Plasma GDF15 increases during hyperinsulinemic hypoglycemia in humans with post-bariatric hypoglycemia and after insulin exposure in mice

Summary

Post-bariatric hypoglycemia (PBH), characterized by excessive postprandial incretin and insulin secretion, is a common complication of bariatric surgery. Here, we investigate the relationship between PBH and growth differentiation factor 15 (GDF15) in individuals with PBH after Roux-en-Y gastric bypass (RYGB), post-RYGB individuals who remain asymptomatic (Asx), and individuals with overweight/obesity but without history of surgery (Ow/Ob). Fasting plasma GDF15 is higher in PBH vs. Ow/Ob and further increases postprandially, coinciding with hypoglycemia symptoms. During a hyperinsulinemic hypoglycemic clamp, GDF15 progressively increases in PBH and correlates with hypoglycemia survey symptoms, including weakness, difficulty concentrating, feeling cold, and tingling lips. In mice, insulin-induced hypoglycemia also results in elevated GDF15 levels, and exogenous recombinant GDF15 (rGDF15) reduces food intake in response to hypoglycemia. Our data suggest that GDF15 modulates the counterregulatory response to hypoglycemia in both PBH individuals and mice and that elevated GDF15 levels contribute to hypoglycemia-related postprandial symptoms. This study was registered at ClinicalTrials.gov (NCT04428866).

Graphical abstract

Highlights

• •GDF15 increases in PBH subjects after a mixed meal and during hyperinsulinemic hypoglycemia
• •GDF15 is strongly correlated with hypoglycemia symptom scores
• •Insulin-induced hypoglycemia in mice increases GDF15 levels
• •GDF15 reduces food intake in response to hypoglycemia

Ferraz-Bannitz et al. demonstrate that GDF15 is elevated in individuals with PBH during mixed meal and hyperinsulinemic hypoglycemia, strongly correlating with hypoglycemia symptoms. In mice, hypoglycemia increases GDF15 and rGDF15 blunts food intake responses to hypoglycemia, implicating GDF15 as a conserved counterregulatory response to hypoglycemia in both PBH individuals and mice.

Introduction

One of the most successful therapeutic approaches to obesity and type 2 diabetes (T2D) is bariatric/metabolic surgery, providing not only weight loss but also rapid and sustained improvements in insulin secretion, insulin sensitivity, and improved glycemic control.¹ Unfortunately, up to 20%–30% of patients develop hypoglycemia, termed post-bariatric hypoglycemia (PBH). Hypoglycemia can be severe and is often accompanied by neuroglycopenia and hypoglycemia unawareness.² PBH is most commonly experienced 1 to 3 h postprandially.² A key feature of bariatric surgery, particularly Roux-en-Y gastric bypass (RYGB), is the rapid emptying of both solids and liquids from the gastric pouch, resulting in early and exaggerated nutrient exposure within the absorptive intestine.^3,4 Multiple studies have demonstrated that individuals with PBH exhibit greater postprandial peak insulin and C-peptide compared with post-RYGB individuals without hypoglycemia, with some studies showing exaggerated postprandial levels of GLP-1 and other incretin hormones as well. Antagonism of GLP-1 signaling may represent a rational therapeutic approach, as noted in a recent review and from data in recent trials.^5,6 Furthermore, reductions in counterregulatory responses, including glucagon, further contribute to recurrent hypoglycemia and unawareness.⁷ Our team has also recently identified an important role for serotonin and bile acid signaling in PBH.^8,9

Growth differentiation factor 15 (GDF15), a member of the transforming growth factor β superfamily, is a cellular stress signal that is expressed in and secreted by multiple tissues^10,11 and signals in part via the cell surface receptor GFRAL. Elevated GDF15 levels have been associated with multiple pathological conditions including inflammation, cancer, metabolic disorders, and cardiovascular diseases^12,13,14 and thus have been widely explored as a promising biomarker for disease and a target for therapy.^15,16 Recent data suggest that GDF15 is also regulated in physiologic states, such as exercise¹⁷ and contributes to systemic glucose metabolism, with weight loss, increased insulin action, and decreased plasma glucagon.¹⁸ Whether alterations in GDF15 contribute to lowering of plasma glucose in individuals with hypoglycemia is unknown.

In this study, we aimed to evaluate the potential relationship between GDF15 and glucose metabolism during hypoglycemia in individuals with PBH and in mice with insulin-induced hypoglycemia. We found that GDF15 levels were higher in individuals with PBH both in the postprandial state and during experimental hypoglycemia (hyperinsulinemic hypoglycemic clamp), and that these elevations were associated with higher hypoglycemia symptom scores. Complementary experiments in mice, both with and without bariatric surgery, revealed increased GDF15 at the time of hypoglycemia. Together, these findings suggest that GDF15 is dynamically regulated during hypoglycemia and may contribute to physiological and symptomatic responses associated with PBH.

