Introduction

Biliopancreatic diversion with duodenal switch (BPD) is the most effective bariatric procedure to induce substantial long-term weight loss and type 2 diabetes remission [1,2,3]. It is reported that between 96 and 100% of individuals undergoing BPD will be free of diabetes at 12 months [3]. Interestingly, reversion of diabetes has been observed as early as 1 week after BPD [4]. This could be explained by enhanced hepatic and adipose tissue insulin sensitivity as well as improved insulin secretion rate (ISR) and perhaps β-cell glucose sensitivity and function in the first few days after BPD [4,5,6,7]. Increased muscle insulin sensitivity is observed later, at 3 months postoperatively [6]. Thus, in the absence of clear mechanisms, several hypotheses have been put forward to explain the early improvement in glucose homeostasis after BPD that is observed before significant weight loss occurs [7]. These include profound caloric restriction [8], increased incretin response, change in bile acid composition and flow, alteration of gut microbiota and intestinal glucose metabolism as well as increased brown adipose tissue metabolic activity [9].

Another possible mechanism that has not been yet explored is the contribution of the bone-derived hormone uncarboxylated osteocalcin (ucOCN). Animal studies have revealed that bone is an endocrine organ able to regulate insulin secretion and insulin sensitivity via ucOCN, the bioactive form of osteocalcin (OCN) [10,11,12]. Indeed, OCN knockout mice display reduced β-cell mass and insulin secretion, and decreased insulin sensitivity [12, 13]. OCN is secreted in its intact carboxylated form by the bone-forming cells (osteoblasts). The process of decarboxylation of OCN into ucOCN requires the acidity of the bone milieu, which is provided by the activity of the bone-resorbing cells (osteoclasts) [11, 14]. In humans, an inverse association between gamma-carboxylated OCN/total OCN ratio (a reflection of the metabolically inactive form of OCN) and several indices of insulin resistance has been corroborated in a cross-sectional study [15]. There is however very limited evidence associating ucOCN and glucose homeostasis in humans due to the unavailability of robust measurement methods for ucOCN [15,16,17,18,19].

In light of preclinical studies and cross-sectional human studies suggesting that bone may alter glucose homeostasis via ucOCN [10,11,12, 15], and of our previous findings demonstrating that bone turnover (mainly bone resorption) increases rapidly and dramatically after BPD even prior to significant weight loss [20], coupled with the availability of a new robust assay to measure directly ucOCN in humans [21], we aimed to explore whether the increase in bone resorption and ucOCN contributes to the improvement in glucose homeostasis observed after BPD. We hypothesized that (1) ucOCN increases after BPD; (2) the increase in ucOCN correlates with the increase in the bone resorption marker C-terminal telopeptide (CTX); (3) the increase in ucOCN and CTX correlate with improved indices of insulin sensitivity, insulin secretion and β-cell function after BPD.

Materials and methods

Study design and population

This is an ancillary study from a 1-year prospective observational study that aimed at evaluating fatty acid metabolism during type 2 diabetes remission after BPD [6]. The study population included 16 men and women aged 18–65 years with severe obesity (BMI ≥ 35 kg/m2), and with or without type 2 diabetes (according to the American Diabetes Association’s criteria) [22]. All participants underwent BPD at the Institut universitaire de cardiologie et de pneumologie de Québec (IUCPQ) between 2012 and 2014. Details on the BPD procedure have been published previously [23]. The 1-year frozen sample of one participant was missing. Moreover, two individuals did not complete the 1-year follow-up due to diagnosis of low-grade breast cancer after the 3-month visit (n = 1) or pregnancy during follow-up (n = 1), leaving 13 participants who completed the study.

