Intensive Direct Cavernous Sinus Sampling ... - Oxford Academic

0013-7227/02/$03.00/0 Printed in U.S.A.

Endocrinology 143(1):117–129 Copyright © 2002 by The Endocrine Society

Intensive Direct Cavernous Sinus Sampling Identifies High-Frequency, Nearly Random Patterns of FSH Secretion in Ovariectomized Ewes: Combined Appraisal by RIA and Bioassay IAIN CLARKE, LLOYD MOORE,

AND

JOHANNES VELDHUIS

Prince Henry’s Institute of Medical Research (I.C.), Clayton, Victoria 3168, Australia; Wallaceville Animal Research Centre (L.M.), Upper Hutt, New Zealand; and Endocrine Division, Department of Internal Medicine, School of Medicine, University of Virginia (J.V.), Charlottesville, Virginia 22908 Analyses of FSH secretion suggest pulsatile, nonpulsatile, or compositely pulsatile and nonpulsatile release modes. This may reflect the reduced signal-to-noise ratio inherent in FSH pulse estimation procedures and/or immunological-biological assay inconsistencies. To address these issues, we sampled cavernous sinus and jugular venous blood concomitantly from ovariectomized sheep at either 5-min or 1-min intervals. Samples from the former were assayed by RIA, and those from the latter by RIA and bioassay. Waveform-independent peak detection revealed FSH pulses occurring at high frequency. Pulsatile FSH secretion accounted for 28% of total secretion. Approximate entropy analysis showed that FSH secretion was nearly random. There was synchronous release of LH and

FSH, but most FSH secretion was not associated with LH release; 13% of discrete FSH and LH pulses were concordant. We infer that FSH secretion exhibits pulsatile and basal/nonpulsatile features, with high-entropy features. Linear and nonlinear statistical measures revealed joint sample-by-sample synchrony of FSH and LH release, indicating pattern coordination despite sparse synchrony of pulses. We postulate that pattern synchrony of FSH and LH release is effected at the level of the gonadotrope. Concordant FSH and LH pulses probably result from pulsatile GnRH input, but other mechanisms could account for independent FSH pulses. (Endocrinology 143: 117–129, 2002)

S

YNTHESIS AND SECRETION of the gonadotropins, LH and FSH, depends upon the pulsatile input of GnRH to the gonadotrope. Thus, when the gonadotrope is deprived of GnRH, the synthesis of the gonadotropin subunits and the secretion of the gonadotropins is reduced (1). This primary secretagogue regulates the secretion of the two gonadotropins differently, however, because the withdrawal of pulsatile GnRH input leads to the immediate cessation of the pulsatile secretion of LH, whereas FSH secretion continues for some time (2). This distinction has prompted the notion that there is a basal, or tonic, secretion of FSH, which is not dependent on the secretagogue function of GnRH (2). Following GnRH withdrawal, FSH secretion persists at a diminishing rate over time, and for this reason (among others), it has been proposed that FSH secretion is a direct reflection of the releasable pool of FSH in the gonadotrope. According to this perspective, the secretion of FSH does not require a pulsatile GnRH stimulus. Whether FSH secretion is pulsatile and is directly linked to GnRH secretory pulses has been addressed in various ways. Padmanabhan et al. (3) collected blood from the anterior face of the pituitary gland, which had been lesioned for the purpose of collecting hypophyseal portal blood. They found that 93% of GnRH pulses were associated with secretory pulses of FSH. This method of blood collection from the lesion site would provide an admixture

of blood that was en passage to the pituitary gland and blood from the sinuses of the secondary capillary bed of the portal system within the pituitary gland. At this sampling site, and in some animals, plasma FSH levels were up to 10 times higher than in the jugular vein plasma. The method was taken to provide a reflection of the secretion of FSH from the gonadotropes, but alterations in blood flow (due to the opening of the pituitary gland at the site of the lesion) and admixture of blood from two sources might have confounded facile interpretation of the results. Using sampling of pituitary venous blood in the mare, Irvine and Alexander (4) reported that the majority of GnRH, LH, and FSH pulses were concurrent and concluded that GnRH is the major secretagogue for both gonadotropins. Although these methods have provided valuable information on the secretory profiles of LH and FSH secretion, the issue of joint secretion of LH and FSH remains a matter of some controversy, and we now examine the issue by sampling as close as possible to the normal point of secretion from the pituitary gland, namely the cavernous sinus. We have determined the patterns of secretion of LH and FSH at this level and compared levels in the cavernous sinus to those in the jugular vein by RIA and bioassay (BIO).

Abbreviations: ApEn, Approximate entropy; BIO, bioassay; CV, coefficient of variation; OVX, ovariectomized; X-ApEn, cross-approximate entropy.

Corriedale ewes, which had been ovariectomized (OVX) for at least 1 month, were used during the time of the year when normal ovaryintact animals were mating. The experiment was carried out according

Materials and Methods Animals and ethics

117

118

Endocrinology, January 2002, 143(1):117–129

to the guidelines for the Care and Maintenance of Experimental Animals and was approved by the Monash Animal Ethics Committee and the Animal Ethics Committee of The Victorian Institute of Animal Science before commencement.

