in dogs. J. Appl. Physiol. 75: 1129–1139, 1993.
Downloaded from http://jap.physiology.org/ by 10.220.33.5 on June 14, 2017