Results

Clinical characteristics

The clinical characteristics of study participants are provided in Table S1. The cohort included a subset of individuals of a larger study¹⁹ who completed both mixed meal tolerance tests and hyperinsulinemic hypoglycemic clamps: 15 individuals post-RYGB with PBH (14 female [F], 1 male [M]; mean age, 54 ± 11; BMI, 30.4 ± 4; 8 ± 5 years post-RYGB), 15 individuals post-RYGB without symptomatic hypoglycemia (Asx) (15F, 0M; age, 53 ± 9; BMI, 34 ± 7; 11 ± 5 years post-RYGB), and 10 overweight/obese individuals without history of upper gastrointestinal surgery or T2D (Ow/Ob) (6F, 4M; age, 47 ± 13; BMI, 32 ± 5). Hemoglobin A1C did not differ between groups.

Circulating GDF15 levels are increased in the fasting and postprandial state in individuals with PBH

Blood samples were collected from participants in the fasting state and at 30 and 120 min after liquid mixed meal (Figure 1A). In the fasting state, glucose levels did not differ between groups, consistent with prior findings.⁸ However, as expected, postprandial glucose differed in both surgical groups, increasing significantly at 30 min in both PBH (201 ± 43 mg/dL) and Asx (186 ± 32 mg/dL) compared with Ow/Ob (138 ± 18 mg/dL). By 120 min, glucose was significantly lower in PBH (77 ± 30 mg/dL) compared with Ow/Ob (124 ± 31 mg/dL) but did not differ from Asx (90 ± 22 mg/dL) (Figure 1B). However, nadir glucose was significantly lower in PBH vs. Asx (55.4 mg/dL vs. 71.2 mg/dL p = 0.011) (Figure 1C). Although fasting insulin, C-peptide, and GLP-1 levels did not differ between groups, all increased significantly at 30 min in both surgical groups, returning to baseline after 2 h (Figures 1D–1F). Insulin and C-peptide levels at 30 min were numerically higher in PBH vs. Asx and significantly different from Ow/Ob.

Figure 1.

Circulating GDF15 levels increase in individuals with PBH after a mixed meal

(A) Experimental design. Blood samples were collected from 15 individuals with PBH, 15 asymptomatic (Asx) individuals, and 10 overweight/obese (Ow/Ob) controls at 0, 30, and 120 min during a mixed-meal test.

(B) Plasma glucose measured by yellow springs instruments (YSI).

(C) Glucose nadir.

(D–G) Plasma insulin, C-peptide, GLP-1, and GDF15 measured by ELISA.

(H–P) Pearson correlation analyses between GDF15 and plama glucose (H–J), insulin (K–M), and GLP-1 (N–P) at fasting state, 30 min, and 120 min after meal digestion.

Data are represented as mean ± SD. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗p < 0.0001.

Fasting GDF15 levels, measured by ELISA, were significantly elevated in PBH (718 ± 381 pg/mL) compared with Ow/Ob (331 ± 108 pg/mL). Meal intake elicited a significant increase in GDF15 levels only in PBH, evident at 2 h postprandially (p < 0.01). At 30 min post-meal, GDF15 levels remained numerically higher in PBH, but differences were not significant. At 120 min, GDF15 was again significantly higher in PBH (914 ± 397 pg/mL) vs. both Asx (476 ± 188 pg/mL) and Ow/Ob (353 ± 142 pg/mL) (Figure 1G).

To identify potential contributors to increased GDF15, we analyzed correlation between GDF15 and metabolic parameters in both fasting and postprandial states. There was a significant positive correlation between fasting glucose and GDF15 in both PBH (r = 0.77, p < 0.01) and Asx groups (r = 0.79 p < 0.01) (Figure 1H). This pattern did not persist at 30 min (Figure 1I). At 120 min post-meal, there was a significant inverse correlation between glucose and GDF15 only in PBH (r = −0.65, p = 0.03) (Figure 1J). Insulin and GDF15 were not associated in the fasting state (Figure 1K), but were positively correlated at 30 min in PBH (r = 0.83, p < 0.01) (Figure 1L) and at 120 min in Asx (r = 0.68, p < 0.01) (Figure 1M). Similar patterns were observed for GLP-1 and GDF15, with no correlation in the fasting state (Figure 1N), positive correlation in PBH at 30 min (r = 0.60, p = 0.01) (Figure 1O), and positive correlation in Ow/Ob (r = 0.73, p = 0.01) at 120 min (Figure 1P).

Patients with PBH often experience multiple symptoms after meals, both in the early postprandial phase prior to hypoglycemia and with subsequent hypoglycemia. Correlation analysis revealed a significant positive relationship between plasma GDF15 and self-reported symptom of “feeling cold” (r = 0.89 p = 0.01) at 120 min post-meal, exclusively in PBH (Table S2).

