Differential Association of Basal and Postprandial Plasma Ghrelin With Leptin, Insulin, and Type 2 Diabetes

A total of 410 patients who visited the outpatient clinic of the department of nutritional medicine were evaluated. The majority of patients were admitted for the treatment of obesity. The NW subgroup (BMI 18.5–24.9 kg/m²) consisted of patients who came for a routine checkup, exclusion of hyper- or dyslipidemia, celiac disease, and lactose or fructose intolerance. As part of the routine examinations, carbohydrate metabolism was evaluated by a standard test meal consisting of bread, butter, and marmalade (260 kcal, 62% carbohydrate, 32% fat, and 6% protein). In patients in whom the postprandial plasma glucose levels exceeded 150 mg/dl, an additional oral glucose tolerance test was performed to exclude impaired glucose tolerance or diabetes mellitus.

For a better discrimination between insulin and leptin effects on ghrelin, the obese nondiabetic and diabetic subjects were divided according to their fasting basal insulin level. Fasting insulin <6 μU/ml (mean ± 2 SD of NW subjects) was considered as cutoff. Type 2 diabetes was either newly diagnosed in patients in the outpatient center (76%) or diagnosed up to 9 months before admission. The patients included in the study population had neither diabetes-related complications nor were being treated with metformin or any other antidiabetic drug.

Patients who were taking medication that is supposed to influence carbohydrate metabolism and related endocrine functions (e.g., β-blocker, steroids, diuretics, and antidiabetic drugs) were excluded from the analysis unless the respective substances were withdrawn at least for 2 weeks before the test meal. The demographic data of the patients are summarized in Table 1.

After informed consent was obtained, all examinations were performed according to the guidelines of the Ethical Committee of the Technical University of Munich and in accordance to the principles of the Declaration of Helsinki.

Examinations started at 8.00 a.m., after a 12-h overnight fast. An indwelling catheter was inserted into a forearm vein for collection of blood samples at −15, 0, 15, 30, 60, 90, 120, 150, and 180 min. The test meal had to be consumed within 10 min. Samples were collected into plastic tubes containing 1.2 mg of EDTA and 500 kIU of Trasylol for hormone analysis and into NaF-containing tubes for the determination of glucose. They were kept chilled until centrifugation at 2000 rpm for 15 min at 4°C. The separated plasma was stored at −20°C until the time of assay. All samples of one subject were run in duplicate in the same assay.

Plasma ghrelin levels were determined with a commercial radioimmunoassay that has been used in several previous studies (10,13,14,17,24,25) (Phoenix Pharmaceuticals, Belmont, CA). The assay uses ¹²⁵I-labeled bioactive ghrelin as a tracer molecule and a polyclonal antibody raised in rabbits against full-length octanylated human ghrelin, which detects both active and inactive ghrelin. The interassay and intra-assay coefficients of variation were 10 and 4%, respectively. No cross-reactivity was observed with gastrin, somatostatin, gastric inhibitory polypeptide, glucagon-like peptide 1(7–36)amide, neuromedin C, cholecystokinin, and insulin, respectively.

Insulin was determined using a radioimmunoassay from DPC (Los Angeles, CA). Leptin was measured by a radioimmunoassay purchased from Linco Research (St. Charles, MO). Insulin sensitivity was determined by the homeostasis model assessment for insulin resistance (HOMA-IR) (43). The HOMA-IR was calculated as [fasting blood glucose (milligrams per deciliter) × fasting insulin (microunits per milliliter)]/405. Glucose was measured by the hexokinase method (Roche Diagnostics, Mannheim, Germany).

Normally distributed data are expressed as means ± SD, if not otherwise stated. In the case of non-Gaussian distribution (ghrelin, leptin, and insulin) mean, SD, and interquartile range are stated. Incremental levels were calculated for the 180-min postprandial period as the sum of differences between each time point and the mean baseline level. To compare means, the t test for independent samples or Welch test was used. Comparisons of means within groups were performed by t test for paired data considering multiple testing according to the procedure of Bonferroni. Multiple regression analysis using forward and backward variable selection was used to investigate the influence of independent variables including BMI, leptin, insulin, sex, age, and the interaction term of insulin and leptin on the dependent variable ghrelin. Ghrelin, leptin, and insulin levels were log-transformed to meet linear model assumptions. All P values given are two-sided and subject to a significance level of 0.05. Statistical examinations were performed using SPSS (version 11.5).

