0021-972X/91/7204-0854$03.00/0 Journal of Clinical Endocrinology and Metabolism Copyright © 1991 by The Endocrine Society
Vol. 72, No. 4 Printed in U.S.A.
Thirty-Second Sampling of Plasma Growth Hormone in Man: Correlation with Sleep Stages* REINHARD W. HOLLf, MARK L. HARTMAN, JOHANNES D. VELDHUIS, WILLIAM M. TAYLOR, AND MICHAEL 0. THORNER Division of Endocrinology and Metabolism, Department of Medicine, University of Virginia Health Sciences Center, Charlottesville, Virginia 22908
ABSTRACT. Both fasting and sleep increase the secretion of human GH and, therefore, might explain its predominantly nocturnal release. To study the precise temporal relationship between GH secretory episodes and cortical activity, GH measurements and electroencephalogram sleep stage recordings were performed every 30 s for 8 h in six young male volunteers fasted for 24 h. GH was measured in two drops of whole blood, which were directly sampled into the assay tube using a continuous blood withdrawal pump and a fraction collector. Concomitant serum sampling during a GH-releasing hormone test (n = 4) revealed a high correlation (r = 0.98) between GH measurements in serum and whole blood. GH pulses were objectively identified with Cluster analysis, and GH secretion rates were calculated with a waveform-independent deconvolution algorithm. When
data were analyzed as replicates with 1-min intervals, the nocturnal pulse frequency was 1.2 pulses/h. Elimination of data points demonstrated 43% and 64% reductions in the number of GH pulses detected for 5- and 20-min sampling intervals, respectively. Mean GH concentrations and secretory rates were significantly higher during stage 3 and 4 sleep compared to stage 1, 2, and rapid eye movement sleep. GH secretory rates and peripheral GH concentrations were maximally correlated with sleep stage, with lags of 4.5 and 16 min, respectively, suggesting that maximal GH release occurs within minutes of the onset of stage 3 or 4 sleep. This temporal coincidence between pituitary GH secretion and A sleep is consistent with cortical control over hypothalamic-pituitary function. (J Clin Endocrinol Metab 72: 854-861, 1991)
T
tions. Rather than processing blood to serum or plasma, which is then assayed for the hormone, we measured GH directly in drops of heparinized whole blood. This technique, which was first applied in studies of GH pulsatility in conscious rats (2), allowed us to determine circulating GH concentrations in man at 30-s intervals. Spontaneous release of GH is augmented within 1-2 days of fasting. In particular, pulse frequency and pulse amplitude are both increased (3, 4). The confounding influences of different nutrients on the rhythmicity of GH release are avoided when fasting subjects are studied (5). It has long been recognized that GH release is also increased during the night (6, 7). Thus, nocturnal blood sampling for GH has been proposed as a diagnostic alternative to pharmacological testing of GH secretory capacity (8, 9). The physiological basis for the predominantly nocturnal release of GH is still not completely understood. Both metabolic effects (reduced suppression of GH release by decreased nutritional intake) as well as reduced inhibitory brain activity during sleep have been considered. Previous studies with relatively infrequent sampling in a limited number of subjects found a relationship between the onset of slow wave sleep and the release of GH (10-12). However, the exact relationship between sleep and pituitary hormone secretion remains
HERE has been considerable interest in the study of spontaneous pulsatile GH release in recent years. However, it is still a matter of debate how frequently blood must be sampled for optimal detection of serum GH pulses. While a sampling interval of 20 min is the most widely used, a recent study demonstrated that sampling every 5 min detects a significantly larger number of pulses compared to sampling at 10- or 20-min intervals (1). It is impractical to further increase the frequency at which conventional serum samples are drawn. Therefore, we developed a different approach to enable more frequent measurements of GH concentra-
Received June 22,1990. Address all correspondence and requests for reprints to: Dr. Michael O. Thorner, Kenneth R. Crispell Professor of Medicine, Division of Endocrinology and Metabolism, Box 511, University of Virginia Health Sciences Center, Charlottesville, Virginia 22908. * Presented in part at the 71st Annual Meeting of The Endocrine Society, Seattle, WA, 1989. This work was supported by NIH Grants DK-32632 (to M.O.T.), Research Career Development Awards 1K04HD-00634 and HD-16806 (to J.D.V.), Grant KO8-HD-00860 (to M.L.H.), Diabetes and Endocrinology Research Center Grant DK38942, Grant RR-0847 (to University of Virginia General Clinical Research Center, including CLINFO), the Pratt Foundation, and the University of Virginia Academic Enhancement Fund (to the Biodynamics Institute). t Supported by DFG (Ho 1041). Present address: Department of Pediatrics I, University of Ulm Medical Center, Ulm/Donau, Germany.
854
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 19 July 2015. at 05:21 For personal use only. No other uses without permission. . All rights reserved.
30-s SAMPLING OF GH IN MAN
to be fully understood. We, therefore, studied fasting male volunteers during sleep by recording the stages of sleep and measuring GH concentrations every 30 s. The aim of this study was to answer the following questions. 1) Does an increase in sampling frequency result in enhanced identification of nocturnal GH pulses? 2) Can the presumptive relation between GH release and slow wave sleep be confirmed by data using high frequency sampling and deconvolution-based estimates of GH secretion?
Subjects and Methods
855
On the night of the study, blood withdrawal via the pump was started at 2200 h. In addition, a heparin lock was inserted in an arm vein opposite the pump and electrodes were attached for the recording of electroencephalogram (EEG), electromyogram, electrooculogram, and electrocardiogram. These parameters were recorded continuously by instruments in a separate room. During the night, the volunteers were monitored by microphone as well as optically with an infrared camera. The room was entered only when collection tubes were exchanged (every 2 h). At 0600 h the subjects were awakened, and an iv bolus of human GH-releasing hormone [GHRH-(1-40)-OH; 1 /xg/kg BW] was administered. For 2 h thereafter, blood was drawn both continuously through the pump to allow 30-s sampling as well as through a separate heparin lock every 20 min.
Subjects Six healthy male volunteers, aged 21-30 yr, were studied after giving informed written consent. The study protocol was approved by the Human Investigation and General Clinical Research Center Advisory Committees of our institution. All subjects were within 10% of ideal body weight and had unremarkable clinical histories and normal physical examinations. They were nonsmokers, were not taking any medication, and denied the use of any (including recreational) drugs. They had normal biochemical indices of renal, hepatic, and hematological function, including partial thromboplastin time and normal fasting concentrations of glucose, T4, and TSH. 30-s sampling of GH A Kowarski-Cormed (ML6-5, Medina, New York, NY) blood withdrawal pump (maximal pump flow, 525 mL/24 h, which was reduced to 67%) was used to remove blood samples through a thrombo-resistant tubing (length, 3 ft) and needle set, which was inserted into a dorsal hand vein of the volunteers (13). Small systemic doses of heparin (50 U/kg BW every 4 h) were administered to prevent clotting of blood in the catheter (Kowarski, A., personal communication). The Cormed pump was connected to a fraction collector (LKB, Uppsala, Sweden) with a drop-counting device positioned beside the bed under a hood for noise reduction. The total amount of blood needed for nearly 1000 GH determinations in an 8-h period was less than 120 mL. No significant decrease in hematocrit was observed. The volunteers were discharged after normalization of the partial thromboplastin time at the conclusion of the study. Study protocol The subjects were admitted to the University of Virginia Clinical Research Center for 2 consecutive nights. During the first night, the volunteers became accustomed to polygraphic recording and the restraint and noise generated by the blood withdrawal pump and fraction collector. The subjects were fasted for 24 h before the second night, during which the actual blood withdrawal and polygraph recordings were performed. The subjects were fasted for 24 h to ensure that all serum GH concentrations would be detectable and that any immediate suppressive effect of food as a confounding variable would be removed. Fasting was verified by measurement of urinary ketones.