GDF15 levels are elevated in PBH individuals during hypoglycemia induced by hyperinsulinemic hypoglycemic clamp

To determine whether GDF15 is modulated by experimentally induced hypoglycemia, we performed hyperinsulinemic hypoglycemic clamps and collected blood samples in the fasting state, at 40 min (hyperinsulinemia, normoglycemia) and 90, 110, and 120 min (hyperinsulinemia, hypoglycemia) (Figure 2A). Glucose did not differ between groups in fasting state (PBH, 90 ± 8 mg/dL; Asx, 93 ± 6 mg/dL; Ow/Ob, 98 ± 14 mg/dL) or during experimental hypoglycemia ranging from 90 to 120 min (e.g., 110 min: PBH, 55 ± 5 mg/dL; Asx, 53 ± 6 mg/dL; Ow/Ob, 51 ± 6 mg/dL) (Figure 2B). Likewise, insulin levels achieved during the clamp were similar (90 min: PBH, 210 ± 60 pg/mL; Asx, 234 ± 99 pg/mL; Ow/Ob, 245 ± 37 pg/mL; Figure S1A) and did not correlate with GDF15 (Figures S1B–S1D). While fasting plasma epinephrine levels did not differ between groups, epinephrine was significantly lower at 90 min in PBH (1.3 ± 0.9 nmol/L) vs. Ow/Ob (4.1 ± 1.5 nmol/L) and Asx (2.4 ± 1.4 nmol/L) (Figure 2C). This pattern was sustained at 110 min (PBH, 1.3 ± 0.9 nmol/L; Asx, 2.5 ± 1.3 nmol/L; Ow/Ob, 4.1 ± 1.5 nmol/L) but not at 120 min. There were no significant differences in norepinephrine (Figure 2D), glucagon, cortisol, or growth hormone (Figures S2A, S2E, and S2I) between groups.

Figure 2.

GDF15 levels are elevated in PBH individuals during hypoglycemia induced by hyperinsulinemic hypoglycemic clamp

(A) Experimental design. Blood samples were collected from 15 PBH, 15 Asx, and 10 Ow/Ob patients at time 0, 40, 90, 110, and 120 min during hyperinsulinemic hypoglycemic clamp.

(B) Plasma glucose.

(C) Plasma epinephrine.

(D) Plasma norepinephrine.

(E) Plasma GDF15.

(F–H) Pearson correlation between glucose and GDF15.

(I–K) Pearson correlation between epinephrine and GDF15.

For all panels, asterisks denote significant difference between PBH vs. Ow/Ob (blue color), between PBH vs. Asx (green), or between PBH vs. Asx (orange). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗p < 0.0001. Data are represented as mean ± SD.

While fasting GDF15 did not differ between groups, GDF15 levels began to rise exclusively in PBH with the onset of hypoglycemia (Figure 2E). At 90 min, GDF15 was significantly elevated in PBH (830 ± 390 pg/mL) compared to Ow/Ob (474 ± 204 pg/mL; p = 0.03) but did not differ from Asx (746 ± 328 pg/mL; p = 0.82). With increasing duration of hypoglycemia, GDF15 increased in PBH and diverged from other groups; at 110 min, GDF15 was significantly elevated in both surgical groups vs. Ow/Ob (PBH, 929 ± 270 pg/mL; p = 0.02; Asx, 744 ± 84 pg/mL; p = 0.01; Ow/Ob, 475 ± 220 pg/mL) but did not differ between PBH and Asx (p = 0.07). At 120 min, GDF15 was significantly elevated in PBH (1,176 ± 122 pg/mL) vs. both Asx (794 ± 100 pg/mL; p = 0.04) and Ow/Ob (531 ± 254 pg/mL; p = 0.04) (Figure 2E).

GDF15 was inversely correlated with glucose during experimental hypoglycemia only in individuals with PBH: 90 min (r = 0.69, p < 0.01), 110 min (r = 0.85, p < 0.01), and 120 min (r = 0.99, p = 0.03) (Figures 2G, 2H, and S3A). GDF15 was also significantly inversely correlated with epinephrine in PBH at 90 min (r = 0.71, p < 0.01) (Figure 2J) and 110 min (r = 0.75, p < 0.01) (Figure 2K). Similarly, GDF15 was inversely correlated with norepinephrine only at 110 min in PBH (r = 0.59, p = 0.01) (Figure S3F).

Prior studies in mice indicate that oral delivery of free fatty acids (FFAs) contributes to the regulation of GDF15.^20,21 Our prior studies⁸ demonstrated that meal ingestion decreased the abundance of lipid species clusters, consistent with insulin-stimulated repression of lipolysis. In the current cohort, we tested whether alterations in circulating FFAs in either the fasting or hyperinsulinemic state, in the absence of oral intake, would be associated with GDF15. Fasting FFA concentrations did not differ among groups (PBH: 0.41 nmol, Asx: 0.39 nmol, Ow/Ob: 0.28 nmol). During the hyperinsulinemic clamp, insulin infusion markedly suppressed FFAs at 40, 90, and 110 min in all groups, with reductions of 44% in PBH, 56% in Asx, and 57% in Ow/Ob, p < 0.05 as expected (Figure S4). However, there was no significant association between FFAs and GDF15 in either the fasting state or any clamp time point in PBH.