Surgery Placement of a cannula into the cavernous sinus was achieved by using a transnasal, transsphenoidal approach to the anterior face of the pituitary gland as previously described (5), with exposure of the cavernous sinuses anterior to the gland. The cavernous sinuses are lateral to the pituitary fossa and extend anteriorly (6). In this species, the carotid arteries enter the sinuses and become the carotid rete (a dense plexus of fine blood vessels) that lie within the sinuses (7). Ascending blood then exits the rete through a common carotid artery (7). The cavernous sinuses collect venous blood from the brain via the transverse sinuses and from other venous outflow of the head, e.g. the facial vein, which passes into the jugular veins (6). The pituitary gland secretes into the cavernous sinuses by passage of blood from the gland via the medial walls of the sinuses. The medial wall of one cavernous sinus was punctured anterior to the body of the pituitary gland with an 18-gauge hypodermic needle and a cannula of polyethylene (internal diameter, 1.57 mm; external diameter, 2.08 mm; WF Scientific, Huntingdale, Victoria, Australia) was inserted into the sinus to a depth of 5 mm. The cannula was advanced so that it lay alongside the wall of the sinus closest to the pituitary gland, and care was taken not to rupture the carotid rete. The entry point to the sinus was packed with GelFoam (Upjohn Pty Ltd., Rydalmere, New South Wales, Australia), and the cannula was secured within the operative tunnel in the sphenoid bone by dental acrylic (Dentsply Ltd., Weybridge, Surrey, UK) introduced by syringe. The cannula was then laid along the ethmoid plate and exited through a burr hole in the nasal bone. It was passed sc to the back of the skull, tied to the animal’s back, and closed with a three-way tap. The cannula was kept patent with heparinized (50 U/ml) normal saline until used (1–3 d later). The animal was allowed to recover for 2– 4 d, and on the day before sampling, a jugular venous cannula was inserted (DwellCath; Tuta Laboratories, Pty Ltd., Lane Cove, New South Wales, Australia) and connected to a manometer line that was also secured on the animal’s back (as described above). Animals were sampled one at a time, and two operators withdrew simultaneous blood samples (2 ml) from the cavernous sinus and the jugular vein. The blood was immediately centrifuged at 4 C, and the plasma was stored at ⫺18 C until assayed. Six sheep were sampled at 5-min intervals over 6 h in the first part of the study. In four additional sheep, 1 ml paired blood samples were taken at 1-min intervals over 4 h for both RIA and BIO of FSH. In the former animals, blood volumes allowed measurement of LH in both cavernous and jugular plasma. In the latter, LH assays were performed only on the jugular samples (to define the position of the secretory pulses), and the cavernous samples were devoted to the dual assay of FSH. There was insufficient plasma for the assay of LH in the cavernous samples taken at 1 min.

RIA of LH and FSH For the samples in Exp 1, plasma levels of LH were assayed in duplicate (100 ␮l) using the method of Lee et al. (8). The standard used was ovine NIH-oLH-S18. For 12 assays, the average sensitivity was 0.1 ng/ml, the intra-assay coefficient of variation (CV) was less than 10% over the range of 0.9 –24 ng/ml, and the interassay CV was 10%. Plasma levels of FSH in the samples in Exp 1 were determined by RIA as previously described (9) with the ovine standard NIAMMD oFSH-RP-1. The sensitivity of the assay was 0.2 ng/ml, the intra-assay CV was less than 10% between 0.5 and 33 ng/ml, and the interassay CV was 23%. Because the samples of Exp 2 were analyzed in a different laboratory than those of the first, a different RIA was used (9). The RIA was similar to the assay used for Exp 1 and was performed as previously described with the exception that the iodinated FSH was purified by ion exchange chromatography (10). The RIA used USDA-oFSH-19-SIAFP-I-2 as the iodination reagent and NIAMMD-anti-oFSH-1 as the antiserum. The minimum detectable concentration of FSH was 0.2 ng/ml (USDA-oFSH19-SIAFP-RP2) as defined by the concentration 2⫻ sd from the zero standard. The mean intra- and interassay CV were 9% and 7%, respectively.

Clarke et al. • Secretion of LH and FSH

FSH BIO The in vitro bioactivity was determined by examining the ability of FSH to stimulate cAMP production by cultured Chinese hamster ovary cells that express recombinant human FSH receptors (11). The cAMP RIA was performed as described (12). The assay had a sensitivity of 0.7 ng of FSH/ml (USDA-oFSH-19-SIAFP-RP2) and an ED50 of 76 ng of FSH/ml. The intra- and interassay CVs were 12% and 13%, respectively.

Statistical procedures for two-hormone synchrony analysis Three mathematically independent statistical approaches were applied to appraise bivariate LH-FSH synchrony, cross-approximate entropy (X-ApEn), cross-correlation analysis, and discrete peak coincidence detection (13). Information gained is complementary and nonredundant, because: a) X-ApEn quantifies lag-independent and nonlinear synchrony of bihormonal release (14); b) cross-correlation analysis detects lag-specific linear relationships between successively paired sample measurements; and c) discrete peak coincidence testing requires a priori pulse identification followed by statistical concordance analysis (13, 15). The complementarity of these analytical tools have been reviewed recently (13, 14).

ApEn Univariate ApEn comprises a class of model- and scale-independent irregularity statistics designed to monitor the relative orderliness of patterns in individual time series (16, 17). ApEn quantifies the subpattern reproducibility of successive measurements and, thus, differs from pulse-detection algorithms (14, 16, 17, 18). ApEn is a single non-negative number, which provides an ensemble estimate of the regularity of any given sequence. Higher ApEn denotes greater disorderliness or randomness of patterns. Technically, ApEn quantifies the logarithmic likelihood that patterns recur in the data (18, 19, 20). ApEn is a family of three-parameter statistics, with members defined by N, m, and r (below). For any given data series containing N observations, two input parameters, namely m and r, are defined, where m represents the pattern (or window) length and r denotes the tolerance (or threshold) for detecting pattern recurrence. To maintain scale invariance, normalized r is defined as a percentage of the between-sample sd of each time series (e.g. 20%), and m is assigned a value of 1 or 2, designating consecutive vectors of length 1 or 2 data points, respectively. For the present time series, we calculated ApEn based on r ⫽ 20% and m ⫽ 1, which is abbreviated as ApEn (1,20%). This parameter set provides a sensitive and replicable ApEn statistic, which has been validated for hormonal time series of this length (14, 16, 21, 22, 23).