GH determination in whole blood A modification of the Tandem-R hGH immunoradiometric assay (Hybritech, San Diego, CA) was found suitable for this technique. Two drops of whole blood (equivalent to 120 nL) were collected into each assay tube precoated with 1 U heparin in 50 /iL. Tubes were covered with Parafilm (American National Can Co., Greenwich, CT), stored at 4 C until the end of the collection period, and then assayed immediately for GH using the procedure recommended for serum samples. Experiments with GH-free serum (to which different amounts of GH were added) were used to establish the accuracy of this test. GH measurements in whole blood were corrected for the individual hematocrit of the volunteers. In addition, the time delay between removal of blood from the circulation and arrival in the assay tube was taken into consideration before GH concentrations were correlated with the corresponding sleep stage. Data analysis Pulse analysis. An objective, statistically based pulse detection algorithm, Cluster, was used to define significant pulses in the data (14). Consecutive 30-s data points were used as pseudoreplicates with 1-min intervals for the Cluster analysis. The optimal t statistics and test sample nadir and peak configuration for Cluster analysis were chosen based on computer simulations of synthetic GH concentration-time series generated by a multiple parameter convolution model (15). For the 1-min data, a 6 x 6 test sample nadir and peak with t statistics of 2.0/ 2.0 for up-stroke/down-stroke were used. To study the effect of sampling frequency, data points were omitted to simulate series with less frequent GH measurements. To keep both the false positive and the false negative peak detection error rate nearly constant, the following Cluster parameters were used: 2min data, 4 X 4 test sample nadir and peak, t = 2.5/2.5; 5-min data, 2 X 2, t = 3.0/3.0; 10- or 15-min data, 2 X 2, t = 1.0/1.0; 20- or 30-min data, 1 x 1, t = 1.0/1.0. Deconvolution analysis. Deconvolution analysis was used to calculate GH secretion rates associated with each 30-s observation by removing mathematically the effect of metabolic clearance (16). A waveform-independent deconvolution method (17) was employed, assuming a two-component GH disappearance model consisting of a distributional half-life of 3.5 ± 0.7 min and a second component half-life of 20.7 ± 0.7 min. The
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 19 July 2015. at 05:21 For personal use only. No other uses without permission. . All rights reserved.
HOLL ET AL.
856
latter had a fractional relationship of 0.63 to the total amplitude (18). Dose-dependent assay variance was modeled as a power function based on 4751 pseudoreplicates. Statistical analysis. Multifactorial analysis of variance and linear regression analysis were performed using the SAS statistical software package. P < 0.05 was considered significant. Crosscorrelations were carried out to relate sample GH concentrations or secretion rates with digitized sleep state considered at various lags (time in minutes separating the two measures of interest) (19).
JCE & M • 1991 Vol 72 • No 4
CLUSTER SLEEP STAGE
Results Evaluation of GH immunoradiometric determination in drops of whole blood In four subjects, both serum samples and whole blood measurements via the withdrawal pump were obtained in parallel from separate bilateral venipuncture sites before and after the administration of GHRH (Fig. 1). These GH measurements were highly correlated (r = 0.98). The median intraassay coefficients of variation determined from 90 replicates of heparinized pools of whole blood were 11% and 16.9% at corrected GH concentrations of 12.9 and 18.9 Mg/L, respectively. This variability reflects both the variation in drop size as well as the variability of the assay. There was no significant streaming during the transport of liquid in the tubing, as demonstrated by stepwise addition of [125I]GH to a proteincontaining test liquid. Nocturnal pattern of GH release Figure 2 shows the nocturnal profiles of GH concentrations in 2 subjects. By visual inspection, between 1-3
GHRH-TEST 30 — :WHOL£ BLOOD D : SERUM
0
30
60
90
120
150
1B0
210
240
270
TUBE NUMBER FIG. 1. Comparison between serial GH concentrations in samples obtained every 30 s and assayed with the whole blood method (—) and conventional serum samples (D) withdrawn every 20 min during a GHRH test (1 ^g/kg BW as an iv bolus) in one subject.
CLUSTER SLEEP STAGE
2300
0100
0300
0500
CLOCKTIME FIG. 2. Two examples of nocturnal pulsatile GH concentration profiles in fasting male subjects. The top line identifies significant GH pulses, as detected by Cluster analysis. The second line depicts sleep stages, while the bottom curve represents the individual GH determinations.
major secretory episodes were observed during the collection period, each composed of several smaller peaks. For quantitation, the objective pulse detection algorithm Cluster was applied (14). Since the volume of blood withdrawn for each sample precluded duplicate measurements of GH, consecutive samples were considered to be pseudoduplicates with 1-min intervals for purposes of Cluster analysis. Using appropriate parameters for cluster size and t statistics (see Materials and Methods), a total of 47 spontaneous GH pulses were identified in the data from all 6 volunteers. As the total duration of sampling was 39.8 h (not including the GHRH tests), the mean nocturnal pulse frequency in fasting men was estimated at 1.2 pulses/h, with a range between 0.6-1.8 in individual subjects. Pertinent attributes of pulsatile GH release are listed in Table 1. Relationship between blood sampling frequency and the number of detectable GH pulses The effect of sampling intensity on the apparent GH pulse frequency was determined by progressively deleting values from the parent series (preserving the pseudoduplicates, as explained above), thus creating series with
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 19 July 2015. at 05:21 For personal use only. No other uses without permission. . All rights reserved.
30-s SAMPLING OF GH IN MAN
857
TABLE 1. Mean parameters of pulsatile GH release in normal men studied overnight using 1-min pseudoreplicates Mean no. of peaks/h Mean maximal peak ht (/ug/L) Mean peak width (min) Mean peak area (min/^g-L) Mean % increase Interpeak valley duration (min) Mean valley cone. (Mg/L)
innniiinrLrn_
1.2 11 ± 1.6 30.9 ± 4.1 150 ± 35 330 ± 170 14.8 ± 3.5
min
2 min
5.1 ± 1.1
5 min
10 min
GH pulse parameters at night in fasting youth adult men, as defined by Cluster analysis on data with a 1-min sampling interval. Data are the mean ± SEM (n = 6 men).
2-, 5-, 10-, 15-, 20-, 30-, and 45-min intervals. The test cluster sizes and t statistics for Cluster analysis were optimized for each sampling interval to result in similar sensitivities, as described previously (15). The profiles of these series and the corresponding significant pulses detected by Cluster analysis from subject 1 are shown in Fig. 3. The mean number of GH pulses per h plotted against the number of samples collected per h is shown in Fig. 4. When the sampling frequency was reduced to 3 or fewer samples/h, the GH pulse frequency was below 0.4 pulses/h. However, when sampling was further intensified, there was a steep increase in the number of pulses resolved up to a frequency of 1.2 pulses/h when 60 GH determinations/h were performed. Deconvolution analysis of 30-s GH data Subject to an assumption of concentration-independent clearance of GH, deconvolution allows all individual sample secretory rates to be calculated from all sample GH concentrations measured peripherally. One example of the 30-s peripheral GH concentrations {upper panel) and calculated 1-min pituitary GH secretion rates (lower panel) is shown in Fig. 5. Correlation of GH secretion with sleep stages
LJ O
innnnr~L
J U LTU I 20 min
15 min
10
O
o X O
is 10 5 0
1
2
3
r 5
0
1 2
3
TIME (HOURS) FIG. 3. Influence of sampling frequency on the pulsatile GH profile and the number of significant GH pulses detected in one subject. The top line identifies significant GH pulses, as detected by Cluster analysis. For 1-min data, consecutive 30-s data points were used as pseudoreplicates in Cluster analysis. Data were then progressively reduced by omitting duplicate values, thus generating sampling intervals of 2, 5, 10, 15, 20, 30, and 45 min. 01
o X
or LJ Q. (/) LJ ID QL
Multifactorial analysis of variance was performed on a total of 4751 pairs of GH determinations and corresponding sleep stages (Fig. 6, top panel). The mean GH concentrations during sleep stages 0 (awake; 4.09 ± 0.15 /ug/L; mean ± SE), 1 (4.91 ± 0.29 Mg/L), and rapid eye movement (REM) sleep (2.69 ± 0.12 /xg/L) were not different. However, a 3-fold increase in GH levels was encountered during A sleep (12.92 ± 0.76 ixg/h in stage 3 and 11.87 ± 0.56 ixg/h in stage 4; P < 0.0001). Mean GH concentrations during stage 2 sleep (3.07 ± 0.08 ng/ L) differed significantly from those during stage 1 (P < 0.05) and stages 3 and 4 (P < 0.0001) as well as those during REM sleep (P < 0.0005). GH secretion rates calculated by deconvolution analysis were 4-fold greater
LJ CD
10
20
30
40
50
60
70
NUMBER OF SAMPLES PER HOUR FIG. 4. Influence of sampling intensity on the number of GH pulses per h detected by Cluster analysis. Shown are pooled data from six volunteers. The data were evaluated as described in Fig. 3.
during stages 3 and 4 sleep compared to those during stage 0,1, and 2 sleep (Fig. 6, bottom panel). It is proposed that the EEG event (slow wave sleep or sleep stages 3 and 4) should precede both the pituitary
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 19 July 2015. at 05:21 For personal use only. No other uses without permission. . All rights reserved.
