Nocturnal Increase of Plasma Testosterone in Men: Relation to Gonadotropins and Prolactin ROBERT T. RUBIN, PAUL R. GOUIN, ARDIE LUBIN, RUSSELL E. POLAND, AND KARL M. PIRKE1 Departments of Psychiatry, Pharmacology, and Medicine, UCLA School of Medicine, Harbor General Hospital Campus, Torrance, California 90509; and Navy Medical Neuropsychiatric Research Unit, San Diego, California ABSTRACT. The nocturnal increase of plasma testosterone (T) in adult men has been well established. Luteinizing hormone (LH) does not show a similar increase throughout the night, whereas prolactin (PRL) does, suggesting the possibility of other hormone influences on T secretion. To investigate this possibility, 8 young adult men were studied for 4 consecutive nights in the sleep laboratory (2 nights adaptation, 2 nights blood sampling), by blood samples taken every 30 min during the 8-h sleep period, for measurement of LH, follicle stimulating hormone (FSH), PRL, and T. LH and FSH were secreted episodically, with little or no change in baseline levels during the
night. PRL and T also were secreted episodically, but their baseline levels increased as the night progressed. Both LH and PRL had maximum within-subject correlations (averages = +0.35 and +0.48 respectively) with T when they led T by 60 min. Within-subject correlations done on first differences (to remove the effect of slow trends) were near zero. LH and PRL had larger correlations with T than did FSH, for both calculations. These data suggest that both LH and PRL levels precede T levels by about 60 min. PRL thus may participate in the regulation of nocturnal T secretion in adult men. (J Clin Endocrinol Metab 40: 1027, 1975)
T
HE rise of plasma testosterone (T) during the night in normal young adult men has been well established. Judd et al. (1) cite relevant references to mid1973; more recent studies (1-7) also confirm this finding. The mechanisms responsible for this nocturnal increase, however, are not completely understood. Evans et al. (8,9) suggested a relationship between T release and rapid eye movement (REM) sleep episodes. Judd et al. (1) found that in 4 adult men the nocturnal secretion of T was episodic, but it was unrelated to sleep onset or REM episodes. They noted that rises of T were always preceded by luteinizing hormone
(LH) peaks, with a lead time of 20-140 min (median time 80 min). Conversely, only one-third of the LH peaks were followed by distinct T secretory episodes. These findings were consistent with earlier data presented by Naftolin et al. (10), who showed that, in 4 adult men during the daytime hours, T release was always preceded by an LH secretory episode, with a lead time of 45-80 min. Again, during the day, only one-third of LH peaks were followed by a distinct T release. Judd et al. (1) also measured follicle stimulating hormone (FSH) and prolactin (PRL) in their sleeping subjects but found no relation between these hormones and T. This lack of a visually observable relation Received December 2, 1974. between T and PRL is of particular inSupported by NIMH Research Scientist Developterest, because PRL shows a consistent rise ment Award K1-MH47363 (to RTR) and by Office of during the hours of sleep, similar to T, Naval Research Contract N00014-73-C-0127. Address reprint requests to: Robert T. Rubin, M.D., whereas LH shows little or no increase in B-4 Neurology Laboratory, Harbor General Hospital, integrated levels as the night progresses Torrance, California 90509. (11,12). Furthermore, FSH has been impli1 Present address: Max Planck Institut fur Psychiat- cated in augmenting T release in both rie, Munich, Germany. animals and humans, and PRL has been Presented in part at the Fourth Congress of the International Society for Psychoneuroendocrinology, implicated in T release in animals (13-18). We therefore undertook a quantitative Berkeley, California, September 1973. 1027
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 04 October 2015. at 12:02 For personal use only. No other uses without permission. . All rights reserved.
1028
RUBIN ETAL.