X-ApEn To quantify joint synchrony for paired time series, we used X-ApEn (14 and see definition 5 in Ref. 24). This statistic compares the conditional regularity of bivariate data sets. X-ApEn is analogous to ApEn, except that calculations are performed on the standardized (z score-transformed) sequences. In the present study, we applied X-ApEn using m ⫽ 1 and r ⫽ 0.2, which ensures good statistical replicability for the timeseries lengths studied here. Further mathematical discussion of X-ApEn, and a comparison with bivariate spectral and cross-correlation assessments, is given in the appendix of Pincus et al. (14).

Error estimates for ApEn and X-ApEn Empirically based error estimates for individual ApEn or X-ApEn values were made via Monte Carlo simulations (25). Each simulation consisted of 1,000 random shufflings of the original individual or paired data series without replacement to estimate a mean “random” (as distinguished from “observed”) ApEn or X-ApEn. The resultant “null distribution” for each (paired) time series was used to estimate the mean, sd, and maximal value of random ApEn and X-ApEns. Maximally random was defined as the highest single ApEn or X-ApEn value generated in the 1000 shuffled series. The ratio of the observed (actual) to random (shuffled) ApEn (and the sd of this ratio) was computed analogously by 1,000 reassortments of each data series. This ratio would tend to normalize ApEn comparisons between hormones measured in different assays or based on sequences of different lengths.


Statistical comparisons Statistical contrasts in ApEn ratio values for the two hormones were evaluated via a paired two-tailed unequal-variance t test and were confirmed nonparametrically by the Wilcoxon (Mann-Whitney U) statistic.

Cross-correlation analysis Cross-correlation analysis appraises the linear (Pearson’s) correlation coefficient between successively paired measurements in two time series of equal length and spacing considered simultaneously (zero lag) and at various other (nonzero) lag times (13). For example, LH is compared with FSH measured concurrently (zero lag), one sample later (or unit sample lag, here 5 min), or one sample earlier. A positive lag interval was defined here as a change in LH preceding that of FSH by that particular lag time (and, vice versa, for a negative lag). Significant r values (crosscorrelation coefficients) at any given time lag were defined analytically based on the corresponding within-sample pooled sd values adjusted for series length and the number of lag units tested (13). The overall significance of a group of r values at any given lag time was examined via the Kolmogorov-Smirnov statistic. Thereby, we test the null hypothesis that the corresponding z score distribution has a zero mean and unit sd. Because multiple comparisons are involved in cross-correlation analysis, statistical significance was construed at the conservative threshold of P ⬍ 0.01. This restricts the type I statistical error to an asymptotic mean of 1 per 100 comparisons.

Discrete peak coincidence testing Cluster analysis was applied as a model-free and baseline-independent technique to detect discrete peaks in the paired LH and FSH time series (26). We used t statistics of 2.0 to identify significant two-point upstrokes and two-point downstrokes in the 5-min data, and at a statistic of 3.0 with three-point test nadirs and peaks for 1-min data, in both cases based on within-series concentration-dependent pooled sample variances (27). Cluster analysis thereby enumerates peak number; interpeak interval length; and maximal (absolute values), fractional (percentage), and incremental peak amplitudes over the preceding nadir. The interpeak interval is computed as the time between successive peak maxima. Pulse coincidence was defined by the simultaneous or time-lagged concordance of peak maxima in the paired series. The hypergeometric (joint binomial) probability density function was applied to calculate the expected random pulse concordance rate and to test the null hypothesis that the number of observed peak coincidences reflects chance associations alone, conditional on the number of peaks identified in each of the paired series and the total series length. Group coincidence probability values were approximated assuming concatenation (13, 15, 28, 29).

Results Five-minute sampling: RIA

As shown in Fig. 1, catheter placement was corroborated by the significantly (P ⬍ 0.0001) higher levels of LH found in cavernous sinus plasma compared with levels in the jugular. There was a smaller (1.3-fold) but significant (P ⫽ 0.0025) differential for plasma FSH. By discrete peak detection analysis, the median LH interpulse intervals in the cavernous sinus and jugular veins were, respectively, 43 and 50 min (P ⫽ NS; Fig. 2A). Median (as well as mean ⫾ sem) interpeak FSH intervals (min) were also comparable statistically at the two sampling locations, 39 in cavernous and 70 in jugular (P ⫽ NS; Fig. 2A). Thus, LH and FSH pulse frequencies were not distinguishable (except for higher variance for FSH pulsatility) in sheep based on a 5-min sampling paradigm. Secretory peak heights, expressed as percentage of increase over preceding nadir (100%), were substantially higher for LH in cavernous than jugular blood; i.e. the me-