JCE & M • 1991 Vol 72 • No 4
HOLL ET AL.
858
20
peripheral GH concentration 15-
;,
*
11 • 1 1 1
pituitary pituitary GH GH secretion secretion
I" 2300
0100
0300
0500
CLOCKTIME FIG. 5. Peripheral plasma GH concentration profile (top panel) and corresponding pituitary GH secretion rate calculated by deconvolution analysis (bottom panel) in one fasting volunteer sampled every 30 s overnight. Lv, Liter of distribution volume.
GH secretory event as well as the peripheral GH concentration pulse, if cortical or midbrain events (monitored here by EEG recordings) influence hypothalamic neurons to increase GHRH release or decrease somatostatin (SRIH) release (or both) into the hypophy seal-portal circulation, and thus increase GH secretion by somatotropes. We, therefore, lagged either the peripheral GH concentrations or the calculated GH secretion rates with respect to the EEG data and examined the resulting correlation with sleep stages (Fig. 7). For pituitary secretion, the closest correlation was found when sample GH secretory rates lagged the EEG recordings by 4.5 min. For the GH concentrations measured in the circulation, the correlation was maximal, with a time lag of 16 min. Discussion Circulating GH concentrations arise from a complex pattern of pulsatile hormone release by the pituitary gland influenced by gonadal steroids, nutrition, body composition, sleep, and possibly additional factors (20, 21). This pattern is due to an interplay between the stimulation by GHRH and the inhibition by SRIH of pituitary GH secretion. Based upon studies in rats using GHRH administration with and without SRIH antiserum (22), as well as direct measurements of GHRH
o
" ••-!!! AWAKE
1
REM
2
FlG. 6. Mean plasma GH concentrations (top panel) and mean GH secretion rates calculated by deconvolution analysis [bottom panel) according to sleep stages. Using Duncan's multiple range test, all groups except the following were different (P < 0.05): awake vs. stage 1 for GH concentration, and awake vs. stage 1 as well as awake vs. REM for pituitary secretion rates. Lv, Liter of distribution volume.
0.6 j O
0.5
* •
A
: peripheral GH concentration
• : pituitary GH secretion
0.4
8
0.3
5 UA
4.5 mtn
0.1
8 °-°
16 min 30
20
10
0
-10
TIMESHIFT (MINUTES) FIG. 7. Correlation between sleep stage and pituitary GH secretion (•) or plasma GH concentrations (A) dependent on the time interval by which the two events are shifted. A positive lag (left) indicates that hormone values are shifted later than the EEG findings, and a negative lag (right) means that hormone values are earlier compared to the EEG. The time lag corresponding to the maximal correlation coefficient is noted for each analysis.
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 19 July 2015. at 05:21 For personal use only. No other uses without permission. . All rights reserved.
30-s SAMPLING OF GH IN MAN and SRIH in the hypophyseal-portal blood of rats and sheep (23, 24), a model for the interaction between these two hypothalamic regulators at the level of pituitary somatotropes has been proposed (22). Simultaneous removal of SRIH inhibition and stimulation by GHRH are thought to trigger GH secretion from the pituitary and generate peripheral GH pulses. Newer studies have modified this model to account for the more complex patterns of GH release that are evident with more frequent blood sampling (25, 26). As a major (37%) component half-life of GH in the circulation is about 2-5 min (18), it is not surprising that earlier studies, using sampling every 20 min or at greater intervals, missed a significant number of GH pulses. When the sampling interval is decreased to every 5 min, an increased number of GH pulses is detected (1, 25). These findings suggested that high frequency pulsatile GH release occurs within major GH secretory episodes and raised the possibility that even more rapid sampling would further increase the number of detectable GH pulses. For conventional radioimmunological determinations of serum GH concentrations, approximately 1 mL whole blood/determination is needed. The possibility of obtaining blood samples at intervals more frequent than 5 min for extended periods of time is, therefore, limited. We present here a novel approach for GH studies in man; rather than drawing blood to prepare serum or plasma which is then used in the GH assay, drops of heparinized whole blood are used directly in the assay system. The erythrocytes do not interfere with the GH assay employed in this study, and the hematocrit and drop size allow calculation of the precise amount of plasma added to the assay tube. This assumes that the rheological properties of blood do not change during the study, so that the drop volume is constant. Previous tests have shown that not all commercially available GH immunoradiometric assay kits are applicable to this procedure. RIA kits were considered less suitable for this approach due to their lower sensitivity and higher variability. The technique of measuring GH in drops of heparinized whole blood was developed by Clark and Robinson to study GH secretion at frequent intervals in chronically cannulated conscious rats (2, 26). Our study demonstrates that this technique can be applied to man, and the results correlate closely with conventional measurements of GH in serum or plasma. It is, therefore, a useful technique for rapid blood sampling over extended periods of time. When an objective pulse detection algorithm was applied to our data, between 4-12 pulses were observed in each of the volunteers between 2200-0600 h. When the total number of pulses in the six fasted male subjects was divided by the total sampling time, the mean hourly nocturnal GH pulse frequency was calculated as 1.2 pulses/h, a number considerably higher than previous
859
reports. By comparison, 7.3 pulses/24 h were reported in men fasted for 1 day and sampled every 20 min (3), while 5.9 pulses/24 h were observed in unfasted men, sampled every 5 min (1). However, nocturnal GH pulse frequencies are not equivalent to 24-h pulse frequencies due to the predominantly nocturnal release of GH. By dropping data points, we increased the apparent sampling interval. There were no major differences in the number of GH pulses detected when samples were obtained every 15, 20, 30, or 45 min, ranging between 0.3-0.4 pulses/h. However, a further reduction of the sampling interval to 10, 5, 2, or 1 min significantly increased the number of detectable pulses to 0.6, 0.7, 0.8, and 1.2 pulses/h, respectively. The apparent nonlinearity of the relationship between observed GH pulse frequency and sample density (number of samples per h) shown in Fig. 4 is consistent with the findings of previous simulation studies in which an increased prevalence of true signal in the data tended to quench false positive errors (27). The individual profiles reveal that GH is released predominantly in major secretory episodes, which usually last more than an hour and are, therefore, detectable with less frequent sampling. We observed that these large pulses are composed of several smaller bursts of secretion, which are resolved only with high frequency sampling. In our study we did not reach a plateau of detected pulses, so it is possible that decreasing the sampling interval further, if feasible, would result in the detection of an even higher number of GH pulses. GH release is increased during the night. This may be partially due to the lack of inhibition by nutrients, since even short periods of fasting are sufficient to increase GH secretion (3, 4, 28). In addition, various studies, using 20-min sampling of GH, reported a relationship between increased nocturnal GH concentrations and slow wave sleep or sleep stages 3 and 4 (10-12), although this finding was not confirmed by others (29). Further circumstantial evidence for a role of cortical activity (as measured by EEG frequency and scored as sleep stages) in influencing GH secretion comes from experimental sleep delay. For example, in the jet-lag model, the timing of the major nocturnal GH peak is shifted in such a way that the association of GH release and slow wave sleep is maintained. In these studies, sleep delay enhanced the total amount of GH released; this correlated negatively with the total duration of REM sleep (30). In temporally isolated (free-running) man, both the duration of slow wave sleep and the amount of GH secreted are reduced (31). Furthermore, GH responses to GHRH are augmented during slow wave sleep compared to those in the waking state or during REM sleep, regardless of whether the time of sleep onset is normal or delayed (32). These results suggest that slow wave sleep is associated with decreased hypothalamic somatostatin secretion, which
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 19 July 2015. at 05:21 For personal use only. No other uses without permission. . All rights reserved.
HOLL ET AL.