statistical assessment of the nocturnal interrelations of LH, FSH, PRL, and T in normal young adult men, with a consideration of the relative influence of each of the anterior pituitary hormones on T secretion. Materials and Methods Eight healthy male volunteers, 20-32 yr of age, not on drugs or medications, were studied for 4 consecutive nights according to an established protocol (19,20). The first 2 nights were for adaptation to the sleep laboratory. On the third and fourth nights, blood was sampled every 30 min via a long indwelling venous catheter during the hours of sleep, 2300 to 0700. Some samples could not be drawn because of transient venospasm, but these were infrequent. The plasma was immediately separated and frozen at —20 C until hormone analyses were performed. Plasma LH, FSH and PRL were analyzed by previously published radioimmunoassay techniques (21-23). LH (LER 960), FSH (LER 1366), and PRL (Lewis 201-195-1) were used for iodination. These materials and the antisera (LH batch # 1 , FSH batch #3, PRL-SLV) were kindly supplied by the National Pituitary Agency. LH and FSH results are expressed as ng/ml based on the LER 907 reference preparation, and PRL results are expressed as ng/ml based on the Lewis preparation. Plasma T also was analyzed by radioimmunoassay (24), using a recently developed antiserum. The preparatory column chromatographic step was not performed, so that both T and dihydro-T were measured in the assay. For each hormone, all samples were run in duplicate, and all samples from the same subject (both nights) were run in the same assay. Intra-assay variability was never greater than 6%, and interassay variability averaged 20%. The 50% intercept and (B/T)o values for each of the assays were within acceptable tolerance ranges. Statistical analyses were undertaken to characterize the release pattern of each of the hormones and their interrelations. The rhythmicity of hormone secretion was measured by fitting sine curves of circadian and ultradian periodicities to the hormone data, as was done in a previous study of nocturnal patterns of LH and FSH release (19,20). Kendall's rank order coefficient of concordance (W) (25) also was used to compare average hormone values by
JCE fit M • 1975 Vol 40 • No 6
thirds of the night across the 8 subjects to test for slow trends in hormone levels as the night progressed. The linear relations of the 3 anterior pituitary hormones to T were measured by calculating the within-subject correlations of LH, FSH, and PRL with T, both simultaneously and with the pituitary hormones leading T by several time intervals. These within-subjects correlations also were calculated on the first differences of the data, to eliminate slow trends and to highlight any covariation of the more rapid hormone secretory episodes (see Appendix).
Results
Figures 1 and 2 illustrate the consecutive nightly pattern of release of the 4 hormones studied, for 2 of the 8 subjects. These subjects were 2 of the 4 with sufficiently complete hormone data to permit reliable correlations. Their hormone secretion patterns are quite representative of the other 4 subjects studied. For all the subjects, both LH and FSH appeared to show episodic release patterns that were different between subjects and inconstant between the 2 consecutive nights for the same subject. PRL showed an episodic release pattern, with increasing blood levels during the first hours of sleep and peak levels occurring between 4-6 AM in most subjects. T also appeared to be released episodically, with increasing blood levels occurring throughout sleep. It is difficult to ascertain by inspection of Figs. 1 and 2 the specific interrelations among the hormones, especially between LH and T. Although some increases in T may be seen to follow secretory episodes of LH at varying time intervals, the overall increases in T levels as the night progressed appear to relate to PRL increases as well as to LH. Rhythm analyses. The presence of circadian rhythms in the hormone patterns was tested for each of the 8 subjects by fitting the 8-h block of data for each night to the best-fit sine curve with a 24-h period. For both LH and FSH the correlations between the hormone data and the 24-h sine curves
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 04 October 2015. at 12:02 For personal use only. No other uses without permission. . All rights reserved.
NOCTURNAL T E S T O S T E R O N E INCREASE IN MEN A. A. 17 MARCH '72 1412-
TESTO. (ng/ml)
A. A. 18 MARCH '72 1412-
10-
10-
8-
8-
6-
6-
170150130-
170150-
LH (ng/ml)
110-
90-
90-
70-
70-
50-
5 0-
400-
TESTO (ng/ml)
LH. (ng/ml)
130-
110-
450-
1029
F.S.K (ng/ml)
3503002 50-
50-
PRL. (ng/ml)
45403530252015105-
2300
2400
0100 0200 0300 0400 0500 0600 0700 TIME OF NIGHT
2300 2400 0100 0200 0300 0400 0500 0G00 0700 TIME OF NIGHT
FIG. 1. Plasma levels of testosterone, luteinizing hormone, follicle stimulating hormone, and prolactin in a sleeping young adult man, sampled every 30 min between 2300 and 0700 on 2 consecutive nights.
were highly variable across subjects (range either of these hormones in the adult male for each hormone = +0.10 to +0.85). In (10,19,20,27,28). addition, the acrotimes (26), or times of On the other hand, both PRL and T maximum amplitude, were not consistent showed consistently high correlations with for the 2 nights, although the LH acrotimes the best-fit 24-h sine curve (range for each did tend to cluster at 0400-0500. These hormone = +0.40 to +0.92). The acrotimes findings indicate a weak or absent circa- for PRL tended to cluster at 0300-0500, dian rhythm for both LH and FSH. A and those for T at 0400-0600. Median Kendall analysis of concordance was com- levels of both hormones increased steadily puted across the 8 subjects for average during the night, and Kendall's W was hormone levels by thirds of the night, as significant for both (Table 1). Thus a clear, shown in Table 1. Median levels of FSH statistically significant trend of increasing were quite constant during the night, and levels throughout the night was present for LH levels increased slightly. Kendall's W both PRL and T; these results also are was nonsignificant for both hormones. consistent with those of previous studies These results are consistent with several (1-7, 29-32). previous studies of LH and FSH which The presence of ultradian rhythms (26) indicate little or no circadian rhythm for in the hormone patterns was tested for each
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 04 October 2015. at 12:02 For personal use only. No other uses without permission. . All rights reserved.