GDF15 levels are associated with higher scores on the Edinburgh Hypoglycemia Symptom Scale during the hypoglycemic clamp

To evaluate whether elevated GDF15 levels contribute to hypoglycemia symptoms, we assessed the relationship between GDF15 and symptoms reported on the Edinburgh Hypoglycemia Symptom Scale (EHSS) during experimental hypoglycemia. GDF15 levels strongly associated with total EHSS scores, exclusively in the PBH group at 40 min (r = 0.70, p < 0.01), 90 min (r = 0.62, p = 0.01), 110 min (r = 0.87, p < 0.01), and 120 min (r = 1.0, p < 0.01) (Figure 3; Table S3). Individual symptom components of the score, including dizziness (r = 0.53, p = 0.04) and difficulty concentrating (r = 0.53, p = 0.04), were positively correlated with fasting GDF15 in PBH. During the hyperinsulinemic clamp, several symptoms were significantly correlated with GDF15 in PBH even at 40 min (before hypoglycemia developed), including weakness, heat sensation, tiredness, dizziness, and tingling of lips (all p < 0.05). Notably these relationships were observed when glucose was stable and normal (mean 87.7 ± 10 mg/dL) but insulin levels were high (mean 228 ± 75 pg/mL). These associations between symptoms and plasma GDF15 persisted during the hypoglycemia portion of the clamp, including drowsiness, difficulty concentrating, and feeling cold at 110 min and weakness and tingling of lips at both 90 and 110 min (all p < 0.05). By contrast, no association between GDF15 levels and hypoglycemia symptoms was observed in either the Asx or Ow/Ob groups (Figure 3; Table S3). These findings suggest a consistent association between GDF15 levels and symptoms in PBH, both at baseline and during progression of insulin-induced hypoglycemia.

Figure 3.

Heatmap showing correlations between GDF15 levels and hypoglycemia symptoms (Edinburgh Hypoglycemia Symptom Scale) during the hyperinsulinemic hypoglycemic clamp

An asterisk (∗) indicates a significant correlation between the symptom and GDF15 levels during the hypoglycemic clamp (Pearson correlation, p < 0.05). Data include 15 individuals with PBH, 15 Asx, and 10 Ow/Ob, assessed at 0, 40, 90, and 110 min.

GDF15 levels are increased in mice with insulin-induced hypoglycemia

To determine whether GDF15 levels are altered during insulin-induced hypoglycemia in mice, we injected chow-fed lean mice with insulin (2 U/kg) or saline and collected blood at time 0, 15, and 30 min later (Figure 4A). By 15 min, glucose was significantly lower in insulin-treated mice (105 ± 22 mg/dL) vs. saline-treated mice (233 ± 52 mg/dL, p < 0.01), and this difference was magnified at 30 min (insulin: 79 ± 18 mg/dL, saline: 199 ± 17 mg/dL; p < 0.01) (Figure 4B). GDF15 levels did not differ at baseline but were significantly elevated in insulin- vs. saline-injected mice at 15 min (43 ± 5 pg/mL vs. 25 ± 2 pg/mL; p = 0.03); this difference was even greater at 30 min (insulin: 84 ± 23 pg/mL, saline: 45 ± 4 pg/mL; p = 0.04 (Figure 4C). While there was no correlation between glucose and GDF15 in the fasting state (Figure 4D), there was a strong inverse correlation in insulin-treated mice (r = 0.90, p = 0.03) at 30 min (Figure 4F).

Figure 4.

GDF15 levels are increased with insulin-induced hypoglycemia, and exogenous GDF15 inhibits food intake after meal-stimulated hypoglycemia

(A) Experimental design for (B and C) (n = 10 per group).

(B and C) (B) Blood glucose and (C) plasma GDF15.

(D–F) Pearson correlation between glucose and GDF15.

(G) Schematic of the insulin-augmented mixed meal tolerance test (n = 5 per group).

(H and I) (H) Blood glucose and (I) body weight after GDF15 or saline treatment.

(J) Blood glucose in male mice (n = 5 per group).

(K) Food intake during insulin-induced hypoglycemia in male mice.

(L) Blood glucose in female mice (n = 5 per group).

(M) Food intake during insulin-induced hypoglycemia in female mice.

In all panels, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001 by two-way ANOVA with Tukey’s multiple-comparison test. Data are presented as mean ± SD from two independent experiments.

Since GLP-1 is a dominant contributor to increased postprandial insulin levels in PBH, we assessed the impact of the GLP-1 receptor agonist semaglutide on GDF15. Semaglutide reduced the glycemic response to exogenous glucose, and this effect was blocked by avexitide, a GLP-1 receptor antagonist, as expected (Figure S5A). However, neither semaglutide nor avexitide had any measurable effect on GDF15 concentrations (Figure S5B). This aligns with existing data indicating that GDF15 and GLP-1 agonists act through distinct neuronal pathways to induce weight loss in mice.²²

We recently demonstrated that serotonin levels are increased in patients with PBH and that exogenous serotonin in mice can induce hypoglycemia.⁸ In mice, serotonin-induced hypoglycemia was blocked by ketanserin, a serotonin receptor 2 antagonist (Figures S6A–S6E). Notably, serotonin-induced hypoglycemia was accompanied by a significant increase in plasma GDF15 compared to saline-treated Ow/Ob (Figure S6C). Ketanserin also suppressed the associated rise in GDF15 (Figure S6F). These findings suggest that elevated GDF15 is linked to hypoglycemia triggered by either exogenous insulin or serotonin.