Main exclusion criteria were: pharmacological treatment with fibrate, insulin, thiazolidinedione, beta-blocker, corticosteroid, oral contraceptive pill, hormone replacement therapy, treatment for osteoporosis, or any other medication known to affect bone or lipid metabolism, uncontrolled thyroid disease, decompensated or severe renal or hepatic disease other than non-alcoholic fatty liver disease, or other serious medical condition or surgical complication following BPD. None of the participants received warfarin. Before surgery, participants were prescribed vitamin D3 to ensure normal preoperative 25-hydroxyvitamin D (25OHD) levels. After BPD, a standard prescription was given to all participants, including daily multivitamin (Centrum Forte®), calcium carbonate 1000 mg, vitamin D2 50,000 IU, vitamin A and iron. Doses were adjusted based on blood results at 3 months and 12 months. As reported previously, 25OHD remained unchanged at 3 days and 3 months after surgery, but then increased significantly at 12 months after surgery (+61%, p < 0.0001). Mean (SD) 25OHD levels were 59.6 (8.3) at baseline, 51.5 (8.7) at 3 days, 67.8 (20.0) at 3 months and 94.4 (33.8) at 12 months after surgery. Besides, parathyroid hormone was within the normal range for all participants at baseline. It increased significantly at 3 days after surgery (+87%, p < 0.0001) but then remained within normal limits for all participants at 3 and 12 months after surgery [20]. The IUCPQ ethical review board approved the protocol. Consent form was signed by all participants before entering the study.

Outcome measurements

Outcome measurements were assessed before surgery (45 ± 8 days before surgery) and at 3 days, 3 months and 12 months after BPD.

Uncarboxylated osteocalcin and bone turnover markers

At each timepoint, 12-h fasting early morning frozen EDTA plasma was analyzed in a single batch to determine levels of CTX (intra-assay CV 1.8–4.5% and inter-assay CV 2.5–6.5%) and 1–43/49 (N-terminal and intact) OCN (intra-assay CV 0.7–2.4% and inter-assay CV 2.0–4.1%) with an automated electrochemiluminescence assay (Elecsys, Roche Diagnostics). ucOCN was assessed with a recently developed Sandwich Enzyme-Linked Immunosorbent Assay recognizing human ucOCN, but not carboxylated OCN (intra-assay CV 3.1–4.1% and inter-assay CV 7.1–8.1%) [21].

Glucose homeostasis markers

Fasting concentrations of glucose (hexokinase method, Roche Glucose Analyzer) and insulin (Luminex LX200 multiplex immunoassays, Millipore, Billerica, MA) were assessed. Details of the methods to assess glucose homeostasis markers have been described in the original paper and summarized here [6]. HOMA-IR, a marker of hepatic insulin resistance, was calculated by multiplying fasting plasma insulin (U/L) by fasting plasma glucose (mmol/L)/22.5 [24]. Each participant completed a 6-h stable isotopic tracer infusion study in which a euglycemic-hyperinsulinemic clamp was maintained for 4 h. The first 2 h was an equilibration period during which subjects received a continuous perfusion of glycerol and palmitate, as previously described [6]. High-dose insulin perfusion (1.2 mUI/kg/min) started 120 min later, and patients received variable glucose infusion to achieve and maintain euglycemia (5 mmol/L) over 4 h. Peripheral insulin sensitivity index (SI), which represents muscle sensitivity to insulin, was determined during the 4-h euglycemic-hyperinsulinemic clamp. Adipose tissue insulin resistance index (ADIPO-IR), which represents adipose tissue resistance to insulin [25], was calculated by multiplying whole-body rate of appearance of glycerol by fasting insulin levels, during the euglycemic-hyperinsulinemic clamp with and without intravenous fat loading [6, 26]. ISR was determined during a hyperglycemic clamp where patients received graded intravenous glucose infusion, reaching and sustaining for 2 h a peak plasma glucose of 10 mmol/L [6, 27]. Insulin disposition index (DI), an indicator of pancreatic β-cell function [28, 29], was assessed by multiplying SI by ISR. Bodyweight was measured using a bioimpedance scale (InBody520, body composition analyzer, Biospace, Los Angeles, California) calibrated for subjects with severe obesity at baseline, 3 months and 12 months. Measurements were obtained in light clothing to the nearest 0.1 kg. Height was measured with a wall-mounted stadiometer to the nearest 0.1 cm. BMI was calculated and reported in kg/m2.

Statistical analyses

Statistical analyses were performed using JMP Pro version 12.1.0 and SAS version 9.4 (SAS Institute Inc, Cary, NC, USA). A general linear mixed model with a variance component covariance structure for repeated measures was used to compare concentrations as well as their changes from baseline in ucOCN, bone remodeling markers and glucose homeostasis indices at each timepoint. The chosen matrix was based on minimal Akaike information criterion. The subject-specific intercept was considered a random effect, and time was considered a fixed effect. When necessary, log transformation of the variables was performed to ensure normality of distribution. The Tukey–Kramer adjustment was used for multiple comparison tests, comparing all timepoints versus all other timepoints. Due to the small sample size, spearman correlations were performed to determine the associations between changes in ucOCN, bone remodeling markers and glucose homeostasis indices, with or without adjustment for weight loss at the specified time of measurement. For all statistical analyses, changes were similar between patients with or without a diagnosis of type 2 diabetes at baseline (data not shown). Data are presented as mean ± SD, unless stated otherwise. A two-tailed p value < 0.05 was considered statistically significant for all analyses.