Endocrinology, January 2002, 143(1):117–129 119

dian was 731 for cavernous and 227 for jugular (P ⫽ 0.0006) samples (Fig. 2B). This 5-fold ratio for LH contrasted with a nonsignificant difference for detectable FSH peak heights; the median was 142 for cavernous and 130 for jugular samples. Thus, compared with LH, FSH peak heights were 8.5fold lower in cavernous blood and 4-fold lower in jugular blood. The estimated percentage of pulsatile secretion (sum of peak area/basal) was, for LH, 63 ⫾ 5% (cavernous) vs. 38 ⫾ 4% [jugular (P ⫽ 0.0074)]. Comparable figures for FSH were 28 ⫾ 8% (cavernous) vs. 20 ⫾ 3% [jugular (P ⫽ 0.0015; Fig. 3]. These prominent distinctions are illustrated in Fig. 1. The ApEn statistic was used to quantify the relative orderliness of LH and FSH secretory patterns (see Materials and Methods and Fig. 4). LH secretion into cavernous blood was significantly more orderly than that into jugular blood; i.e. the median LH ApEn ratio was 0.512 for cavernous vs. 0.794 for jugular samples (P ⫽ 0.017). Lower ApEn ratios denote more regular patterns in the data, and a ratio of 1.0 denotes a mean random pattern (equivalent to shuffled data series). In contrast, the median FSH ApEn ratios approached unity for cavernous (0.948) and jugular (1.015) values (P ⫽ NS). Indeed, jugular FSH fluctuations were statistically indistinguishable from those of randomly shuffled data series (mean random FSH ApEn ratio of 1.000 ⫾ 0.0385). The foregoing inferences were corroborated by expressing LH and FSH ApEn values as sd values removed from the mean value of 1,000 randomly shuffled series (Fig. 4, lower panel). LH ApEn in cavernous blood was 7.4 ⫾ 0.73 sd values removed from the mean random (P ⬍ 10⫺6), but FSH ApEn was only 1 ⫾ 0.45 sd values from the mean random (P ⫽ 0.86). Synchrony of LH and FSH release was quantified by complementary linear cross-correlation, nonlinear (patternsensitive) X-ApEn, and discrete peak coincidence probability analysis. Cross-correlation analysis established strong synchrony of LH release observed at the two different sampling sites at a zero or 5-min time lag only (P ⬍ 104 for both), and between LH and FSH in the cavernous sinus at zero time lag (P ⬍ 10⫺4; Fig. 5). In contrast, FSH secretion patterns in the cavernous and jugular sites were independent, as was FSH and LH as determined at the jugular site (data not shown). Nonlinear X-ApEn demonstrated complete loss of LHFSH pattern synchrony in the jugular vein data from the 5-min sampling series (data not shown). There was, however, significant pattern coordination (compared with randomly shuffled LH and FSH pairs) in the data derived from the samples taken at 5-min intervals from the cavernous sinus (P ⬍ 10⫺3; data not shown). Using data from the 1-min sampling (cavernous sinus), FSH RIA and FSH BIO showed detectable (nonrandom) orderliness (P ⬍ 0.05 for both cases) but remarkably less regularity than the LH taken from the jugular samples (P ⫽ 0.00011; Fig. 6). FSH values from either assay maintained significant (P ⬍ 0.025) nonrandom synchrony with LH values in the jugular samples (1-min sampling) (Fig. 6). Discrete peak coincidence testing via the hypergeometric probability density revealed significant group (total) LH/LH peak concordance at the cavernous and jugular sites (P ⬍ 10⫺8) and nonrandom total LH/FSH peak concordance among the groups of animals at the cavernous site only (P ⫽ 0.042; Fig. 7).

120



FIG. 1. Matching plasma concentrations of LH and FSH in cavernous sinus (left column) and jugular venous (right column) blood samples from six individual ewes (A–F). Samples were collected simultaneously from each site at 5-min intervals (see Materials and Methods for details). Error bars represent the predicted assay sample SD values, and vertical arrows denote identified pulses.



FIG. 1. Continued.

122



FIG. 2. Discrete peak-detection (cluster) analysis of 5-min LH and FSH release profiles sampled in the cavernous sinus and internal jugular veins. The upper panel gives the individual data values of the interpeak intervals (min) with the group mean ⫾ SEM. The lower panel depicts fractional peak heights (percentage increase of the peak maximum over the preceding nadir, with the latter defined as 100%). Values of P denote results of ANOVA on the logarithmically transformed values (NS ⫽ P ⬎ 0.05). Unshared alphabetical superscripts define statistically contrasting means.

FIG. 3. Mean (⫾SEM) percentage of total LH and FSH release due to pulsatile secretion in time series from animals sampled at 5-min intervals. Bars with arrows connect post hoc-paired t test comparisons at the indicated P values. Overall contrasts among the four groups were evident at P ⫽ 0.01 (ANOVA).

One-minute sampling: RIA and BIO

Data obtained from 1-min sampling are illustrated for two animals in Fig. 8. Mean FSH concentrations by RIA and BIO in the cavernous sinus were 5.0 ⫾ 2.0 (median ⫽ 5.0) and

11.5 ⫾ 2.1 (median ⫽ 11.8), respectively (P ⫽ 0.017). The median interpeak interval (min) was 12.5 by RIA and 10.8 by BIO (NS; Fig. 9, upper panel). Fractional peak amplitudes (percentage) for FSH measured every 1 min in the cavernous sinus were (median) 143 by RIA and 214 by BIO (P ⫽ 0.013; Fig. 9, lower panel). LH interpulse intervals and percentage peak heights were analogous in the 1-min and 5-min data (Fig. 9 and above), providing a suitable control. Figure 10 shows the lack of significant (nonrandom) X-ApEn between 5-min jugular LH and 5-min jugular or cavernous FSH series. Regularity of the 1-min FSH secretory profiles (Fig. 8) was assessed by the ApEn ratio of observed to random (shuffled series). FSH ApEn ratios were similar in both assay systems; median ⫽ 0.945 (RIA) or 0.948 (BIO) [NS]. Both values were more irregular than that for LH (median ⫽ 0.655; P ⫽ 0.00011). Likewise, FSH ApEn ratios were only minimally removed from random expectation (mean ApEn of 1,000 shufflings of each original FSH time series). The observed ApEn values for FSH were removed from mean random by 2.33 ⫾ 0.99 sd values (P ⬍ 0.01) for RIA and 1.74 ⫾ 1.1 sd values for BIO (NS). In contrast, 1-min LH secretion was highly ordered based on an LH ApEn ratio that was removed from random by 14.5 ⫾ 1.6 sd values (P ⬍ 10⫺6). Crosscorrelation of FSH RIA and BIO time series was significant



FIG. 4. ApEn ratios (upper panel) and ApEn SD values from the mean random (lower panel) for the indicated sampling sites (jugular or cavernous sinus) and hormones (LH and FSH). The ApEn ratio defines the mean ratio of the individually observed ApEn value to that of 1,000 randomly shuffled renditions of the same time series. The number of SD values by which the observed series ApEn is removed from mean random (shuffled) ApEn provides a complementary measure of relative orderliness. An ApEn ratio of unity and a number of SD values from mean random of zero approach mean random expectation. Accordingly, LH release monitored every 5 min in the cavernous sinus is significantly orderly, whereas FSH patterns in jugular blood are not distinguishable from randomly shuffled data (see Materials and Methods). Unshared alphabetical superscripts define statistically contrasting means Arrows connect post hoc-paired t test comparisons at the indicated P values. Overall contrasts among the four groups were evident at P ⫽ 0.01 (ANOVA).