We also assessed whether GDF15 responses differed in high fat-fed mice subjected to vertical sleeve gastrectomy (VSG) or sham surgery. During oral gavage of a dextrose-supplemented mixed meal, VSG mice exhibited significantly higher GDF15 levels than sham-operated controls at 60 min (p < 0.02; Figure S7A). Moreover, those mice that developed hypoglycemia after the meal had even higher GDF15 levels vs. sham (p < 0.01; Figure S7B). Together, these findings indicate that VSG increases post-meal GDF15, mirroring the heightened GDF15 levels observed in human PBH.

GDF15 reduces food intake response to hypoglycemia

To determine whether acute GDF15 administration could contribute to postprandial glucose dynamics and exaggerated insulin secretion characteristic of PBH in mice, we performed an insulin-augmented mixed meal tolerance test (Figure 4G). Acute administration of GDF15 did not change glucose levels vs. saline-injected mice (Figure 4H).

We next asked whether GDF15 modulated food intake response to hypoglycemia, a key element of the counterregulatory response. To induce more robust hypoglycemia, male mice were again fasted for 4 h and subsequently administered a higher dose of insulin (2 U/kg). Once blood glucose levels reached ≈60 mg/dL, we injected either recombinant GDF15 (rGDF15) (0.5 mg/kg) or saline and provided a pellet of chow in the cage. Saline-treated mice exhibited a steady rise in blood glucose, whereas rGDF15-treated mice experienced a significant delay in hypoglycemia recovery (Figure 4J). Similar results were observed in females (Figure 4L). To determine whether delayed recovery from hypoglycemia was related to rGDF15-mediated alterations in food intake, we assessed food intake. While food intake increased in saline-injected mice, both male and female mice injected with GDF15 had a marked and significant reduction in food intake (males: absent food intake at 105, 120, and 150 min, p < 0.05, with cumulative food intake of 1.5 g in GDF15-injected vs. 3.6 g in saline-injected mice (p < 0.01) (Figure 4K). Similarly, in females, food intake was completely absent at 60 and 120 min and reduced by 57% and 66% at 90 and 150 min, respectively (p < 0.01 for all); cumulative food intake was markedly lower in GDF15 vs. saline (1.2 vs. 3.6 g, p < 0.01) (Figure 4M). Acute reductions in food intake likely also contributed to a 14% lower body weight at 72 h after GDF15 injection in males (p = 0.002) (Figure 4I).

Discussion

Increasing evidence points to GDF15 as a stress-responsive cytokine elevated in humans in a variety of conditions, including cardiac and renal failure, liver disease, cancers, and mitochondrial diseases.²³ Increases in GDF15 have been observed as early as 1 week after both gastric bypass and sleeve gastrectomy forms of bariatric metabolic surgery,^24,25,26 with sustained increases over time and higher levels associated with greater weight loss.²⁵ We now demonstrate that individuals with PBH have significantly higher plasma GDF15 after a mixed meal. GDF15 was inversely correlated with glucose and positively correlated with both insulin and GLP-1 levels at 30 min post-meal. In addition, GDF15 levels increased exclusively in PBH participants even after intravenous insulin-induced hypoglycemia, suggesting that stimulation of GDF15 is not fully dependent on engaging the gastrointestinal tract.

PBH is a multifaceted clinical syndrome, and our findings reflect this complexity. Nadir glucose levels during a standardized liquid mixed meal test were consistently lower in individuals with clinically diagnosed PBH as compared with asymptomatic post-bariatric participants (55.3 vs. 69.7 mg/dL). We recognize that glucose alone is unlikely to explain the full spectrum of PBH symptoms; nearly half of postprandial symptoms occur without hypoglycemia in post-bariatric populations.²⁷ These observations highlight the need to define PBH-specific physiological pathways—beyond glycemia—that contribute to symptom generation and disease heterogeneity.

We do not yet understand why PBH would be associated with more robust increases in GDF15 in response to both oral and intravenous provocation of hypoglycemia in comparison to other post-surgical patients. One potential factor common to both experimental paradigms would be insulin; plasma insulin is higher in PBH than in other groups in the post-meal state and marked hyperinsulinemia is present during the hyperinsulinemic clamp (but similar between groups). Moreover, insulin administration in mice increases GDF15, even before severe hypoglycemia is achieved. GDF15 has been shown to regulate insulin secretion in human pancreatic islets²⁸ and to enhance insulin action in hepatic and adipose tissue,¹⁸ indicating a potential interplay between hyperinsulinemia and increased GDF15 in PBH.

Additional contributors to higher levels of GDF15 in PBH could include other peptides or metabolites. For example, oral delivery of fatty acids has been shown to increase expression of GDF15 in several tissues, including intestinal segments in mice, and to increase levels in the portal vein, suggesting that luminal intestinal stimuli could increase systemic levels of GDF15.^20,21 This remains an intriguing possibility, but our prior studies show that lipid species are coordinately downregulated in the post-meal state in humans with PBH, consistent with meal suppression of lipolysis.⁸ In the present study, we found no association between plasma FFA and GDF15 during experimental hyperinsulinemic euglycemia, in the absence of complicating meal stimuli. Taken together, these data do not suggest that circulating FFAs are a likely driver of PBH-specific increases in GDF15 in the fasting, post-meal, or hyperinsulinemic hypoglycemic conditions. However, these data do not exclude the possibility that oral lipid sensing by the intestine could be altered after bariatric surgery and contribute to increases in plasma GDF15. Additional studies will be needed to test this possibility.