Results

Study population

Study population included eleven men and five women with a mean age of 41.6 ± 8.8 years and a mean BMI of 49.4 ± 5.6 kg/m2. At baseline, 40% of the women (n = 2) were postmenopausal. Besides, 11 individuals were diagnosed with type 2 diabetes and treated with diet, metformin or a sulfonylurea, as monotherapy or in combination, and 5 did not have diabetes but were insulin-resistant based on HOMA-IR.

Changes in total bodyweight, ucOCN and bone remodeling markers after BPD

Details of the changes in total bodyweight, CTX and OCN (not ucOCN) have been published previously [20]. Briefly, total bodyweight was not measured at 3 days after BPD but was presumably stable. However, it decreased by 26 ± 7 kg (−19%, p < 0.0001) at 3 months and 59 ± 12 kg (−43%, p < 0.0001) at 12 months. Change in CTX was inversely associated with weight loss at 3 months (r = −0.63, p = 0.009) and 12 months (r = −0.58, p = 0.039). Changes in ucOCN and OCN did not correlate with weight loss. Descriptive analyses of ucOCN and the bone remodeling markers are detailed in Table 1. As shown in Fig. 1, while the individual change in ucOCN at 3 days after BPD was variable (8% reduction on average, p = 0.986), the increase in ucOCN was universal and impressive at 3 months (+257%, p < 0.0001) and 12 months (+498%, p < 0.0001). Similarly, OCN first declined (−19%, p = 0.012) 3 days after BPD, and then rose to a lesser extent than ucOCN at 3 months (+69%, p < 0.0001) and 12 months (+164%, p < 0.0001). CTX increased significantly at 3 days (+66%, p < 0.0001) after BPD, and continued to rise at 3 months (+219%, p < 0.0001) and 12 months (+295%, p < 0.0001) after surgery.

Table 1 Descriptive statistics of bodyweight, ucOCN and bone turnover markers before and at 3 days, 3 months and 12 months after BPD
Full size table
Fig. 1
figure 1

Change in ucOCN at 3 days, 3 months and 12 months after BPD. ucOCN uncarboxylated osteocalcin

Full size image

Changes in glucose homeostasis indices after BPD

Descriptive analyses of the glucose homeostasis indices are shown in Table 2. Fasting plasma glucose decreased 3 days after BPD (−15%, p < 0.0001) and continued to decrease at 3 months (−25%, p < 0.0001) and 12 months (−32%, p < 0.0001). In parallel, insulin concentrations fell as early as 3 days after BPD (−53%, p < 0.0001), and continued to decline throughout follow-up (−71% at 3 months; −84% at 12 months, all p < 0.0001). Both HOMA-IR (−61% at 3 days; −81% at 3 months; −92% at 12 months; all p < 0.0001) and ADIPO-IR (−80% at 3 days; −82% at 3 months; −83% at 12 months; all p < 0.0001) declined at every timepoint after BPD. SI and DI were unchanged at 3 days but increased at 3 months (SI + 194%; DI + 213%; all p < 0.0001) and remained elevated 12 months after surgery (SI + 242%; DI + 120%; all p < 0.0001). While ISR did not change significantly at either 3 days or 3 months after BPD, it decreased significantly at 12 months (−45%, p = 0.003).

Table 2 Descriptive statistics of glucose homeostasis indices before and at 3 days, 3 months and 12 months after BPD
Full size table

Correlations between changes in ucOCN, bone remodeling markers and glucose homeostasis indices after BPD

Significant correlations between changes in ucOCN, bone turnover markers and glucose homeostasis indices before and after adjustment for weight loss are shown in Table 3. Change in ucOCN was positively associated with change in CTX at 3 months and 12 months (Fig. 2). Moreover, change in ucOCN correlated negatively with change in fasting insulin and HOMA-IR, and positively with change in SI at 3 days after BPD (Fig. 3). At 3 months, change in ucOCN was negatively associated with ADIPO-IR and HbA1c (Fig. 3). Furthermore, change in CTX correlated negatively with change in ADIPO-IR at 3 months and at 12 months, and positively with change in DI at 3 months. None of the glucose homeostasis indices correlated significantly with OCN. Adjustment for weight loss at the specified time of measurement did not affect the results.