(P ⬍ 0.01) in three of the four individual animals (maximal z scores ⫹2.3 to ⫹4.1). FSH RIA/BIO X-ApEn for 1-min cavernous data averaged 0.956 ⫾ 0.0250 (P ⬍ 0.01 vs. random). FSH RIA and BIO pulses tended toward nonrandom discrete-peak concordance (P ⫽ 0.11). Discussion

High-intensity sampling from the point of secretion (cavernous sinus) in adult OVX sheep and detailed biomathematical analyses of pituitary FSH secretion unveiled several novel features of the pulsatile and entropic (pattern-sensitive) control of this hormone. By concomitant LH measurements, we showed nonrandom joint synchrony of FSH and LH release in this model. Validation experiments based on 1-min cavernous-sinus sampling data and combined RIA and BIO of the FSH time series further disclosed unexpected high frequency oscillations of FSH and established that the subordinate patterns of FSH release were minimally ordered. Collectively, these data thus illuminate both common and unique properties of FSH and LH neuroregulation in vivo (vide infra). Pulsatile analyses identified consistently episodic release of FSH in both the 5- and 1-min sampling paradigms, as corroborated and by both RIA and BIO in a total of 10 an-

imals. The mean (fractional) FSH pulse amplitude ranged from 145 to 214% (RIA, BIO, and both 5- and 1-min data). The lower mean amplitude of 145% exceeds nominal FSH assay imprecision by 4- to 6-fold. Moreover, in the 5-min data set, when LH was assayed simultaneously in cavernous blood, FSH/LH pulse concordance was markedly nonrandom. Whereas FSH pulsatility was statistically unequivocal, two particular features distinguished FSH pulses from those of LH. Firstly, the relative amplitude of LH pulses was 8.5-fold higher than that of FSH pulses at the same sampling site, and secondly, there were unanticipated high-frequency FSH oscillations superimposed in 1-min sampled RIA and BIO FSH time series. Although mean FSH and LH interpulse intervals (reciprocal of pulse frequencies) were indistinguishable statistically based on a 5-min sampling paradigm (i.e. 42 ⫾ 4.7 min for FSH and 45 ⫾ 1.7 min for LH), FSH oscillations recurred every 11–13 min (RIA and BIO, respectively) in the 1-min time series. When FSH was measured by BIO, the mean amplitude of these rapid pulsations was 214 ⫾ 8.7%. Hormonal oscillations of this high frequency have been recognized previously for GH (based on 30-sec blood sampling; Ref. 30), PTH (31), insulin (based on 1-min blood sampling; Ref. 32), and possibly LH and FSH in postmenopausal

124



FIG. 5. Cross-correlation coefficients (y-axis, r values) are given as the median and absolute range for the indicated paired hormone time series evaluated simultaneously and at 5-min intervals lag times (x-axis min) and denote the sampling intervals separating observed associations at the indicated group significance P value (see Materials and Methods). The upper panel compares LH secretion determined from cavernous and jugular samples. When zero lag was observed (coincident peaks) or when the peak occurred 5 min earlier (lag ⫺5) in the cavernous sample, highly significant cross-correlations were obtained. The lower panel shows cross-correlation for LH and FSH secretion as observed in the cavernous samples, with zero lag indicating coincidence of secretion around the time (⫺20 min to ⫹20 min) of identified LH pulses.

women (27, 33). Except for the present direct cavernous sinus blood sampling protocol, however, other studies relied upon measurements in peripheral/systemic blood wherein the ratio of the signal amplitude to the basal hormone level is markedly damped. Indeed, in the present study, the relative amplitude of LH pulses evaluated at the cavernous vs. jugular sampling sites exceeded 9-fold and for FSH was approximately 1.5-fold. The modest differential in FSH levels illustrates the prominent recirculation properties of this hormone, which limits pulsatility resolution. Complementary to pulse analysis is the quantification of subordinate (nonpulsatile) patterns of ad seriatim hormone release (14, 21, 22, 34, 35, 36). The latter regularity analysis provides information about the complexity and relative strength of input (feedback and feedforward) signals controlling moment-to-moment hormone output, e.g. as quantified by ApEn (25). Mean FSH ApEn values in 5-min and 1-min time series were minimally distinguishable from random. Only FSH ApEn calculated for 1-min FSH time series showed significantly nonrandom patterning, because the observed ApEn ratios were 2.33 sd values removed from the mean random ApEn (obtained by shuffling each of the four data series 1,000 times). For LH secretory patterns in jugular

blood determined by either 5- or 1-min sampling, ApEn ratios were 7.4 and 14.5 sd values removed from mean random expectation (P ⬍ 10⫺6 to P ⬍ 10⫺12). Accordingly, the present data establish unequivocal statistical contrasts in the relative complexity of FSH and LH secretory control, whether measured in the cavernous sinus or the jugular vein in the same animal. FSH ApEn ratios tended to show less disorderliness in cavernous than in jugular blood, in accordance with the earlier findings of Padmanabhan et al. (3). On the other hand, Irvine and Alexander (4) showed a very close relationship between GnRH, LH, and FSH secretion when sampling cavernous sinus blood in the mare. We sought to measure gonadotropin secretion at the point of secretion, downstream from the gonadotrope and at the normal exit point from the pituitary gland. In this way, our study hopefully provides data that are representative of the normal physiological process. In comparison with the data obtained in the horse (4), it seems likely that there is a substantial species difference, such that the majority of LH and FSH is secreted in synchronous pulses in the horse, whereas the same is not true in the OVX ewe. The present analyses differ from the previous reports (3, 4) by way of providing a 3-fold complementary quantification of pairwise LH and FSH syn-



FIG. 6. ApEn data for 1-min sampling of cavernous sinus FSH release as quantified by RIA or BIO and LH secretion as measured by RIA. Data are presented exactly as described in the legend of Fig. 4.