860
facilitates the GH response to GHRH. In patients with depression, alterations have been found both in sleep pattern and GH release (33). Drugs that affect sleep also affect GH release (7). Thus, GH release is influenced by sleep, with differential effects associated with specific sleep stages. Using high frequency sampling, we investigated whether smaller GH secretory episodes thus resolved would coincide with specific sleep stages. The use of deconvolution analysis allowed us to relate not only peripheral hormone concentrations, but also the apparent time of pituitary hormone secretion, with cortical activity. A sleep was positively and REM sleep negatively correlated with peripheral GH concentrations and pituitary GH secretion under these conditions of high temporal resolution. On the assumption that altered cortical activity is transmitted through mediating centers to the hypothalamus, and that such centers then stimulate pituitary hormone release, we were able to estimate the apparent maximal time delay between these events. We suggest that the close temporal correlation between EEG stages and GH secretion (4.5-min time delay) reflects a causal relationship between cortical activity and hypothalamic secretion of SRIH and GHRH; however, we cannot exclude that other centers affect both cortical and hypothalamic activity, and thus, these relationships may only be coincidental. The finding that major episodes of GH release consist of multiple smaller pulses is consistent with a model of multiple GHRH pulses within a period of decreased SRIH secretion, as previously hypothesized (1, 25, 26). The physiological significance of these high frequency, low amplitude GH secretory bursts has not been ascertained, but they may reflect the frequency of hypothalamic GHRH secretion (25). Acknowledgments We thank Dr. A. Kowarski (Baltimore, MD) and Dr. Robert Blizzard for their support with the withdrawal pump, Dr. Donald Kaiser, Dr. M. L. Johnson, and Mr. David Boyd for help with data analysis, and Ms. Ginger Bauler and Ms. Catherine Kern for excellent technical assistance. We also thank Mrs. Sandra Jackson and the staff from the Clinical Research Center, Ms. Suzan Pezzoli for help with study coordination, and Dr. E. Heinze (Ulm) for helpful discussions. GHRH-(1-40)OH was kindly provided by Drs. W. Vale and J. Rivier from the Clayton Foundation Laboratories for Peptide Biology, The Salk Institute (San Diego, CA). References 1. Evans WS, Faria ACS, Christiansen E, et al. Impact of intensive venous sampling on characterization of pulsatile GH release. Am J Physiol. 1987;252:E549-56. 2. Clark RG, Chambers G, Lewin L, Robinson ICAF. Automated repetitive microsampling of blood: growth hormone profiles in conscious male rats. J Endocrinol. 1986;lll:27-35.
JCE & M • 1991 Vol 72 • No 4
3. Ho K, Veldhuis JD, Johnson ML, et al. Fasting enhances growth hormone secretion and amplifies the complex rhythm of growth hormone secretion in man. J Clin Invest. 1988;81:968-75. 4. Hartman ML, Veldhuis JD, Thorner MO. Augmented growth hormone (GH) secretory burst frequency and amplitude mediate enhanced GH secretion during a two-day fast in normal men [Abstract] p. 13. 2nd Int Pituitary Congr; 1989. 5. Goldsmith SJ, Glick SM. Rhythmicity of human growth hormone secretion. Mt Sinai J Med. 1970:37:501-9. 6. Hunter WM, Friend JAR, Strong JA. The diurnal pattern of plasma growth hormone concentration in adults. J Endocrinol. 1966;34:139-46. 7. Takahashi Y, Kipnis M, Daughaday WH. Growth hormone secretion during sleep. J Clin Invest. 1968;47:2079-90. 8. King JM, Price DA. Sleep-induced growth hormone release-evaluation of a simple test for clinical use. Arch Dis Child. 1983;58:2202. 9. Hindmarsh PC, Smith PJ, Taylor BJ, Pringle K, Brook CGD. Comparison between a physiological and a pharmacological stimulus of growth hormone secretion: response to stage IV sleep and insulin-induced hypoglycemia. Lancet. 1985;1033-5. 10. Eastman CJ, Lazarus L. Growth hormone release during sleep in growth retarded children. Arch Dis Child. 1973;48:502-7. 11. Illig R, Stahl M, Henrichs I, Hecker A. Growth hormone release during slow-wave sleep: comparison with insulin and arginine provocation in children with small stature. Helv Paed Acta. 1971;26:665-72. 12. Honda Y, Takahashi K, Takahashi S, et al. Growth hormone secretion during nocturnal sleep in normal subjects. J Clin Endocrinol Metab. 1969;29:20-9. 13. Kowarski A, Thompson RG, Migeon CJ, Blizzard RM. Determination of integrated plasma concentrations and true secretion rates of human growth hormone. J Clin Endocrinol Metab. 1971;32:35660. 14. Veldhuis JD, Johnson ML. Cluster analysis: a simple, versatile and robust algorithm for endocrine pulse detection. Am J Physiol. 1986;250:E486-93. 15. Veldhuis JD, Johnson ML. A novel general biophysical model for simulating episodic endocrine gland signaling. Am J Physiol. 1988;255:E749-59. 16. Veldhuis JD, Carlson ML, Johnson ML. The pituitary gland secretes in bursts: appraising the nature of glandular secretory impulses by simultaneous multiple-parameter deconvolution of plasma hormone concentrations. Proc Natl Acad Sci USA. 1987;84:7686-90. 17. Johnson ML, Lassiter AE, Veldhuis JD. A wave form-independent deconvolution technique to analyze in vivo hormone secretion. Proc of the 72nd Annual Meet of The Endocrine Soc. 1990;549. 18. Faria ACS, Veldhuis JD, Thorner MO, Vance ML. Half-time of endogenous growth hormone (GH) disappearance in normal man after stimulation of GH secretion by GH-releasing hormone and suppression with somatostatin. J Clin Endocrinol Metab. 1989;68:535-41. 19. Veldhuis JD, Johnson ML, Dufau ML. Preferential release of bioactive luteinizing hormone in response to endogenous and low dose exogenous gonadotropin-releasing hormone pulses in man. J Clin Endocrinol Metab. 1987;64:1275-82. 20. Reichlin S. Neuroendocrinology. In: Wilson JD, Foster DW, eds. Williams' textbook of endocrinology, 7th ed. Philadelphia: Saunders; 1985;492-567. 21. Thorner MO, Vance ML, Evans WS, et al. Physiological and clinical studies on GRF and GH. Recent Prog Horm Res. 1986;42:589-640. 22. Tannenbaum GS, Ling N. The interrelationship of growth hormone (GH)-releasing factor and somatostatin in generation of the ultradian rhythm of GH secretion. Endocrinology. 1984;115:19527. 23. Plotsky PM, Vale W. Patterns of growth hormone-releasing factor and somatostatin: secretion into the hypophyseal-portal circulation of the rat. Science. 1985;230:461-3. 24. Thomas GB, Cummins JT, Clarke IJ. Secretion of growth hormone-releasing factor and somatostatin in nutritionally restricted
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 19 July 2015. at 05:21 For personal use only. No other uses without permission. . All rights reserved.
30-s SAMPLING OF GH IN MAN
25. 26. 27. 28.
29.
sheep. Proc of the 72nd Annual Meet of The Endocrine Soc. 1990;521. Hartman ML, Faria ACS, Vance ML, Johnson ML, Thorner MO, Veldhuis JD. Temporal structure of in vivo growth hormone secretory events in man. Am J Physiol. In Press. Clark RG, Carlsson LMS, Robinson ICAF. Growth hormone secretory profiles in conscious female rats. J Endocrinol. 1987; 114:399407. Urban RJ, Johnson ML, Veldhuis JD. Biophysical modeling of sensitivity and positive accuracy of detecting episodic endocrine signals. Am J Physiol. 1989;257:E88-94. Berelowitz M, Szabo M, Frohman LA, Firestone S, Chu L. Somatomedin-C mediates growth hormone negative feedback by effects on both the hypothalamus and pituitary. Science. 1981;212:127981. Jarrett DB, Greenhouse JB, Miewald JM, Fedorka IB, Kupfer DJ.
30. 31. 32.
33.
861
A reexamination of the relationship between growth hormone secretion and slow wave sleep using delta wave analysis. Biol Psychiatry. 1990;27:497-509. Golstein J, Van Cauter E, Desir D, et al. Effects of jet lag on hormonal patterns. IV. Time shifts increase growth hormone release. J Clin Endocrinol Metab. 1983;56:433-40. Moline ML, Monk TH, Wagner DR, et al. Human growth hormone release is decreased during sleep in temporal isolation (free-running). Chronobiologia. 1986;13:13-9. Van Cauter E, Kerkofs M, Van Underbergen A, Caufriez A, Copinschi G. Modulation of spontaneous and GHRH-stimulated GH secretion by sleep. Proc of the 71st Annual Meet of The Endocrine Soc. 1989;792. Mendlewicz J, Linkowski P, Kerkhofs M, et al. Diurnal hypersecretion of growth hormone in depression. J Clin Endocrinol Metab. 1985;60:505-12.