Table 3 Significant spearman correlations between changes in ucOCN, bone remodeling markers and glucose homeostasis indices before and after adjustment for weight loss at the specified time (3 days, 3 months, 12 months)
Full size table
Fig. 2
figure 2

Non-adjusted Spearman correlations between change in ucOCN and change in CTX at 3 months (a) and 12 months (b) after BPD. ucOCN uncarboxylated osteocalcin; CTX C-terminal telopeptide

Full size image
Fig. 3
figure 3

Non-adjusted significant Spearman correlations between change in ucOCN and a change in fasting insulin, b change in HOMA-IR and c change in SI, at 3 days after BPD, and d change in HbA1c and e change in ADIPO-IR at 3 months after BPD. ucOCN uncarboxylated osteocalcin; HOMA-IR Homeostasis Model Assessment of insulin resistance; SI Insulin sensitivity index; HbA1c glycated hemoglobin; ADIPO-IR adipose tissue insulin resistance index

Full size image

Discussion

In the present study, we explored for the first time the association between changes in ucOCN, bone remodeling markers and glucose homeostasis indices following bariatric surgery. We found that even though mean ucOCN levels did not change significantly 3 days after BPD, individual changes were variable and were associated with changes in hepatic and muscle insulin sensitivity. However, ucOCN increased in all participants at 3 months and continued to rise at 12 months. Furthermore, we observed significant associations between the change in ucOCN and improvement in hepatic and muscle insulin sensitivity indices at 3 days, as well as with adipose tissue insulin sensitivity and HbA1c at 3 months after BPD, independent of weight loss. These findings are consistent with previous studies in mice and humans [12, 15, 30, 31], and support a potential role of ucOCN in the regulation of whole-body insulin sensitivity after BPD. In line with our hypothesis, ucOCN correlated positively with CTX, suggesting that bone resorption is required to decarboxylate OCN [14, 30]. Besides, the increase in CTX correlated with the improvement in adipose tissue insulin resistance and pancreatic β-cell function. Altogether, these findings suggest a potential implication of ucOCN in the improvement of glucose homeostasis after BPD, that may be driven by bone resorption.

We could find four cross-sectional studies [16, 17, 32, 33] and one prospective study [31] that examined associations between ucOCN levels and glucose homeostasis in humans. However, the assay used to measure ucOCN in two of these studies [16, 31] tends to overestimate ucOCN levels [15] while the assays in the other studies have not been validated. Nevertheless, the first study showed that lower ucOCN levels were associated with increased fasting insulin and HOMA-IR in 98 obese patients [16] while the second study found a positive correlation between ucOCN levels and insulin secretion indices in 50 Japanese patients with type 2 diabetes [17]. Conversely, the other cross-sectional studies found no correlations between ucOCN levels and insulin resistance indices in 129 patients with type 2 diabetes assessed with a euglycemic-hyperinsulinemic clamp [32], as well as in women with gestational diabetes [33]. Finally, low baseline concentrations of ucOCN were significantly associated with an increased risk of developing type 2 diabetes at 5 years [31]. Furthermore, in line with our results, Ferron et al. found that the metabolically inactive gamma-carboxylated form of OCN/intact OCN ratio correlated negatively with insulin sensitivity assessed by the euglycemic-hyperinsulinemic clamp or SI assessed by the oral glucose tolerance test [15]. Despite some inconsistencies among studies, which could be attributed to different assay methods and study populations, most analyses support that increased ucOCN, or reduced gamma-carboxylated OCN, is associated with greater insulin sensitivity, insulin secretion and possibly lower type 2 diabetes risk.