1030
JCE & M • 1975 Vol 40 • No 6
RUBIN ETAL. M.M. 21 MARCH ' 7 2
141210864150130110907050303503002502 00130100151050-
141210864150I SOII 090705030350300250200150100151050-
TESTO. (ng/ml)
2300 2400
MM. 22 MARCH '72
0100 0200 0300 0400 0500 0600 0700 TIME OF NIGHT
TESTO. (ng/ml)
L.H. (ng/ml)
F.S.H. (ng/ml)
2300 2400
0100 0200 0300 0400 0500 0600 0700 TIME OF NIGHT
FIG. 2. See Legend to Fig. 1.
subject by fitting the 8-h block of data for hormone data sufficiently complete to pereach night to best-fit sine curves with mit cross-correlation analysis. Table 2 preultradian periods of 45, 60, 75, and 90 min sents the average within-subjects correlaand 2, 4, and 8 h. Although many of the tions, using these 4 subjects, of T with the individual correlations between the hor- 3 pituitary hormones. Because of the very mone data and the ultradian sine curves small sample size, these correlations lack were high, there was no consistent pattern power for the testing of statistical sigacross the 8 subjects for any particular time nificance; nevertheless, they are descripperiod for any of the 4 hormones. Thus, tively useful for the elucidation of hormone neither LH, FSH, PRL nor T showed interrelationships. For the raw data (Table consistent ultradian secretion rhythms; 2), the maximum correlation of LH with T again, these findings are consistent with occurred when LHt was correlated with Tt+60; that is, when LH led T by 60 min. those of previous studies. FSH showed relatively low correlations Intercorrelations among hormones. As with T for all lead times. As with LH and mentioned earlier, 4 of the 8 subjects had T, the maximum correlation of PRL and T also occurred when PRL led T by 60 min. TABLE 1. Median hormone levels for eight subjects by The same cross-correlations were perthirds of night, Kendall's W, and significance levels formed on the first differences of the 34 scores for each subject (Table 2). This First Second Third P third third third W eliminated the effect of slow trends (as with PRL) and emphasized the episodic LH (ng/ml) 62 74 NS 74 0.24 fluctuations in hormone levels (as with FSH (ng/ml) 210 NS 230 220 0.15 PRL (ng/ml) 10.6
night. PRL and T also were secreted episodically, but their baseline levels increased as the night progressed. Both LH and PRL had maximum within-subject correlations (averages = +0.35 and +0.48 respectively) with T when they led T by 60 min. Within-subject correlations done on first differences (to remove the effect of slow trends) were near zero. LH and PRL had larger correlations with T than did FSH, for both calculations. These data suggest that both LH and PRL levels precede T levels by about 60 min. PRL thus may participate in the regulation of nocturnal T secretion in adult men. (J Clin Endocrinol Metab 40: 1027, 1975)
T
HE rise of plasma testosterone (T) during the night in normal young adult men has been well established. Judd et al. (1) cite relevant references to mid1973; more recent studies (1-7) also confirm this finding. The mechanisms responsible for this nocturnal increase, however, are not completely understood. Evans et al. (8,9) suggested a relationship between T release and rapid eye movement (REM) sleep episodes. Judd et al. (1) found that in 4 adult men the nocturnal secretion of T was episodic, but it was unrelated to sleep onset or REM episodes. They noted that rises of T were always preceded by luteinizing hormone
(LH) peaks, with a lead time of 20-140 min (median time 80 min). Conversely, only one-third of the LH peaks were followed by distinct T secretory episodes. These findings were consistent with earlier data presented by Naftolin et al. (10), who showed that, in 4 adult men during the daytime hours, T release was always preceded by an LH secretory episode, with a lead time of 45-80 min. Again, during the day, only one-third of LH peaks were followed by a distinct T release. Judd et al. (1) also measured follicle stimulating hormone (FSH) and prolactin (PRL) in their sleeping subjects but found no relation between these hormones and T. This lack of a visually observable relation Received December 2, 1974. between T and PRL is of particular inSupported by NIMH Research Scientist Developterest, because PRL shows a consistent rise ment Award K1-MH47363 (to RTR) and by Office of during the hours of sleep, similar to T, Naval Research Contract N00014-73-C-0127. Address reprint requests to: Robert T. Rubin, M.D., whereas LH shows little or no increase in B-4 Neurology Laboratory, Harbor General Hospital, integrated levels as the night progresses Torrance, California 90509. (11,12). Furthermore, FSH has been impli1 Present address: Max Planck Institut fur Psychiat- cated in augmenting T release in both rie, Munich, Germany. animals and humans, and PRL has been Presented in part at the Fourth Congress of the International Society for Psychoneuroendocrinology, implicated in T release in animals (13-18). We therefore undertook a quantitative Berkeley, California, September 1973. 1027
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 04 October 2015. at 12:02 For personal use only. No other uses without permission. . All rights reserved.