Basal plasma insulin concentrations were 2.00 ± 1.97 (0.53–2.86) μU/ml (mean ± SD [interquartile range]) in NW subjects, 2.84 ± 1.58 (1.45–5.01) μU/ml in normoinsulinemic obese (NO), and 14.1 ± 8.17 (8.0–18.0) μU/ml in hyperinsulinemic obese (HO) subjects (P < 0.001 vs. NW and NO) (Fig. 1).

Basal ghrelin was significantly lower in the NO group (370.3 ± 215.5 [211.8–517.9] pg/ml, P = 0.002) compared with NW (555.7 ± 398 [324.8–659.3] pg/ml), and a further reduction to 237.9 ± 162.7 (135.6–303.6) pg/ml (P = 0.001 vs. NW and NO) was observed in HO subjects (Fig. 1). Corresponding leptin concentrations were 5.6 ± 5.1 (1.89–7.95) ng/ml (NW), 26.1 ± 19.6 (8.5–32.7) ng/ml (NO) (P < 0.001), and 41.3 ± 35.3 (21.9–47.5) ng/ml (HO) (P < 0.001 vs. NW and NO). The mean basal plasma glucose level was 87.6 ± 10.9 (81.5–92.5) mg/dl in NW, 91.4 ± 9.8 (84.5–97.5) mg/dl in NO (P < 0.02), and 95.3 ± 12.2 (86.2–103.5) mg/dl in the HO group (P < 0.001 vs. NW and P = 0.006 vs. NO). HOMA-IR was 0.49 ± 0.58 (0.1–0.74) in NW, 0.65 ± 0.38 (0.32–0.97) in NO (P = 0.02), and 3.34 ± 2.03 (1.83–4.14) in HO subjects (P < 0.001 vs. NW and NO).

Forward and backward variable selection resulted in the same covariate model. Multiple linear regression analysis demonstrated that BMI (standardized regression coefficient β = −0.359, P < 0.001), sex (β = 0.229, P = 0.001; with women having higher levels than men), and insulin (β = −0.151, P = 0.025) but not leptin (β = 0.127, P = 0.152) and age (β = 0.116, P = 0.908) had a significant association with ghrelin. When patients with normal and high insulin levels were matched for BMI, sex, age, and leptin, the normoinsulinemic group consisted of 48 subjects (5 men and 43 women; BMI 39.1 ± 4.2 kg/m²; 43 ± 11.8 years), and the high-insulin group contained 48 subjects (6 men and 42 women; BMI 38.4 ± 2.95 kg/m²; 45.1 ± 2.3 years). In this HO subgroup (insulin, 12.8 ± 6.9 μU/ml; leptin, 33.4 ± 13.1 ng/ml), ghrelin was significantly lower with 261.6 ± 171.1 pg/ml compared with the 334.6 ± 173.9 pg/ml (P = 0.042) of the NO subgroup (insulin, 2.8 ± 1.4 μU/ml; leptin, 32.2 ± 13.9 ng/ml). On the other hand, increasing leptin concentrations in the hyperinsulinemic subjects had no influence on basal ghrelin levels (Table 2).

Multiple regression analysis of the normoinsulinemic subjects demonstrated a significant association of ghrelin with leptin (β = −0.477, P < 0.0001) and sex (β = 0.408, P < 0.0001; with women having higher levels than men) but not with insulin (β = −0.084, P = 0.578) and age (β = −0.052, P = 0.516). The highest ghrelin concentration of 658.6 ± 507.5 pg/ml was observed in the men with leptin levels <2.0 ng/ml (Fig. 2), and ghrelin concentration significantly decreased to 444.6 ± 392.6 pg/ml (P < 0.05) at leptin levels of 2–4.9 ng/ml and further to 302.4 ± 276.9 pg/ml at leptin levels of 5–9.9 ng/ml (P < 0.01). Thereafter, ghrelin values were at a plateau around 400 pg/ml despite substantially higher leptin concentrations. In the women, a similar tendency was observed, but because of the small number of subjects in the low leptin group, no statistics were calculated.