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 19 July 2015. at 05:21 For personal use only. No other uses without permission. . All rights reserved.
Vol. 72, No. 4 Printed in U.S.A.
Thirty-Second Sampling of Plasma Growth Hormone in Man: Correlation with Sleep Stages* REINHARD W. HOLLf, MARK L. HARTMAN, JOHANNES D. VELDHUIS, WILLIAM M. TAYLOR, AND MICHAEL 0. THORNER Division of Endocrinology and Metabolism, Department of Medicine, University of Virginia Health Sciences Center, Charlottesville, Virginia 22908
ABSTRACT. Both fasting and sleep increase the secretion of human GH and, therefore, might explain its predominantly nocturnal release. To study the precise temporal relationship between GH secretory episodes and cortical activity, GH measurements and electroencephalogram sleep stage recordings were performed every 30 s for 8 h in six young male volunteers fasted for 24 h. GH was measured in two drops of whole blood, which were directly sampled into the assay tube using a continuous blood withdrawal pump and a fraction collector. Concomitant serum sampling during a GH-releasing hormone test (n = 4) revealed a high correlation (r = 0.98) between GH measurements in serum and whole blood. GH pulses were objectively identified with Cluster analysis, and GH secretion rates were calculated with a waveform-independent deconvolution algorithm. When
data were analyzed as replicates with 1-min intervals, the nocturnal pulse frequency was 1.2 pulses/h. Elimination of data points demonstrated 43% and 64% reductions in the number of GH pulses detected for 5- and 20-min sampling intervals, respectively. Mean GH concentrations and secretory rates were significantly higher during stage 3 and 4 sleep compared to stage 1, 2, and rapid eye movement sleep. GH secretory rates and peripheral GH concentrations were maximally correlated with sleep stage, with lags of 4.5 and 16 min, respectively, suggesting that maximal GH release occurs within minutes of the onset of stage 3 or 4 sleep. This temporal coincidence between pituitary GH secretion and A sleep is consistent with cortical control over hypothalamic-pituitary function. (J Clin Endocrinol Metab 72: 854-861, 1991)
T
tions. Rather than processing blood to serum or plasma, which is then assayed for the hormone, we measured GH directly in drops of heparinized whole blood. This technique, which was first applied in studies of GH pulsatility in conscious rats (2), allowed us to determine circulating GH concentrations in man at 30-s intervals. Spontaneous release of GH is augmented within 1-2 days of fasting. In particular, pulse frequency and pulse amplitude are both increased (3, 4). The confounding influences of different nutrients on the rhythmicity of GH release are avoided when fasting subjects are studied (5). It has long been recognized that GH release is also increased during the night (6, 7). Thus, nocturnal blood sampling for GH has been proposed as a diagnostic alternative to pharmacological testing of GH secretory capacity (8, 9). The physiological basis for the predominantly nocturnal release of GH is still not completely understood. Both metabolic effects (reduced suppression of GH release by decreased nutritional intake) as well as reduced inhibitory brain activity during sleep have been considered. Previous studies with relatively infrequent sampling in a limited number of subjects found a relationship between the onset of slow wave sleep and the release of GH (10-12). However, the exact relationship between sleep and pituitary hormone secretion remains
HERE has been considerable interest in the study of spontaneous pulsatile GH release in recent years. However, it is still a matter of debate how frequently blood must be sampled for optimal detection of serum GH pulses. While a sampling interval of 20 min is the most widely used, a recent study demonstrated that sampling every 5 min detects a significantly larger number of pulses compared to sampling at 10- or 20-min intervals (1). It is impractical to further increase the frequency at which conventional serum samples are drawn. Therefore, we developed a different approach to enable more frequent measurements of GH concentra-
Received June 22,1990. Address all correspondence and requests for reprints to: Dr. Michael O. Thorner, Kenneth R. Crispell Professor of Medicine, Division of Endocrinology and Metabolism, Box 511, University of Virginia Health Sciences Center, Charlottesville, Virginia 22908. * Presented in part at the 71st Annual Meeting of The Endocrine Society, Seattle, WA, 1989. This work was supported by NIH Grants DK-32632 (to M.O.T.), Research Career Development Awards 1K04HD-00634 and HD-16806 (to J.D.V.), Grant KO8-HD-00860 (to M.L.H.), Diabetes and Endocrinology Research Center Grant DK38942, Grant RR-0847 (to University of Virginia General Clinical Research Center, including CLINFO), the Pratt Foundation, and the University of Virginia Academic Enhancement Fund (to the Biodynamics Institute). t Supported by DFG (Ho 1041). Present address: Department of Pediatrics I, University of Ulm Medical Center, Ulm/Donau, Germany.
854
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 19 July 2015. at 05:21 For personal use only. No other uses without permission. . All rights reserved.
30-s SAMPLING OF GH IN MAN
to be fully understood. We, therefore, studied fasting male volunteers during sleep by recording the stages of sleep and measuring GH concentrations every 30 s. The aim of this study was to answer the following questions. 1) Does an increase in sampling frequency result in enhanced identification of nocturnal GH pulses? 2) Can the presumptive relation between GH release and slow wave sleep be confirmed by data using high frequency sampling and deconvolution-based estimates of GH secretion?
Subjects and Methods
855
On the night of the study, blood withdrawal via the pump was started at 2200 h. In addition, a heparin lock was inserted in an arm vein opposite the pump and electrodes were attached for the recording of electroencephalogram (EEG), electromyogram, electrooculogram, and electrocardiogram. These parameters were recorded continuously by instruments in a separate room. During the night, the volunteers were monitored by microphone as well as optically with an infrared camera. The room was entered only when collection tubes were exchanged (every 2 h). At 0600 h the subjects were awakened, and an iv bolus of human GH-releasing hormone [GHRH-(1-40)-OH; 1 /xg/kg BW] was administered. For 2 h thereafter, blood was drawn both continuously through the pump to allow 30-s sampling as well as through a separate heparin lock every 20 min.
Subjects Six healthy male volunteers, aged 21-30 yr, were studied after giving informed written consent. The study protocol was approved by the Human Investigation and General Clinical Research Center Advisory Committees of our institution. All subjects were within 10% of ideal body weight and had unremarkable clinical histories and normal physical examinations. They were nonsmokers, were not taking any medication, and denied the use of any (including recreational) drugs. They had normal biochemical indices of renal, hepatic, and hematological function, including partial thromboplastin time and normal fasting concentrations of glucose, T4, and TSH. 30-s sampling of GH A Kowarski-Cormed (ML6-5, Medina, New York, NY) blood withdrawal pump (maximal pump flow, 525 mL/24 h, which was reduced to 67%) was used to remove blood samples through a thrombo-resistant tubing (length, 3 ft) and needle set, which was inserted into a dorsal hand vein of the volunteers (13). Small systemic doses of heparin (50 U/kg BW every 4 h) were administered to prevent clotting of blood in the catheter (Kowarski, A., personal communication). The Cormed pump was connected to a fraction collector (LKB, Uppsala, Sweden) with a drop-counting device positioned beside the bed under a hood for noise reduction. The total amount of blood needed for nearly 1000 GH determinations in an 8-h period was less than 120 mL. No significant decrease in hematocrit was observed. The volunteers were discharged after normalization of the partial thromboplastin time at the conclusion of the study. Study protocol The subjects were admitted to the University of Virginia Clinical Research Center for 2 consecutive nights. During the first night, the volunteers became accustomed to polygraphic recording and the restraint and noise generated by the blood withdrawal pump and fraction collector. The subjects were fasted for 24 h before the second night, during which the actual blood withdrawal and polygraph recordings were performed. The subjects were fasted for 24 h to ensure that all serum GH concentrations would be detectable and that any immediate suppressive effect of food as a confounding variable would be removed. Fasting was verified by measurement of urinary ketones.