1028
RUBIN ETAL.
statistical assessment of the nocturnal interrelations of LH, FSH, PRL, and T in normal young adult men, with a consideration of the relative influence of each of the anterior pituitary hormones on T secretion. Materials and Methods Eight healthy male volunteers, 20-32 yr of age, not on drugs or medications, were studied for 4 consecutive nights according to an established protocol (19,20). The first 2 nights were for adaptation to the sleep laboratory. On the third and fourth nights, blood was sampled every 30 min via a long indwelling venous catheter during the hours of sleep, 2300 to 0700. Some samples could not be drawn because of transient venospasm, but these were infrequent. The plasma was immediately separated and frozen at —20 C until hormone analyses were performed. Plasma LH, FSH and PRL were analyzed by previously published radioimmunoassay techniques (21-23). LH (LER 960), FSH (LER 1366), and PRL (Lewis 201-195-1) were used for iodination. These materials and the antisera (LH batch # 1 , FSH batch #3, PRL-SLV) were kindly supplied by the National Pituitary Agency. LH and FSH results are expressed as ng/ml based on the LER 907 reference preparation, and PRL results are expressed as ng/ml based on the Lewis preparation. Plasma T also was analyzed by radioimmunoassay (24), using a recently developed antiserum. The preparatory column chromatographic step was not performed, so that both T and dihydro-T were measured in the assay. For each hormone, all samples were run in duplicate, and all samples from the same subject (both nights) were run in the same assay. Intra-assay variability was never greater than 6%, and interassay variability averaged 20%. The 50% intercept and (B/T)o values for each of the assays were within acceptable tolerance ranges. Statistical analyses were undertaken to characterize the release pattern of each of the hormones and their interrelations. The rhythmicity of hormone secretion was measured by fitting sine curves of circadian and ultradian periodicities to the hormone data, as was done in a previous study of nocturnal patterns of LH and FSH release (19,20). Kendall's rank order coefficient of concordance (W) (25) also was used to compare average hormone values by
JCE fit M • 1975 Vol 40 • No 6
thirds of the night across the 8 subjects to test for slow trends in hormone levels as the night progressed. The linear relations of the 3 anterior pituitary hormones to T were measured by calculating the within-subject correlations of LH, FSH, and PRL with T, both simultaneously and with the pituitary hormones leading T by several time intervals. These within-subjects correlations also were calculated on the first differences of the data, to eliminate slow trends and to highlight any covariation of the more rapid hormone secretory episodes (see Appendix).
Results
Figures 1 and 2 illustrate the consecutive nightly pattern of release of the 4 hormones studied, for 2 of the 8 subjects. These subjects were 2 of the 4 with sufficiently complete hormone data to permit reliable correlations. Their hormone secretion patterns are quite representative of the other 4 subjects studied. For all the subjects, both LH and FSH appeared to show episodic release patterns that were different between subjects and inconstant between the 2 consecutive nights for the same subject. PRL showed an episodic release pattern, with increasing blood levels during the first hours of sleep and peak levels occurring between 4-6 AM in most subjects. T also appeared to be released episodically, with increasing blood levels occurring throughout sleep. It is difficult to ascertain by inspection of Figs. 1 and 2 the specific interrelations among the hormones, especially between LH and T. Although some increases in T may be seen to follow secretory episodes of LH at varying time intervals, the overall increases in T levels as the night progressed appear to relate to PRL increases as well as to LH. Rhythm analyses. The presence of circadian rhythms in the hormone patterns was tested for each of the 8 subjects by fitting the 8-h block of data for each night to the best-fit sine curve with a 24-h period. For both LH and FSH the correlations between the hormone data and the 24-h sine curves
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 04 October 2015. at 12:02 For personal use only. No other uses without permission. . All rights reserved.