Interestingly, the decrease of ghrelin levels was not only associated with an increase of leptin but also with a small but significant rise of insulin (Fig. 2). To exclude an additional effect of insulin, men with low (<5 ng/ml) or high leptin were matched for insulin. Ghrelin levels were significantly lower in the high leptin group (Fig. 3). They remained unchanged in a third group matched for high leptin and BMI that had higher insulin levels, indicating that in normoinsulinemia, the inverse leptin-ghrelin relation is independent of small but significant alterations of insulin.

Similar to the nondiabetic obese subjects, patients with type 2 diabetes were divided into two groups according to basal insulin levels below (normoinsulinemic subjects with type 2 diabetes [DNO]) or above (hyperinsulinemic subjects with type 2 diabetes [DHO]) 6 μU/ml, respectively. Insulin concentrations were 3.25 ± 1.73 (2.2–4.7) μU/ml (DNO) and 16.8 ± 7.9 (10–21.6) μU/ml (DHO) (P < 0.001), respectively. The corresponding glucose levels were 148.0 ± 54.2 (106.8–175.5) mg/dl and 168.4 ± 63.8 (122.7–200.4) mg/dl (P = 0.13, NS; Fig. 1). Basal ghrelin levels were significantly lower in the DNO group compared with the DHO group (DNO 300.9 ± 172.1 [161.7–395.9] pg/ml vs. DHO 216.5 ± 150.6 [110.2–275.7] pg/ml, P = 0.015). Leptin in DHO patients (29.9 ± 22.5 [15.7–395.9] ng/ml) was significantly higher compared with DNO patients (21.4 ± 18.18.2 [9.5–27.1] ng/ml, P = 0.04).

In comparison with the nondiabetic NO subjects, ghrelin was significantly lower in the DNO patients (DNO 300.9 ± 172.1 [161.7–395.9] pg/ml vs. NO 370.3 ± 215.5 [211.8–518] pg/ml, P < 0.001). No significant difference was observed between DHO and HO patients (237.9 ± 162.7 [135.6–303.6] pg/ml, NS). Plasma leptin concentration in DNO was lower but not significantly different from the 26.1 ± 9.6 (8.5–32.7) ng/ml of the NO group (P = 0.13, NS), whereas DHO leptin was significantly lower than HO (41.3 ± 35.3 (21.9–47.5) ng/ml, P < 0.001; Fig. 1).

After meal ingestion, insulin levels were significantly elevated above baseline between 15 and 150 min with a maximum at 60 min in all three groups (Fig. 1)

In NW subjects, postprandial ghrelin remained unchanged during the first 15 min decreasing thereafter significantly between 30 and 120 min (P < 0.01). In the NO group, ghrelin rose significantly by 22 pg/ml during the first 15 min (P = 0.007), decreasing thereafter significantly between 60 and 120 min (P < 0.001). At 180 min, ghrelin levels were again significantly above the mean basal value by 23 pg/ml (P = 0.049). In the HO group, there was a small but significant increase by 17 pg/ml (P = 0.007) at 15 min, decreasing thereafter to significantly lower values at 60 and 90 min (P < 0.001).

The maximal decrease of ghrelin levels below baseline was significantly attenuated in NO compared with NW patients and was significantly postponed by 12 min (Table 3). A further attenuation of ghrelin suppression was observed in HO subjects, whereas the maximal rise of insulin showed the inverse pattern and occurred 10–12 min earlier than trough levels of ghrelin. The patients of the NO and HO subgroup, respectively, matched for age, sex, BMI, and leptin had comparable Δ-insulin and Δ-ghrelin levels with similar significant differences.

Corresponding to the different trough values of ghrelin was the steepness of the slope during the fall and rise of postprandial ghrelin concentrations in the respective groups (Table 4). The initial decrease of ghrelin was significantly steeper than the subsequent rise back to baseline in all three groups. Similar to the maximal changes of insulin and ghrelin, the steepness of the respective slopes was also inversely related.

Plasma glucose rose by 27 mg/dl (P < 0.001) within 30 min in NW subjects, returning to baseline levels at 90 min. In NO and HO subjects, the maximal increase of glucose was 33 and 31 mg/dl within 60 min, respectively (P < 0.001; Fig. 1). In comparison with the NW group, glucose levels of both obese groups were significantly higher at 60, 90, and 120 min (P < 0.01). Plasma leptin concentrations remained unchanged during the entire postprandial period in all three groups (Fig. 1).