FIG. 7. Total number of observed (asterisks) and randomly expected (horizontal dashed lines) coincident peaks for the indicated hormone pairs in the entire group of animals studied. Values of P denote the probability that the observed rate of peak concordance reflects chance associations alone.

chrony, namely cross-correlation, X-ApEn, and discrete peak coincidence probability analysis. Furthermore, with respect to FSH, comparable inferences were made by RIA and BIO. Although agreement among these methodologies offers an important basis for our inferences, comparably comprehensive assessments of data collected in other laboratories and

in additional animals will be important to extend these findings. The highly irregular (virtually random) subpatterns of FSH release, but not LH release, provide unequivocal evidence of unequal control of these two hormones but do not establish the precise cellular or neuromodulatory mecha-

126



FIG. 8. Plasma LH (jugular) and FSH RIA or BIO (cavernous sinus) concentration profiles observed by sampling at 1-min intervals from the cavernous sinus in two ewes. Error bars represent the predicted assay sample SD values.

nisms that drive this distinction. Indeed, analogous dissociation between the relative orderliness of FSH (highly irregular) and LH (more orderly) is evident in the human but vanishes with increasing age in both women and men (37, 38). Highly irregular subpatterns of (nonpulsatile) hormone secretion are also prominent in neuroendocrine tumoral states, such as acromegaly, Cushing’s disease, and prolactinoma (34, 39, 40). Tumoral conditions further emulate FSH output patterns in their predominance of nonpulsatile release; only 28% of cavernous FSH secretion was pulsatile vs. 63% for LH in the same animals. These features suggest that FSH production is driven at a high basal nonpulsatile level. The high ApEn points further to a composite of interactive factors, which jointly confer a highly irregular release process. Plausible candidate regulators include intrapituitary activin, follistatin, and inhibin; extrahypophyseal, adrenogonadal signals (e.g. sex-steroid hormones); and other hy-

pothalamic inputs (e.g. GnRH and other possible modulators of FSH release or inhibition). Concomitant measurements of cavernous-sinus FSH and LH secretion at 5-min intervals allowed us to appraise coordinate secretion of these two gonadotropins. We used three complementary strategies. First, via the linear estimation procedure of cross-correlation analysis, we identified highly significant (P ⬍ 10⫺4) covariations in sample-by-sample FSH and LH release. Secondly, the analytically independent XApEn statistic delineated nonrandom synchrony of the patterns of FSH and LH secretion. This finding was unexpected, given the near-randomness of FSH release considered alone. Indeed, this observation unmasks quantifiable (P ⬍ 0.01 vs. random) pattern-specific coupling between the basal/nonpulsatile control of these two hormones. Thirdly, discrete FSH and LH peak-coincidence analysis established nonrandom co-pulsatility (P ⫽ 0.042). Albeit statistically significant,



FIG. 9. Discrete peak detection (cluster) analysis of the 1-min serum FSH concentration time series that are shown in Fig. 8. Data are presented exactly as described in the legend of Fig. 4.

FIG. 10. X-ApEn values for samples taken simultaneously from the cavernous sinus (FSH) and jugular vein (LH or FSH) at 5-min intervals (data series shown in Fig. 1). X-ApEn ratios of unity denote essentially mean random patterns considered pair-wise (see legend of Fig. 4). P ⫽ NS denotes P ⬎ 0.05.

concordant FSH and LH pulses constituted the minority of events (i.e. 13% observed vs. 5% expected on the basis of change alone). Accordingly, the foregoing perspectives provide evidence for both common (cross-correlation and XApEn) and distinct (peak discordance rate of 87%) mecha-

nisms of FSH and LH neuroregulation. These data agree with those of Padmanabhan et al. (3) but do not show the vast discrepancy between FSH levels at the point of secretion and in jugular blood. We conclude that the majority of FSH secretion is not associated with GnRH/LH pulses. The present data provide a platform for the examination of upstream and downstream implications of high frequency pulsatile secretion of FSH. We speculate that the latter derives from local intrapituitary regulation of FSH secretion by autocrine and paracrine factors. This may involve localized cell clusters being acted upon by such factors. Mathematical modeling would support the plausibility of this conjecture, but studies of the intact pituitary gland and/or regions of gonadotrope cells in normal clusters would be required to affirm or refute this notion. In contrast, the mechanism of LH release, based on extensive studies with GnRH antisera or antagonist peptides (41, 42, 43), appears to be exclusively dependent upon the pulsatile secretion of GnRH from the hypothalamus. In essence, these conclusions are in broad agreement with Padmanabhan et al. (3). Fewer data are available to appraise the downstream effects of pulsatile LH and FSH secretion on gonadal target cells. In some studies, episodic stimulation drives biochemical responses more effectively, and in others the reverse is true (27, 33).