GH determination in whole blood A modification of the Tandem-R hGH immunoradiometric assay (Hybritech, San Diego, CA) was found suitable for this technique. Two drops of whole blood (equivalent to 120 nL) were collected into each assay tube precoated with 1 U heparin in 50 /iL. Tubes were covered with Parafilm (American National Can Co., Greenwich, CT), stored at 4 C until the end of the collection period, and then assayed immediately for GH using the procedure recommended for serum samples. Experiments with GH-free serum (to which different amounts of GH were added) were used to establish the accuracy of this test. GH measurements in whole blood were corrected for the individual hematocrit of the volunteers. In addition, the time delay between removal of blood from the circulation and arrival in the assay tube was taken into consideration before GH concentrations were correlated with the corresponding sleep stage. Data analysis Pulse analysis. An objective, statistically based pulse detection algorithm, Cluster, was used to define significant pulses in the data (14). Consecutive 30-s data points were used as pseudoreplicates with 1-min intervals for the Cluster analysis. The optimal t statistics and test sample nadir and peak configuration for Cluster analysis were chosen based on computer simulations of synthetic GH concentration-time series generated by a multiple parameter convolution model (15). For the 1-min data, a 6 x 6 test sample nadir and peak with t statistics of 2.0/ 2.0 for up-stroke/down-stroke were used. To study the effect of sampling frequency, data points were omitted to simulate series with less frequent GH measurements. To keep both the false positive and the false negative peak detection error rate nearly constant, the following Cluster parameters were used: 2min data, 4 X 4 test sample nadir and peak, t = 2.5/2.5; 5-min data, 2 X 2, t = 3.0/3.0; 10- or 15-min data, 2 X 2, t = 1.0/1.0; 20- or 30-min data, 1 x 1, t = 1.0/1.0. Deconvolution analysis. Deconvolution analysis was used to calculate GH secretion rates associated with each 30-s observation by removing mathematically the effect of metabolic clearance (16). A waveform-independent deconvolution method (17) was employed, assuming a two-component GH disappearance model consisting of a distributional half-life of 3.5 ± 0.7 min and a second component half-life of 20.7 ± 0.7 min. The
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 19 July 2015. at 05:21 For personal use only. No other uses without permission. . All rights reserved.
HOLL ET AL.
856
latter had a fractional relationship of 0.63 to the total amplitude (18). Dose-dependent assay variance was modeled as a power function based on 4751 pseudoreplicates. Statistical analysis. Multifactorial analysis of variance and linear regression analysis were performed using the SAS statistical software package. P < 0.05 was considered significant. Crosscorrelations were carried out to relate sample GH concentrations or secretion rates with digitized sleep state considered at various lags (time in minutes separating the two measures of interest) (19).
JCE & M • 1991 Vol 72 • No 4
CLUSTER SLEEP STAGE
Results Evaluation of GH immunoradiometric determination in drops of whole blood In four subjects, both serum samples and whole blood measurements via the withdrawal pump were obtained in parallel from separate bilateral venipuncture sites before and after the administration of GHRH (Fig. 1). These GH measurements were highly correlated (r = 0.98). The median intraassay coefficients of variation determined from 90 replicates of heparinized pools of whole blood were 11% and 16.9% at corrected GH concentrations of 12.9 and 18.9 Mg/L, respectively. This variability reflects both the variation in drop size as well as the variability of the assay. There was no significant streaming during the transport of liquid in the tubing, as demonstrated by stepwise addition of [125I]GH to a proteincontaining test liquid. Nocturnal pattern of GH release Figure 2 shows the nocturnal profiles of GH concentrations in 2 subjects. By visual inspection, between 1-3
GHRH-TEST 30 — :WHOL£ BLOOD D : SERUM
0
30
60
90
120
150
1B0
210
240
270
TUBE NUMBER FIG. 1. Comparison between serial GH concentrations in samples obtained every 30 s and assayed with the whole blood method (—) and conventional serum samples (D) withdrawn every 20 min during a GHRH test (1 ^g/kg BW as an iv bolus) in one subject.
CLUSTER SLEEP STAGE
2300
0100
0300
0500
CLOCKTIME FIG. 2. Two examples of nocturnal pulsatile GH concentration profiles in fasting male subjects. The top line identifies significant GH pulses, as detected by Cluster analysis. The second line depicts sleep stages, while the bottom curve represents the individual GH determinations.
major secretory episodes were observed during the collection period, each composed of several smaller peaks. For quantitation, the objective pulse detection algorithm Cluster was applied (14). Since the volume of blood withdrawn for each sample precluded duplicate measurements of GH, consecutive samples were considered to be pseudoduplicates with 1-min intervals for purposes of Cluster analysis. Using appropriate parameters for cluster size and t statistics (see Materials and Methods), a total of 47 spontaneous GH pulses were identified in the data from all 6 volunteers. As the total duration of sampling was 39.8 h (not including the GHRH tests), the mean nocturnal pulse frequency in fasting men was estimated at 1.2 pulses/h, with a range between 0.6-1.8 in individual subjects. Pertinent attributes of pulsatile GH release are listed in Table 1. Relationship between blood sampling frequency and the number of detectable GH pulses The effect of sampling intensity on the apparent GH pulse frequency was determined by progressively deleting values from the parent series (preserving the pseudoduplicates, as explained above), thus creating series with
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 19 July 2015. at 05:21 For personal use only. No other uses without permission. . All rights reserved.
30-s SAMPLING OF GH IN MAN
857
TABLE 1. Mean parameters of pulsatile GH release in normal men studied overnight using 1-min pseudoreplicates Mean no. of peaks/h Mean maximal peak ht (/ug/L) Mean peak width (min) Mean peak area (min/^g-L) Mean % increase Interpeak valley duration (min) Mean valley cone. (Mg/L)
innniiinrLrn_
1.2 11 ± 1.6 30.9 ± 4.1 150 ± 35 330 ± 170 14.8 ± 3.5
min
2 min
5.1 ± 1.1
5 min
10 min
GH pulse parameters at night in fasting youth adult men, as defined by Cluster analysis on data with a 1-min sampling interval. Data are the mean ± SEM (n = 6 men).
2-, 5-, 10-, 15-, 20-, 30-, and 45-min intervals. The test cluster sizes and t statistics for Cluster analysis were optimized for each sampling interval to result in similar sensitivities, as described previously (15). The profiles of these series and the corresponding significant pulses detected by Cluster analysis from subject 1 are shown in Fig. 3. The mean number of GH pulses per h plotted against the number of samples collected per h is shown in Fig. 4. When the sampling frequency was reduced to 3 or fewer samples/h, the GH pulse frequency was below 0.4 pulses/h. However, when sampling was further intensified, there was a steep increase in the number of pulses resolved up to a frequency of 1.2 pulses/h when 60 GH determinations/h were performed. Deconvolution analysis of 30-s GH data Subject to an assumption of concentration-independent clearance of GH, deconvolution allows all individual sample secretory rates to be calculated from all sample GH concentrations measured peripherally. One example of the 30-s peripheral GH concentrations {upper panel) and calculated 1-min pituitary GH secretion rates (lower panel) is shown in Fig. 5. Correlation of GH secretion with sleep stages
LJ O
innnnr~L
J U LTU I 20 min
15 min
10
O
o X O
is 10 5 0
1
2
3
r 5
0
1 2
3
TIME (HOURS) FIG. 3. Influence of sampling frequency on the pulsatile GH profile and the number of significant GH pulses detected in one subject. The top line identifies significant GH pulses, as detected by Cluster analysis. For 1-min data, consecutive 30-s data points were used as pseudoreplicates in Cluster analysis. Data were then progressively reduced by omitting duplicate values, thus generating sampling intervals of 2, 5, 10, 15, 20, 30, and 45 min. 01
o X
or LJ Q. (/) LJ ID QL
Multifactorial analysis of variance was performed on a total of 4751 pairs of GH determinations and corresponding sleep stages (Fig. 6, top panel). The mean GH concentrations during sleep stages 0 (awake; 4.09 ± 0.15 /ug/L; mean ± SE), 1 (4.91 ± 0.29 Mg/L), and rapid eye movement (REM) sleep (2.69 ± 0.12 /xg/L) were not different. However, a 3-fold increase in GH levels was encountered during A sleep (12.92 ± 0.76 ixg/h in stage 3 and 11.87 ± 0.56 ixg/h in stage 4; P < 0.0001). Mean GH concentrations during stage 2 sleep (3.07 ± 0.08 ng/ L) differed significantly from those during stage 1 (P < 0.05) and stages 3 and 4 (P < 0.0001) as well as those during REM sleep (P < 0.0005). GH secretion rates calculated by deconvolution analysis were 4-fold greater
LJ CD
10
20
30
40
50
60
70
NUMBER OF SAMPLES PER HOUR FIG. 4. Influence of sampling intensity on the number of GH pulses per h detected by Cluster analysis. Shown are pooled data from six volunteers. The data were evaluated as described in Fig. 3.
during stages 3 and 4 sleep compared to those during stage 0,1, and 2 sleep (Fig. 6, bottom panel). It is proposed that the EEG event (slow wave sleep or sleep stages 3 and 4) should precede both the pituitary
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 19 July 2015. at 05:21 For personal use only. No other uses without permission. . All rights reserved.