In both diabetic groups, postprandial insulin was significantly elevated between 15 and 180 min (P = 0.01; Fig. 1). In comparison with the nondiabetic obese subjects, the initial rise was attenuated. In the DNO group, the maximal rise of insulin was similar to NO subjects but occurred nearly 30 min later (Table 3). In the DHO group, the maximal rise of insulin was significantly less in comparison with HO, but the occurrence was similarly delayed.

In DNO patients, ghrelin levels rose significantly by 36 pg/ml at 15 min (P < 0.001), decreasing thereafter below baseline levels between 60 and 90 min (P = 0.02). Ghrelin in the DHO group showed a similar pattern with a significant rise at 15 min by 24 pg/ml (P = 0.007). Thereafter, ghrelin decreased but was significantly below baseline only at 90 min (P < 0.001; Fig. 1).

The maximal decrease of ghrelin in the DNO group (−131 ± 95.3 pg/ml) was not different from the NO group. In the DHO group, the ghrelin decrease was significantly attenuated compared with the DNO subjects (−82.9 ± 81.4 pg/ml, P = 0.01), but it was identical to the HO group (Table 3).

The slope of decreasing and increasing ghrelin levels in both diabetic groups was similar to the nondiabetic subjects (Table 4), whereas the corresponding slopes of rising insulin were significantly attenuated in diabetic patients. The subsequent fall of insulin was identical in DNO but attenuated in DHO subjects.

The rise of plasma glucose was similar in both groups with significantly elevated levels over the entire 3-h experimental period (Fig. 1). Postprandial leptin concentrations did not change significantly from baseline in both diabetic groups.

The present data demonstrate that obesity (BMI >25 kg/m²) is associated with reduced basal plasma ghrelin levels and an attenuated postprandial decrease, which confirms several previously published observations (19–23,44). Some but not all studies support the concept that insulin is an inhibitor of ghrelin secretion (24–30;33). Thus, obesity-related hyperinsulinemia could at least in part be responsible for ghrelin suppression. On the other hand, leptin has to be considered as a tonic regulator of the gastrohypothalamic axis involved in short-term feeding regulation (45,46). Leptin provides information about the energy reserve stored in adipocytes, because plasma leptin concentrations correlate well with the body fat cell mass (35,36). From experiments in the isolated stomach and from animal studies, there is substantial evidence that leptin can exert an inhibitory effect on gastric ghrelin release (38–40), and in some but not in all studies in obese humans, an inverse relation between leptin and ghrelin levels has been described (21,44,47). This raises the possibility that not only insulin but also leptin contributes to ghrelin suppression in obese subjects. On the other hand, Chan et al. (42) recently presented evidence that leptin does not suppress ghrelin levels in NW subjects.

The reduction of basal ghrelin levels is not only associated with elevated insulin and leptin but also with higher BMI levels. This as well as age and sex are potential confounding factors (22,42,47). When patients were matched for age, sex, BMI, and also leptin, the difference in insulin was still associated with a significant reduction of ghrelin, whereas subgroups with comparable high insulin levels but progressively increasing BMI and leptin concentrations did not show any alteration of basal ghrelin. This supports the concept that in obese subjects with associated hyperinsulinemia, ghrelin suppression is due to insulin, whereas leptin is of only minor or no importance.

The multiple regression analysis of the normoinsulinemic subjects demonstrates that ghrelin is negatively associated with leptin but not insulin. It must be noted, though, that the decrease of basal ghrelin is paralleled not only by an increase of leptin but also by a significant rise of insulin. The subsequent analysis of subgroups matched for insulin or leptin, respectively, suggests that such small, although significant, changes of insulin have no effect on the suppression of ghrelin levels, which supports the concept that leptin is important for the negative feedback regulation of ghrelin in states of moderately increasing body weight. The previously reported inefficacy of recombinant leptin on plasma ghrelin in NW subjects (42) is in contrast to this concept and is as yet unexplained.