Additional factors have recently been demonstrated to be exclusively perturbed in PBH, including increases in postprandial GLP-1, bile acids, serotonin, and other intestinally derived factors.^8,9 We recently demonstrated that exogenous serotonin induces hypoglycemia in mice; this effect is blocked by ketanserin, a serotonin receptor 2 antagonist.⁸ Interestingly, serotonin increases GDF15 levels in mice. We do not know whether this results from a direct effect of serotonin administration, induction of insulin secretion, or low blood glucose. Since GDF15 is recognized as a stress sensor, it is possible that elevated serotonin or downstream signaling may induce physiological stress resulting in increased GDF15. This is supported by the efficacy of ketanserin to block both serotonin-induced hypoglycemia and serotonin-induced increases in GDF15 in mice. Finally, it is possible that patients with PBH have increased overall sensitivity to induction of GDF15 in response to multiple diverse stimuli, but additional studies will be required to evaluate this possibility.

Our data in humans do not permit assessment of whether increases in GDF15 are a marker of or a contributor to hypoglycemia. Acute exogenous administration of rGDF15 did not affect glucose levels in mice, suggesting that it is unlikely to be a key direct contributor to acute hypoglycemia. Rather, GDF15 may impact a key element of the counterregulatory response to hypoglycemia. Normally, hypoglycemia triggers a robust increase in food intake, contributing to normalization of glucose levels. rGDF15 administration during insulin-induced hypoglycemia suppressed food intake during hypoglycemia in both male and female mice. Our findings align with prior studies in mice demonstrating that rGDF15 induces sickness-like behavior, alters food intake, and leads to reduced body weight.^29,30,31 GDF15 may also modulate additional components of the counterregulatory response; GDF15 has been shown to not only reduce plasma glucagon¹⁸ but also increase insulin secretion and action.^18,32 This constellation of responses would be expected to decrease net postprandial hepatic glucose production,¹⁸ a key contributor to PBH.³³ The relative contribution of increased GDF15 to each of these mechanisms to sustain hypoglycemia in the postprandial state in PBH will need to be evaluated in future experiments.

Prior studies have reported that plasma GDF15 is increased in multiple conditions,^24,34,35 and elevated levels are linked to symptoms such as anorexia, nausea, and fatigue.^16,36,37 Our findings also suggest a potential relationship between GDF15 and the multiple symptoms experienced by patients with PBH, particularly early in the postprandial state when hypoglycemia has not yet developed. Interestingly, only individuals with PBH exhibited a strong positive correlation between GDF15 and higher scores on the EHSS, particularly for symptoms such as weakness, difficulty concentrating, and cold sensation. While some of these symptoms can occur at the time of concurrent hypoglycemia, these symptoms often also are reported by patients when plasma glucose values are falling but not yet low. It is important to recognize that these symptoms are not specific for hypoglycemia and may instead reflect the impact of additional postprandial metabolic changes characteristic of PBH, including rapid gastric emptying and altered splanchnic flow and high levels of insulin, incretins, serotonin, and GDF15. Thus, we hypothesize that GDF15 may contribute to at least some of the postprandial symptom complex in PBH. Intriguingly, acute activation of neurons expressing the GDF15 receptor GFRAL not only reduces food intake and body weight in mice, potentially via nausea-associated autonomic and behavioral responses, but also reduces body temperature.³⁸ Further studies are needed to explore parallels with postprandial symptoms in human PBH and to define the contribution of GDF15.

In summary, our study identifies GDF15 as a potential biological marker of the hypoglycemic state in both humans and mice, positioning it as a sensor of glucose metabolism and a component of the counterregulatory response. Moreover, elevated GDF15 levels were associated with higher postprandial scores on the EHSS in individuals with PBH, linking this cytokine to hypoglycemia-related postprandial symptomatology. Together, these results provide insight into the role of GDF15 in hypoglycemia and its potential implications for metabolic regulation.

Limitations of the study

This study has several limitations. Recruitment of individuals with PBH, asymptomatic post-surgical controls, and non-surgical comparators willing to undergo two prolonged metabolic assessments (mixed meal testing and hyperinsulinemic hypoglycemic clamp) was challenging, resulting in a modest sample size that limits generalizability. In addition, additional studies will be required to elucidate mechanisms underlying the exaggerated GDF15 responses to both oral and intravenous hypoglycemic stimuli in PBH.

Resource availability

Lead contact

Requests for further information, resources, and reagents should be directed to and will be fulfilled by lead contact, Mary-Elizabeth Patti (mary.elizabeth.patti@joslin.harvard.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

All data reported in this paper will be shared by the lead contact upon request. This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

We thank the participants in the human studies and the excellent support from Clinical Research Center nursing staff. We gratefully acknowledge funding from the National Institutes of Health, including R01DK121995 (to M.-E.P. and D.A.S.), K26DK138368 (to D.A.S.), and P30DK036836 (Diabetes Research Center).