JCE & M • 1991 Vol 72 • No 4
HOLL ET AL.
858
20
peripheral GH concentration 15-
;,
*
11 • 1 1 1
pituitary pituitary GH GH secretion secretion
I" 2300
0100
0300
0500
CLOCKTIME FIG. 5. Peripheral plasma GH concentration profile (top panel) and corresponding pituitary GH secretion rate calculated by deconvolution analysis (bottom panel) in one fasting volunteer sampled every 30 s overnight. Lv, Liter of distribution volume.
GH secretory event as well as the peripheral GH concentration pulse, if cortical or midbrain events (monitored here by EEG recordings) influence hypothalamic neurons to increase GHRH release or decrease somatostatin (SRIH) release (or both) into the hypophy seal-portal circulation, and thus increase GH secretion by somatotropes. We, therefore, lagged either the peripheral GH concentrations or the calculated GH secretion rates with respect to the EEG data and examined the resulting correlation with sleep stages (Fig. 7). For pituitary secretion, the closest correlation was found when sample GH secretory rates lagged the EEG recordings by 4.5 min. For the GH concentrations measured in the circulation, the correlation was maximal, with a time lag of 16 min. Discussion Circulating GH concentrations arise from a complex pattern of pulsatile hormone release by the pituitary gland influenced by gonadal steroids, nutrition, body composition, sleep, and possibly additional factors (20, 21). This pattern is due to an interplay between the stimulation by GHRH and the inhibition by SRIH of pituitary GH secretion. Based upon studies in rats using GHRH administration with and without SRIH antiserum (22), as well as direct measurements of GHRH
o
" ••-!!! AWAKE
1
REM
2
FlG. 6. Mean plasma GH concentrations (top panel) and mean GH secretion rates calculated by deconvolution analysis [bottom panel) according to sleep stages. Using Duncan's multiple range test, all groups except the following were different (P < 0.05): awake vs. stage 1 for GH concentration, and awake vs. stage 1 as well as awake vs. REM for pituitary secretion rates. Lv, Liter of distribution volume.
0.6 j O
0.5
* •
A
: peripheral GH concentration
• : pituitary GH secretion
0.4
8
0.3
5 UA
4.5 mtn
0.1
8 °-°
16 min 30
20
10
0
-10
TIMESHIFT (MINUTES) FIG. 7. Correlation between sleep stage and pituitary GH secretion (•) or plasma GH concentrations (A) dependent on the time interval by which the two events are shifted. A positive lag (left) indicates that hormone values are shifted later than the EEG findings, and a negative lag (right) means that hormone values are earlier compared to the EEG. The time lag corresponding to the maximal correlation coefficient is noted for each analysis.
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 19 July 2015. at 05:21 For personal use only. No other uses without permission. . All rights reserved.
30-s SAMPLING OF GH IN MAN and SRIH in the hypophyseal-portal blood of rats and sheep (23, 24), a model for the interaction between these two hypothalamic regulators at the level of pituitary somatotropes has been proposed (22). Simultaneous removal of SRIH inhibition and stimulation by GHRH are thought to trigger GH secretion from the pituitary and generate peripheral GH pulses. Newer studies have modified this model to account for the more complex patterns of GH release that are evident with more frequent blood sampling (25, 26). As a major (37%) component half-life of GH in the circulation is about 2-5 min (18), it is not surprising that earlier studies, using sampling every 20 min or at greater intervals, missed a significant number of GH pulses. When the sampling interval is decreased to every 5 min, an increased number of GH pulses is detected (1, 25). These findings suggested that high frequency pulsatile GH release occurs within major GH secretory episodes and raised the possibility that even more rapid sampling would further increase the number of detectable GH pulses. For conventional radioimmunological determinations of serum GH concentrations, approximately 1 mL whole blood/determination is needed. The possibility of obtaining blood samples at intervals more frequent than 5 min for extended periods of time is, therefore, limited. We present here a novel approach for GH studies in man; rather than drawing blood to prepare serum or plasma which is then used in the GH assay, drops of heparinized whole blood are used directly in the assay system. The erythrocytes do not interfere with the GH assay employed in this study, and the hematocrit and drop size allow calculation of the precise amount of plasma added to the assay tube. This assumes that the rheological properties of blood do not change during the study, so that the drop volume is constant. Previous tests have shown that not all commercially available GH immunoradiometric assay kits are applicable to this procedure. RIA kits were considered less suitable for this approach due to their lower sensitivity and higher variability. The technique of measuring GH in drops of heparinized whole blood was developed by Clark and Robinson to study GH secretion at frequent intervals in chronically cannulated conscious rats (2, 26). Our study demonstrates that this technique can be applied to man, and the results correlate closely with conventional measurements of GH in serum or plasma. It is, therefore, a useful technique for rapid blood sampling over extended periods of time. When an objective pulse detection algorithm was applied to our data, between 4-12 pulses were observed in each of the volunteers between 2200-0600 h. When the total number of pulses in the six fasted male subjects was divided by the total sampling time, the mean hourly nocturnal GH pulse frequency was calculated as 1.2 pulses/h, a number considerably higher than previous
859
reports. By comparison, 7.3 pulses/24 h were reported in men fasted for 1 day and sampled every 20 min (3), while 5.9 pulses/24 h were observed in unfasted men, sampled every 5 min (1). However, nocturnal GH pulse frequencies are not equivalent to 24-h pulse frequencies due to the predominantly nocturnal release of GH. By dropping data points, we increased the apparent sampling interval. There were no major differences in the number of GH pulses detected when samples were obtained every 15, 20, 30, or 45 min, ranging between 0.3-0.4 pulses/h. However, a further reduction of the sampling interval to 10, 5, 2, or 1 min significantly increased the number of detectable pulses to 0.6, 0.7, 0.8, and 1.2 pulses/h, respectively. The apparent nonlinearity of the relationship between observed GH pulse frequency and sample density (number of samples per h) shown in Fig. 4 is consistent with the findings of previous simulation studies in which an increased prevalence of true signal in the data tended to quench false positive errors (27). The individual profiles reveal that GH is released predominantly in major secretory episodes, which usually last more than an hour and are, therefore, detectable with less frequent sampling. We observed that these large pulses are composed of several smaller bursts of secretion, which are resolved only with high frequency sampling. In our study we did not reach a plateau of detected pulses, so it is possible that decreasing the sampling interval further, if feasible, would result in the detection of an even higher number of GH pulses. GH release is increased during the night. This may be partially due to the lack of inhibition by nutrients, since even short periods of fasting are sufficient to increase GH secretion (3, 4, 28). In addition, various studies, using 20-min sampling of GH, reported a relationship between increased nocturnal GH concentrations and slow wave sleep or sleep stages 3 and 4 (10-12), although this finding was not confirmed by others (29). Further circumstantial evidence for a role of cortical activity (as measured by EEG frequency and scored as sleep stages) in influencing GH secretion comes from experimental sleep delay. For example, in the jet-lag model, the timing of the major nocturnal GH peak is shifted in such a way that the association of GH release and slow wave sleep is maintained. In these studies, sleep delay enhanced the total amount of GH released; this correlated negatively with the total duration of REM sleep (30). In temporally isolated (free-running) man, both the duration of slow wave sleep and the amount of GH secreted are reduced (31). Furthermore, GH responses to GHRH are augmented during slow wave sleep compared to those in the waking state or during REM sleep, regardless of whether the time of sleep onset is normal or delayed (32). These results suggest that slow wave sleep is associated with decreased hypothalamic somatostatin secretion, which
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 19 July 2015. at 05:21 For personal use only. No other uses without permission. . All rights reserved.
HOLL ET AL.