The inverse relation between leptin and ghrelin reaches its maximum at a leptin level of 10 ng/ml, which corresponds to a BMI of 28 kg/m². Despite substantially higher leptin and BMI, no further decrease of ghrelin was observed. If this effect of leptin is a mirror image of the action of leptin on the other regulatory pathways of feeding behavior, then the direct feedback control by the energy reserve would be limited to a rather small range of body weight. On the other hand, the limitation of the regulatory role of leptin to BMI 25–30 kg/m² could be indicative for the upper weight range that occurred during human evolution. At higher body weight and more severe metabolic deterioration, high insulin could in part compensate for the loss of leptin regulation.

In patients with type 2 diabetes, higher basal insulin was also associated with reduced ghrelin levels. Interestingly in normoinsulinemic subjects with type 2 diabetes, ghrelin was significantly lower compared with the nondiabetic group. Because insulin concentrations were comparable and leptin was even lower, other as yet unidentified factors related to the diabetic state must be responsible. The elevated glucose levels are most likely of no major importance considering previous in vivo studies (33). Furthermore, in the isolated rat stomach changes of perfusate glucose over a wide range between 30 and 300 mg/dl had no effect on ghrelin secretion (F.L, J.E., V.S., unpublished observations).

In contrast to the normoinsulinemic diabetic patients, those with high basal insulin had no further reduction of ghrelin compared with the nondiabetic group. This could at least in part be due to the previously reported attenuated suppressive action of insulin on ghrelin in diabetic patients (48).

The postprandial decrease of plasma ghrelin levels in NW subjects is in accordance with previously reported data. The reduction of ghrelin depends on an adequate basal and postprandial insulin secretion because ghrelin levels increase after test meals that elicit only a small rise of plasma insulin (10,11,13,14,20,31,32). In addition, intestinal hormones contribute to the suppression of ghrelin secretion (29,34). The present data show that in obese subjects, the postprandial decrease is attenuated, which is in line with previous observations (20,44,49,50). It extends these data by demonstrating an inverse relation between postprandial ghrelin and insulin. Because the postprandial insulin response is substantially greater in both obese groups, one would expect the opposite effect on postprandial ghrelin, considering the inverse relationship between postprandial insulin and ghrelin in NW subjects (10,11,13,14,20). The mechanisms responsible for this attenuated postprandial suppression of ghrelin with increasing obesity are as yet unknown.

In contrast to NW subjects, there was an early rise of postprandial plasma ghrelin levels in obese subjects with and without diabetes before its subsequent suppression. A tendency toward an increase of ghrelin immediately after carbohydrate-rich test meals has also been observed in previous studies in NW subjects, although statistical significance has never been demonstrated (10,13,14). In the diabetic subjects, the rapid increase of ghrelin could at least in part be explained by the attenuated early-phase insulin response.

The reduction of basal ghrelin secretion in response to an increasing energy reserve of the body seems to be a reasonable negative feedback loop that helps to turn down further energy intake. Therefore, it is somewhat surprising that the postprandial decrease of ghrelin is attenuated in obese subjects. The disadvantage of this phenomenon could be a delayed and reduced support of neurally mediated gastric satiety signals. It should be noted, however, that hunger/satiety sensations and the early ghrelin response correlate poorly and that trough levels of ghrelin occur ∼1 h after maximal satiety has been reached (14). Thus, the return of ghrelin back to baseline might be more important (10) as shown by a better correlation between these levels and subsequent meal size (14).

In the event that the ameliorated steepness of returning ghrelin levels in conjunction with lower absolute preprandial concentrations is important for feeding behavior, then obesity-related attenuation of postprandial ghrelin suppression could be a meaningful important regulatory feature. The available infusion studies with ghrelin do not permit an adequate answer to these questions (9), and more sophisticated dose-response studies are required.

In conclusion, the present data suggest that during moderate increases of body weight, leptin can be important for reduction of ghrelin release, whereas in more severe obesity, hyperinsulinemia without leptin is most likely responsible for the cross talk between the energy reserve and short-term feeding regulation. The greater postprandial rise of insulin in obesity is associated with an attenuated suppression of postprandial ghrelin levels, suggesting that in obese and type 2 diabetic subjects, a certain insulin resistance exists at the ghrelin cell that is different from the impairment of glucose disposal.

We thank Margit Hausmann, Christine Herda, Sylvia Tholl, and Jens Peter Zimmermann for their expert technical assistance and Angelina Bockelbrink for expert assistance in data analysis.