Author contributions

Conceptualization, R.F.-B. and M.-E.P.; methodology, R.F.-B., L.P., H.W., B.O., P.C.Q., A.G.d.C., T.M.C., H. Saeed, L.P., H. Saifeldin, L.O., C.C., M.F., A.A., A.S., D.C.S., D.A.S., and M.-E.P.; investigation, R.F.-B., L.P., H.W., B.O., P.C.Q., A.G.d.C., T.M.C., H. Saeed, L.P., H. Saifeldin, L.O., C.C., M.F., A.A., A.S., D.C.S., D.A.S., M.-E.P.; visualization, R.F.-B.; funding acquisition, M.-E.P. and D.A.S.; supervision, M.-E.P.; writing – original draft, R.F.-B. and M.-E.P.; and writing – review and editing, all authors.

Declaration of interests

The authors declare that no conflict of interest exists.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Biological samples
Human plasma	This paper	N/A
Mouse plasma	This paper	N/A
Chemicals, peptides, and recombinant proteins
Regular insulin (insulin human)	Lilly	Humulin R
rGDF15	PeproTech	120-28C-100UG
Semaglutide	Cayman Chemical	40231
Avexitide	Eiger Biopharmaceuticals	10020
Serotonin hydrochloride	Sigma-Aldrich	H9523-100MG
Ketanserin	Millipore&Sigma	S006-50MG
DPP IV Inhibitor	Millipore	DPP4
Critical commercial assays
Human GDF15 ELISA Kit	R&D systems	DGD150
Mouse GDF15 ELISA Kit	R&D systems	MGD150
Human Cortisol Competitive ELISA Kit	Invitrogen	EIAHCOR
Human Growth Hormone Quantikine ELISA Kit	R&D systems	DGH00
Glucagon ELISA Kit	Mercodia	10-1271-01
Free Fatty Acid Colorimetric Assay Kit	Abcam	ab282927
Experimental models: Organisms/strains
C57bl/6J mice	Jackson Labs	000664
Software and algorithms
GraphPad Prism v. 9.2.0	GraphPad	https://www.graphpad.com/features
BioRender	BioRender	N/A
Other
Ensure Compact – (liquid mixed meal - MMTT)	Abbott	64356
Prodigy AutoCode®	Prodigy Diabetes Care	N/A
High Fat Diet (HFD)	Research Diets	D12492
DietGel® Boost	Clear H2O	72-04-5022

Experimental model and study participant details

Participants

Participants were recruited between 2020 and 2023 into a study aimed at understanding additional metabolic features of post-bariatric hypoglycemia. Participants with PBH were recruited from the Hypoglycemia Clinic at Joslin Diabetes Center. Participants in the nonsurgical control group and in the post-bariatric asymptomatic group were recruited by advertisement via social media. We recruited 15 individuals with PBH (14 females and 1 male; mean age 54 ± 11 years), 15 Asx (all females; mean age 53 ± 9 years), and 10 individuals with Ow/Ob (6 females and 4 males; mean age 47 ± 13 years). This recruitment strategy may have biased toward recruitment of more severely affected individuals with PBH given that they presented for evaluation at the specialized hypoglycemia clinic. All participants provided written informed consent. The study was approved by the Joslin Diabetes Center Committee on Human Investigation CHI and registered on clinical trials.gov (NCT04428866).

Animals

All animal studies were approved by the Institutional Animal Care and Use Committee. C57BL/6J male and female mice, age 6 weeks, were purchased from JAX. Mice were maintained on a 12 h light-dark cycle with ad libitum access to water and Purina 9F chow diet. The experiments were approved by the IACUC at Joslin Diabetes Center.

Method details

Liquid mixed meal test

Blood samples were obtained after an overnight fast and at 30 and 120 min after consumption of a liquid mixed meal (MMTT) (2 bottles of Ensure Compact (64 g carbohydrates, 18 g protein, volume 236 mL) over 10 min).

Hyperinsulinemic hypoglycemic clamp

A primed, continuous infusion of insulin (80 mU/m²/min) was initiated at time 0, together with a 20% glucose infusion, titrated every 5 min to gradually lower the glucose from fasting levels to a target of 50 mg/dL at a rate of 1 mg/dL/min,³⁹ the target glucose was maintained for 30 min to ensure a consistent stimulus for counterregulatory responses. Blood samples were obtained after an overnight fast, and at 40, 90, 110 and 120 min after the start of the hyperinsulinemic hypoglycemic clamp.

Animal experiments

Vertical sleeve gastrectomy

C57Bl/6J male mice were singly housed and fed a 60% high fat diet (HFD; Research Diets) for 8 weeks to induce obesity before surgery. HFD was removed the day before surgery, and mice were given Diet Gel Boost. On the day of surgery, mice were anesthetized via isofluroane inhalation and given buprenorphine SR and carprofen for pain management. For VSG surgery, a laparotomy was performed, the abdominal cavity was flushed with saline, and approximately 80% of the stomach along the greater curvature was clamped via LigaClips and stabilized with sutures. For sham surgery, laparotomy and abdominal flushing were performed similarly, without ligating the stomach. The abdominal cavity was then closed with sutures and mice recovered on a warm heating pad. For three days post-operation, mice were given fresh diet gel, carprofen for pain, and gentamycin to prevent infection. Three days after surgery HFD was resumed.