860
facilitates the GH response to GHRH. In patients with depression, alterations have been found both in sleep pattern and GH release (33). Drugs that affect sleep also affect GH release (7). Thus, GH release is influenced by sleep, with differential effects associated with specific sleep stages. Using high frequency sampling, we investigated whether smaller GH secretory episodes thus resolved would coincide with specific sleep stages. The use of deconvolution analysis allowed us to relate not only peripheral hormone concentrations, but also the apparent time of pituitary hormone secretion, with cortical activity. A sleep was positively and REM sleep negatively correlated with peripheral GH concentrations and pituitary GH secretion under these conditions of high temporal resolution. On the assumption that altered cortical activity is transmitted through mediating centers to the hypothalamus, and that such centers then stimulate pituitary hormone release, we were able to estimate the apparent maximal time delay between these events. We suggest that the close temporal correlation between EEG stages and GH secretion (4.5-min time delay) reflects a causal relationship between cortical activity and hypothalamic secretion of SRIH and GHRH; however, we cannot exclude that other centers affect both cortical and hypothalamic activity, and thus, these relationships may only be coincidental. The finding that major episodes of GH release consist of multiple smaller pulses is consistent with a model of multiple GHRH pulses within a period of decreased SRIH secretion, as previously hypothesized (1, 25, 26). The physiological significance of these high frequency, low amplitude GH secretory bursts has not been ascertained, but they may reflect the frequency of hypothalamic GHRH secretion (25). Acknowledgments We thank Dr. A. Kowarski (Baltimore, MD) and Dr. Robert Blizzard for their support with the withdrawal pump, Dr. Donald Kaiser, Dr. M. L. Johnson, and Mr. David Boyd for help with data analysis, and Ms. Ginger Bauler and Ms. Catherine Kern for excellent technical assistance. We also thank Mrs. Sandra Jackson and the staff from the Clinical Research Center, Ms. Suzan Pezzoli for help with study coordination, and Dr. E. Heinze (Ulm) for helpful discussions. GHRH-(1-40)OH was kindly provided by Drs. W. Vale and J. Rivier from the Clayton Foundation Laboratories for Peptide Biology, The Salk Institute (San Diego, CA). References 1. Evans WS, Faria ACS, Christiansen E, et al. Impact of intensive venous sampling on characterization of pulsatile GH release. Am J Physiol. 1987;252:E549-56. 2. Clark RG, Chambers G, Lewin L, Robinson ICAF. Automated repetitive microsampling of blood: growth hormone profiles in conscious male rats. J Endocrinol. 1986;lll:27-35.
JCE & M • 1991 Vol 72 • No 4
3. Ho K, Veldhuis JD, Johnson ML, et al. Fasting enhances growth hormone secretion and amplifies the complex rhythm of growth hormone secretion in man. J Clin Invest. 1988;81:968-75. 4. Hartman ML, Veldhuis JD, Thorner MO. Augmented growth hormone (GH) secretory burst frequency and amplitude mediate enhanced GH secretion during a two-day fast in normal men [Abstract] p. 13. 2nd Int Pituitary Congr; 1989. 5. Goldsmith SJ, Glick SM. Rhythmicity of human growth hormone secretion. Mt Sinai J Med. 1970:37:501-9. 6. Hunter WM, Friend JAR, Strong JA. The diurnal pattern of plasma growth hormone concentration in adults. J Endocrinol. 1966;34:139-46. 7. Takahashi Y, Kipnis M, Daughaday WH. Growth hormone secretion during sleep. J Clin Invest. 1968;47:2079-90. 8. King JM, Price DA. Sleep-induced growth hormone release-evaluation of a simple test for clinical use. Arch Dis Child. 1983;58:2202. 9. Hindmarsh PC, Smith PJ, Taylor BJ, Pringle K, Brook CGD. Comparison between a physiological and a pharmacological stimulus of growth hormone secretion: response to stage IV sleep and insulin-induced hypoglycemia. Lancet. 1985;1033-5. 10. Eastman CJ, Lazarus L. Growth hormone release during sleep in growth retarded children. Arch Dis Child. 1973;48:502-7. 11. Illig R, Stahl M, Henrichs I, Hecker A. Growth hormone release during slow-wave sleep: comparison with insulin and arginine provocation in children with small stature. Helv Paed Acta. 1971;26:665-72. 12. Honda Y, Takahashi K, Takahashi S, et al. Growth hormone secretion during nocturnal sleep in normal subjects. J Clin Endocrinol Metab. 1969;29:20-9. 13. Kowarski A, Thompson RG, Migeon CJ, Blizzard RM. Determination of integrated plasma concentrations and true secretion rates of human growth hormone. J Clin Endocrinol Metab. 1971;32:35660. 14. Veldhuis JD, Johnson ML. Cluster analysis: a simple, versatile and robust algorithm for endocrine pulse detection. Am J Physiol. 1986;250:E486-93. 15. Veldhuis JD, Johnson ML. A novel general biophysical model for simulating episodic endocrine gland signaling. Am J Physiol. 1988;255:E749-59. 16. Veldhuis JD, Carlson ML, Johnson ML. The pituitary gland secretes in bursts: appraising the nature of glandular secretory impulses by simultaneous multiple-parameter deconvolution of plasma hormone concentrations. Proc Natl Acad Sci USA. 1987;84:7686-90. 17. Johnson ML, Lassiter AE, Veldhuis JD. A wave form-independent deconvolution technique to analyze in vivo hormone secretion. Proc of the 72nd Annual Meet of The Endocrine Soc. 1990;549. 18. Faria ACS, Veldhuis JD, Thorner MO, Vance ML. Half-time of endogenous growth hormone (GH) disappearance in normal man after stimulation of GH secretion by GH-releasing hormone and suppression with somatostatin. J Clin Endocrinol Metab. 1989;68:535-41. 19. Veldhuis JD, Johnson ML, Dufau ML. Preferential release of bioactive luteinizing hormone in response to endogenous and low dose exogenous gonadotropin-releasing hormone pulses in man. J Clin Endocrinol Metab. 1987;64:1275-82. 20. Reichlin S. Neuroendocrinology. In: Wilson JD, Foster DW, eds. Williams' textbook of endocrinology, 7th ed. Philadelphia: Saunders; 1985;492-567. 21. Thorner MO, Vance ML, Evans WS, et al. Physiological and clinical studies on GRF and GH. Recent Prog Horm Res. 1986;42:589-640. 22. Tannenbaum GS, Ling N. The interrelationship of growth hormone (GH)-releasing factor and somatostatin in generation of the ultradian rhythm of GH secretion. Endocrinology. 1984;115:19527. 23. Plotsky PM, Vale W. Patterns of growth hormone-releasing factor and somatostatin: secretion into the hypophyseal-portal circulation of the rat. Science. 1985;230:461-3. 24. Thomas GB, Cummins JT, Clarke IJ. Secretion of growth hormone-releasing factor and somatostatin in nutritionally restricted
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 19 July 2015. at 05:21 For personal use only. No other uses without permission. . All rights reserved.
30-s SAMPLING OF GH IN MAN
25. 26. 27. 28.
29.
sheep. Proc of the 72nd Annual Meet of The Endocrine Soc. 1990;521. Hartman ML, Faria ACS, Vance ML, Johnson ML, Thorner MO, Veldhuis JD. Temporal structure of in vivo growth hormone secretory events in man. Am J Physiol. In Press. Clark RG, Carlsson LMS, Robinson ICAF. Growth hormone secretory profiles in conscious female rats. J Endocrinol. 1987; 114:399407. Urban RJ, Johnson ML, Veldhuis JD. Biophysical modeling of sensitivity and positive accuracy of detecting episodic endocrine signals. Am J Physiol. 1989;257:E88-94. Berelowitz M, Szabo M, Frohman LA, Firestone S, Chu L. Somatomedin-C mediates growth hormone negative feedback by effects on both the hypothalamus and pituitary. Science. 1981;212:127981. Jarrett DB, Greenhouse JB, Miewald JM, Fedorka IB, Kupfer DJ.
30. 31. 32.
33.
861
A reexamination of the relationship between growth hormone secretion and slow wave sleep using delta wave analysis. Biol Psychiatry. 1990;27:497-509. Golstein J, Van Cauter E, Desir D, et al. Effects of jet lag on hormonal patterns. IV. Time shifts increase growth hormone release. J Clin Endocrinol Metab. 1983;56:433-40. Moline ML, Monk TH, Wagner DR, et al. Human growth hormone release is decreased during sleep in temporal isolation (free-running). Chronobiologia. 1986;13:13-9. Van Cauter E, Kerkofs M, Van Underbergen A, Caufriez A, Copinschi G. Modulation of spontaneous and GHRH-stimulated GH secretion by sleep. Proc of the 71st Annual Meet of The Endocrine Soc. 1989;792. Mendlewicz J, Linkowski P, Kerkhofs M, et al. Diurnal hypersecretion of growth hormone in depression. J Clin Endocrinol Metab. 1985;60:505-12.
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 19 July 2015. at 05:21 For personal use only. No other uses without permission. . All rights reserved.