NOCTURNAL T E S T O S T E R O N E INCREASE IN MEN A. A. 17 MARCH '72 1412-
TESTO. (ng/ml)
A. A. 18 MARCH '72 1412-
10-
10-
8-
8-
6-
6-
170150130-
170150-
LH (ng/ml)
110-
90-
90-
70-
70-
50-
5 0-
400-
TESTO (ng/ml)
LH. (ng/ml)
130-
110-
450-
1029
F.S.K (ng/ml)
3503002 50-
50-
PRL. (ng/ml)
45403530252015105-
2300
2400
0100 0200 0300 0400 0500 0600 0700 TIME OF NIGHT
2300 2400 0100 0200 0300 0400 0500 0G00 0700 TIME OF NIGHT
FIG. 1. Plasma levels of testosterone, luteinizing hormone, follicle stimulating hormone, and prolactin in a sleeping young adult man, sampled every 30 min between 2300 and 0700 on 2 consecutive nights.
were highly variable across subjects (range either of these hormones in the adult male for each hormone = +0.10 to +0.85). In (10,19,20,27,28). addition, the acrotimes (26), or times of On the other hand, both PRL and T maximum amplitude, were not consistent showed consistently high correlations with for the 2 nights, although the LH acrotimes the best-fit 24-h sine curve (range for each did tend to cluster at 0400-0500. These hormone = +0.40 to +0.92). The acrotimes findings indicate a weak or absent circa- for PRL tended to cluster at 0300-0500, dian rhythm for both LH and FSH. A and those for T at 0400-0600. Median Kendall analysis of concordance was com- levels of both hormones increased steadily puted across the 8 subjects for average during the night, and Kendall's W was hormone levels by thirds of the night, as significant for both (Table 1). Thus a clear, shown in Table 1. Median levels of FSH statistically significant trend of increasing were quite constant during the night, and levels throughout the night was present for LH levels increased slightly. Kendall's W both PRL and T; these results also are was nonsignificant for both hormones. consistent with those of previous studies These results are consistent with several (1-7, 29-32). previous studies of LH and FSH which The presence of ultradian rhythms (26) indicate little or no circadian rhythm for in the hormone patterns was tested for each
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 04 October 2015. at 12:02 For personal use only. No other uses without permission. . All rights reserved.
1030
JCE & M • 1975 Vol 40 • No 6
RUBIN ETAL. M.M. 21 MARCH ' 7 2
141210864150130110907050303503002502 00130100151050-
141210864150I SOII 090705030350300250200150100151050-
TESTO. (ng/ml)
2300 2400
MM. 22 MARCH '72
0100 0200 0300 0400 0500 0600 0700 TIME OF NIGHT
TESTO. (ng/ml)
L.H. (ng/ml)
F.S.H. (ng/ml)
2300 2400
0100 0200 0300 0400 0500 0600 0700 TIME OF NIGHT
FIG. 2. See Legend to Fig. 1.
subject by fitting the 8-h block of data for hormone data sufficiently complete to pereach night to best-fit sine curves with mit cross-correlation analysis. Table 2 preultradian periods of 45, 60, 75, and 90 min sents the average within-subjects correlaand 2, 4, and 8 h. Although many of the tions, using these 4 subjects, of T with the individual correlations between the hor- 3 pituitary hormones. Because of the very mone data and the ultradian sine curves small sample size, these correlations lack were high, there was no consistent pattern power for the testing of statistical sigacross the 8 subjects for any particular time nificance; nevertheless, they are descripperiod for any of the 4 hormones. Thus, tively useful for the elucidation of hormone neither LH, FSH, PRL nor T showed interrelationships. For the raw data (Table consistent ultradian secretion rhythms; 2), the maximum correlation of LH with T again, these findings are consistent with occurred when LHt was correlated with Tt+60; that is, when LH led T by 60 min. those of previous studies. FSH showed relatively low correlations Intercorrelations among hormones. As with T for all lead times. As with LH and mentioned earlier, 4 of the 8 subjects had T, the maximum correlation of PRL and T also occurred when PRL led T by 60 min. TABLE 1. Median hormone levels for eight subjects by The same cross-correlations were perthirds of night, Kendall's W, and significance levels formed on the first differences of the 34 scores for each subject (Table 2). This First Second Third P third third third W eliminated the effect of slow trends (as with PRL) and emphasized the episodic LH (ng/ml) 62 74 NS 74 0.24 fluctuations in hormone levels (as with FSH (ng/ml) 210 NS 230 220 0.15 PRL (ng/ml) 10.6