128


In summary, the present experimental analyses combine highly intensive, central cavernous-sinus blood sampling with RIA and BIO of FSH to appraise the regulation of in vivo FSH secretion in the gonadectomized sheep. Thereby, we delineate unequivocal FSH pulsatility, but at 8.5-fold lower relative (fractional) mean amplitude than that of LH. FSH is secreted primarily in a basal/nonpulsatile mode with essentially random patterning of this component. High sampling density (1-min intervals) and sophisticated analysis has revealed rapid oscillations in FSH release occurring every 11–13 min. Synchrony of FSH and LH secretion points to sample-by-sample covariation of regulatory processes. In contrast, albeit nonrandomly associated, discrete pulses of FSH are discordant from those of LH in 87% of cases, thus unveiling distinctions in pulse generation. In ensemble, we infer from our data that both common and distinct neuroregulatory mechanisms govern pituitary FSH and LH secretion in the OVX ewe. Whether analogous complexity of bihormonal secretory control operates in the gonad-intact animal is not known. Acknowledgments We thank Bruce Doughton and Karen Perkins for animal care and assistance with the experiments and Alix Rao, Winny Ng Chie, and Wayne Young for RIAs and BIOs. Received June 29, 2001. Accepted October 17, 2001. Address all correspondence and requests for reprints to: Prof. Iain J. Clarke, Prince Henry’s Institute of Medical Research, P.O. Box 5152, Clayton, Victoria 3168, Australia. E-mail: [email protected]. edu.au. This work was supported by The Australian National Health and Medical Research Council NIH U54 Reproduction Center.

References 1. Clarke IJ, Cummins JT, de Krester DM 1983 Pituitary gland function after disconnection from direct hypothalamic influences in the sheep. Neuroendocrinology 36:376 –384 2. Clarke IJ 1987 New concepts in GnRH action on the pituitary gland. In: Asch R, ed. Seminars in reproductive endocrinology. Thieme Medical Publishers; vol 5:345–352 3. Padmanabhan V, McFadden K, Mauger DT, Karsch FJ, Rees Midgley Jr A 1997 Neuroendocrine control of follicle-stimulating hormone (FSH) secretion. I. Direct evidence for separate episodic and basal components of FSH secretion. Endocrinology 138:424 – 432 4. Irvine HG, Alexander SL 1993 Secretory patterns and rates of gonadotropinreleasing hormone, follicle stimulating hormone and luteinizing hormone revealed by intensive sampling of pituitary venous blood in the luteal phase mare. Endocrinology 132:212–218 5. Clarke IJ, Brown BW, Tran VV, Scott CJ, Fry R, Millar RP, Rao A 1998 Neonatal immunization against gonadotropin releasing hormone (GnRH) results in diminished GnRH secretion in adulthood. Endocrinology 139:2007– 2014 6. Daniel PM, Pritchard MML 1975 Studies of the hypothalamus and pituitary gland with special reference to the effects of transection of the pituitary stalk. Oxford: Alden Press 7. Baker MA 1979 A brain-cooling system in mammals. Sci Am 240:114 –122 8. Lee VWK, Cumming IA, de Kretser DM, Findlay JK, Hudson B, Keogh EJ 1976 Regulation of gonadotrophin secretion in rams from birth to sexual maturity I. Plasma LH, FSH and testosterone levels. J Reprod Fertil 46:1– 6 9. Bremner WJ, Findlay JK, Lee VWK, de Kretser DM, Cumming IA 1980 Feedback effects of the testis on pituitary responsiveness to luteinizing hormone-releasing hormone infusions in the ram. Endocrinology 106:329 – 336 10. McNatty KP, Hudson N, Henderson KM, Gibb M, Morrison L, Ball K, Smith P 1987 Differences in gonadotrophin concentrations and pituitary responsiveness to GnRH between Booroola ewes which were homozygous (FF), heterozygous (F⫹) and non carriers (⫹⫹) of a major gene influencing their ovulation rate. J Reprod Fertil 80:577–588