Insulin tolerance testing

To determine whether insulin could impact GDF15 levels in mice, mice were fasted for 4 h and then injected intraperitoneally with insulin (2U/kg, Humulin R, Lilly, Indianapolis, IN, USA) or saline. Blood was collected via tail vein at time 0 and 15 min and 30 min post-injection.

Meal testing

To determine whether acute GDF15 administration could contribute to postprandial glucose dynamics and exaggerated insulin secretion characteristic of PBH in mice, we performed an insulin-augmented mixed meal tolerance test after a 4 h fast. rGDF15 (0.5mg/Kg, PeproTech, Cranbury, NJ) or saline was injected 30 min later. For studies in VSG mice, mice were fasted for 6 h prior to gavage with 200 μL of Ensure with 30 mg dextrose added.

GDF15 injection

To assess the effects of GDF15 on the counterregulatory response to hypoglycemia, both male and female mice were fasted for 4 h, after which blood glucose levels were measured every 15 min using a glucometer. When blood glucose reached approximately 60 mg/dL, mice received an intraperitoneal injection of recombinant human GDF15 (rGDF15; 0.5 mg/kg; PeproTech, Cranbury, NJ) or an equal volume of saline control. Immediately following injection, a pre-weighed chow pellet was provided and food intake was recorded at 15 min intervals for up to 2 h to assess feeding behavior and glycemic recovery. All experiments were conducted during the light phase (12:00–16:00) to minimize circadian influences.

To evaluate whether GDF15 levels could be modulated by GLP-1 action, mice were treated with semaglutide (3 mmol/kg; Cayman Chemical) and/or avexitide, a GLP-1 receptor antagonist (30 mmol/kg; Eiger Biopharmaceuticals). Mice were pretreated with intraperitoneal (i.p.) avexitide or an equal volume of saline 2 h prior to injection with either semaglutide or saline and intraperitoneal glucose. Blood glucose levels were measured 15 min after semaglutide or saline administration.

Metabolic assays

Plasma collected at baseline (fasting) and at 30 and 120 min following the mixed meal, or plasma collected at baseline (fasting) and at 40 and 100 min after initiation of the hyperinsulinemic hypoglycemic clamp was assayed by human GDF15 ELISA (R&D systems, Minneapolis, MN, USA, DGD150) following the manufacturer’s instructions.

In humans, plasma glucose was measured using a YSI analyzer; in mice, tail vein blood glucose was measured with a Prodigy AutoCode meter.

Plasma insulin and C-peptide were quantified using ELISA, and epinephrine and norepinephrine concentrations were quantified by LC/MS at the University of Pennsylvania Diabetes Research Center Core. Cortisol levels were measured using a commercial ELISA kit (Invitrogen, EIAHCOR), human growth hormone (hGH) using the Human Growth Hormone Quantikine ELISA Kit (R&D Systems, DGH00) and glucagon using the Mercodia Glucagon ELISA (10-1271-01), following manufacturers’ instructions.

Plasma collected from mice at fasting (time 0) and subsequently at 15, 30 and 60 min after insulin injections was assayed by GDF15 ELISA (R&D systems, Minneapolis, MN, USA, MGD150). All sampled blood was collected via tail vein or cardiac puncture in EDTA-coated microtubes. All assays were performed according to manufacturers’ instructions.

Blood from animals subjected to VSG or Sham sugery were collected from the tail in EDTA tubes at baseline and 60 min after mixed meal gavage and plasma was collected after centrifugation at 4°C. GDF15 ELISA (R&D systems, Minneapolis, MN, USA, MGD150) was performed on plasma samples according to manufacturer’s instructions.

Quantification and statistical analysis

Power calculation: Sample size and power were evaluated using two-sided tests with α = 0.05. With the available cohort sizes (PBH, n:15; asymptomatic post-RYGB, Asx, n:15; overweight/obese, Ow/Ob, n:10), the study has 80% power to detect large effect sizes. Specifically, detectable standardized effects were Cohen’s d ≥ 1.06 for PBH vs. Asx and d ≥ 1.19 for comparisons involving Ow/Ob. Thus, the study is powered to detect large between-group differences, while smaller effects are considered exploratory.

Between-group comparisons used two-sided tests. Clamp studies with repeated measurements were analyzed using linear mixed-effects models including group, time and group-by-time interaction, with participant as a random effect. Single time-point outcomes were analyzed using ANOVA, followed by prespecified pairwise comparisons. Correlations were assessed using Pearson methods.

All results were presented as the mean ± standard desviation (SD). Differences in the means among multiple groups were analyzed using two-way analysis of variance (ANOVA) and for analysis of two groups, the Student’s t test was performed using GraphPad Prism 10.0. Significance was defined as p ≤ 0.05 (∗), p ≤ 0.01 (∗∗), and p ≤ 0.001 (∗∗∗).

Additional resources

This study was registered at ClinicalTrials.gov (identifier: NCT04428866). Additional information regarding the trial is available: https://clinicaltrials.gov/study/NCT04428866?term=NCT04428866&rank=1.

Published: March 9, 2026