11. Moore LG, Ng Chie W, Lun S, Lawrence SB, Heath DA, McNatty KP 1997 Isolation, characterization and radioimmunoassay of luteinizing hormone in the brushtail possum. Reprod Fertil Dev 9:419 – 425 12. Albanese C, Christin-Maitre SP, Crowley WF, Jameson JL 1994 Development of a bioassay for FSH using a recombinant human FSH receptor and a cAMP responsive luciferase reporter gene. Mol Cell Endocrinol 101:211–219 13. Veldhuis JD, Johnson ML, Faunt LM, Seneta E 1994 Assessing temporal coupling between two, or among three or more, neuroendocrine pulse trains: cross-correlation analysis, simulation methods, and conditional probability testing. Methods Neurosci 20:336 –376 14. Pincus SM, Mulligan T, Iranmanesh A, Gheorghiu S, Godschalk M, Veldhuis JD 1996 Older males secrete luteinizing hormone and testosterone more irregularly, and jointly more asynchronously, than younger males. Proc Natl Acad Sci USA 93:14100 –14105 15. Veldhuis JD, Johnson ML, Seneta E 1991 Analysis of the co-pulsatility of anterior pituitary hormones. J Clin Endocrinol Metab 73:569 –576 16. Pincus SM, Hartman ML, Roelfsema F, Thorner MO, Veldhuis JD 1999 Hormone pulsatility discrimination via coarse and short time sampling. Am J Physiol 277:E948 –E957 17. Pincus SM 1991 Approximate entropy as a measure of system complexity. Proc Natl Acad Sci USA 88:2297–2301 18. Pincus SM, Goldberger AL 1994 Physiological time-series analysis: what does regularity quantify? Am J Physiol 266:H1643–H1656 19. Pincus SM, Huang WM 1992 Approximate entropy: statistical properties and applications. Commun Stat Theory Methods 21:3061–3077 20. Pincus SM 1995 Approximate entropy (ApEn) as a complexity measure. Chaos 5:110 –117 21. Gevers E, Pincus SM, Robinson ICAF, Veldhuis JD 1998 Differential orderliness of the GH release process in castrate male and female rats. Am J Physiol 274:R437–R444 22. Pincus SM, Gevers E, Robinson ICAF, van den Berg G, Roelfsema F, Hartman ML, Veldhuis JD 1996 Females secrete growth hormone with more process irregularity than males in both human and rat. Am J Physiol 270: E107–E115 23. Veldhuis JD, Metzger DL, Martha Jr PM, Mauras N, Kerrigan JR, Keenan B, Rogol AD, Pincus SM 1997 Estrogen and testosterone, but not a non-aromatizable androgen, direct network integration of the hypothalamo-somatotrope (growth hormone)-insulin-like growth factor I axis in the human: evidence from pubertal pathophysiology and sex-steroid hormone replacement. J Clin Endocrinol Metab 82:3414 –3420 24. Pincus SM, Singer BH 1996 Randomness and degrees of irregularity. Proc Natl Acad Sci USA 93:2083–2088 25. Veldhuis JD, Straume M, Iranmanesh A, Mulligan T, Jaffe CA, Barkan A, Johnson ML, Pincus SM 2001 Secretory process regularity monitors neuroendocrine feedback and feedforward signaling strength in humans. Am J Physiol 280:R721–R729 26. Veldhuis JD, Johnson ML 1986 Cluster analysis: a simple, versatile and robust algorithm for endocrine pulse detection. Am J Physiol 250:E486 –E493 27. Urban RJ, Evans WS, Rogol AD, Kaiser DL, Johnson ML, Veldhuis JD 1988 Contemporary aspects of discrete peak detection algorithms. I. The paradigm of the luteinizing hormone pulse signal in men. Endocr Rev 9:3–37 28. Veldhuis JD, Iranmanesh A, Clarke IA, Kaiser DL, Johnson ML 1989 Random and non-random coincidence between luteinizing hormone peaks and follicle-stimulating hormone, alpha subunit, prolactin, and gonadotropinreleasing hormone pulsations. J Neuroendocrinol 1:185–194 29. Veldhuis JD, Lassiter AB, Johnson ML 1990 Operating behavior of dual or multiple endocrine pulse generators. Am J Physiol 259:E351–E361 30. Holl RW, Hartman ME, Veldhuis JD, Taylor WM, Thorner MO 1991 Thirtysecond sampling of plasma growth hormone in man: correlations with sleep stages. J Clin Endocrinol Metab 72:854 – 861 31. Schmitt CP, Schaefer F, Bruch A, Veldhuis JD, Schmidt-Gayk H, Stein G, Ritz E, Mehis O 1996 Control of pulsatile and tonic parathyroid hormone secretion by ionized calcium. J Clin Endocrinol Metab 81:4236 – 4243 32. Porksen N, Nyholm B, Veldhuis JD, Butler PC, Schmitz O 1997 In humans at least seventy-five percent of overall insulin secretion arises from punctuated high-frequency insulin secretory bursts. Am J Physiol 273:E908 –E914 33. Evans WS, Sollenberger MJ, Booth Jr RA, Rogol AD, Urban RJ, Carlsen EC, Johnson ML, Veldhuis JD 1992 Contemporary aspects of discrete peak detection algorithms. II. The paradigm of the luteinizing hormone pulse signal in women. Endocr Rev 13:81–104 34. Hartman ML, Pincus SM, Johnson ML, Matthews DH, Faunt LM, Vance ML, Thorner MO, Veldhuis JD 1994 Enhanced basal and disorderly growth hormone secretion distinguish acromegalic from normal pulsatile growth hormone release. J Clin Invest 94:1277–1288 35. Pincus SM, Hartman ML, Roelfsema F, Thorner MO, Veldhuis JD 1999 Hormone pulsatility discrimination via coarse and short time sampling. Am J Physiol 277:E947–E957 36. Veldhuis JD, Pincus SM 1998 Orderliness of hormone release patterns; a complementary measure to conventional pulsatile and circadian analyses. Eur J Endocrinol 138:358 –362 37. Pincus SM, Veldhuis JD, Mulligan T, Iranmanesh A, Evans WS 1997 Effects


of age on the irregularity of LH and FSH serum concentrations in women and men. Am J Physiol 273:E989 –E995 38. Veldhuis JD, Iranmanesh A, Mulligan T, Pincus SM 1999 Disruption of the young-adult synchrony between luteinizing hormone release and oscillations in follicle-stimulating hormone, prolactin, and nocturnal penile tumescence (NPT) in healthy older men. J Clin Endocrinol Metab 84:3498 –3505 39. Roelfsema F, Pincus SM, Veldhuis JD 1998 Patients with Cushing’s disease secrete adrenocorticotropin and cortisol jointly more asynchronously than healthy subjects. J Clin Endocrinol Metab 83:688 – 692 40. Veldman RG, van den Berg G, Pincus SM, Frolich M, Veldhuis JD, Roelfsema F 1999 Increased episodic release and disorderliness of prolactin


secretion in both micro- and macroprolactinomas. Eur J Endocrinol 140: 192–200 41. Clarke IJ, Fraser HM, McNeilly AS 1978 Active immunization of ewes against luteinizing hormone releasing hormone, and its effects on ovulation and gonadotrophin, prolactin and ovarian steroid secretion. J Endocrinol 78:39 – 47 42. Fraser HM, McNeilly AS 1983 Differential effects of LH-RH immunoneutralization on LH and FSH secretion in the ewe. J Reprod Fertil 69:569 –577 43. Lincoln GA, Fraser HM 1987 Compensatory response of the luteinizinghormone (LH)-releasing hormone (LHRH)/LH pulse generator after administration of a potent LHRH antagonist in the ram. Endocrinology 120:2245–2250