To our knowledge, no study addressed specifically whether changes in ucOCN are associated with glucose homeostasis markers after bariatric surgery or medical weight loss. Our findings support the existence of a bone-pancreas-adipose tissue endocrine loop regulating whole-body glucose homeostasis in humans, as previously described in animals. Indeed, Ferron et al. demonstrated that insulin signaling in osteoblasts decreases the expression of osteoprotegerin [34]. This reduction in osteoprotegerin activates bone resorption, which creates a highly acid environment in the bone milieu that facilitates OCN decarboxylation and the release of ucOCN. Activation of bone resorption is thus a necessary step to increase bioactive ucOCN, which then stimulates insulin secretion in pancreatic β-cells and enhances adiponectin expression in adipose tissue, a well-known insulin-sensitizing hormone in muscle and liver [35,36,37]. Indeed, we found a significant correlation between changes in ucOCN and changes in high molecular weight adiponectin at 3 days after BPD (r = 0.57, p = 0.021), but correlations were not significant at 3 months and 12 months. Our hypothesis is also supported by the earlier increase in CTX, a marker of bone resorption, and by the correlation we found between CTX and ucOCN at 3 and 12 months, as well as with several indices of insulin sensitivity. It is not possible to determine from our study the sequence of events leading to increased unOCN and insulin sensitivity. Indeed, the increase in bone resorption may activate decarboxylation of available OCN, increase unOCN and then improve insulin sensitivity. Concomitantly, osteoblasts may secrete more OCN as they become more sensitive to insulin after BPD, similar to what is observed in other insulin-sensitive organs [38,39,40,41]. Hence, it has been shown in animal studies that bone is a site of insulin resistance [42]. Thus, one of the possible mechanisms responsible for the improvement in whole-body insulin sensitivity after BPD may be the increase in ucOCN levels resulting from the activation of bone resorption and the increased bone sensitivity to insulin. Therefore, while the increase in bone resorption resulting from BPD might be detrimental to bone health [43, 44], it may be favorable at the metabolic level by enhancing systemic insulin sensitivity through decarboxylation of OCN.

Regarding the associations that we found between the change in CTX and glucose homeostasis markers, we could find only one study reporting a correlation between the increase in CTX and the reduction in HbA1c 1 year after Roux-in-Y gastric bypass (r = −0.74, p = 0.023) [45]. Moreover, CTX levels were found to correlate positively with insulin sensitivity indices derived from a euglycemic-hyperinsulinemic clamp in 14 overweight and obese individuals, with or without type 2 diabetes [46]. Those results are in line with a few studies suggesting that bone may influence insulin sensitivity, independently of OCN [47, 48]. Moreover, mice studies have shown that bone resorption is modulated by insulin sensitivity in the bone cells, and insulin secretion by the pancreas [34, 42]. Therefore, the findings that changes in CTX correlate with improvements in adipose tissue insulin resistance and DI may reflect those associations. Nonetheless, it also supports our general hypothesis that the increase in bone resorption enhances the bioavailability of ucOCN, which then improves whole-body insulin sensitivity. Besides, it is worth mentioning that we did not find any association between the change in OCN and glucose homeostasis markers after BPD, further supporting that ucOCN is the metabolically-active form of OCN and the importance of accurately measuring it. Our results are in contrast with another study in which higher total OCN levels were significantly associated with improved glucose tolerance, higher SI and lower HOMA-IR in overweight and obese postmenopausal women during an oral glucose tolerance test [15]. The lack of correlation with OCN in our study may be explained by the fact that the assay we used to measure OCN does not detect ucOCN, or by a lack of statistical power compared with the latter study (n = 16 vs. n = 132) [15].

This study explored prospectively for the first time the associations between short- and medium-term changes in ucOCN, bone remodeling markers and glucose homeostasis indices after BPD, measured precisely using gold standard methods. There are limitations that need to be highlighted including the small and heterogeneous study population comprising men as well as pre- and postmenopausal women with and without T2D. Indeed, bone turnover and glucose homeostasis are affected by both diabetes and prediabetes [49] as well as menopausal status [50]. Moreover, the observational nature of our study does not enable us to establish a causal link between ucOCN, bone remodeling markers and glucose homeostasis after BPD.

In conclusion, the increase in ucOCN may be associated with the improvement in insulin sensitivity and glucose homeostasis observed after BPD, independently of weight loss. We hypothesized that this relationship may be driven by bone resorption. Our study lends further support to the existence of endocrine communication between bone and organs involved in glucose metabolism in humans. It also provides additional novel hypotheses that could explain the improvement in glucose homeostasis after bariatric surgery. Our exploratory findings however need to be confirmed in a larger and less heterogenous study population. Moreover, interventional studies should investigate specifically the involvement of ucOCN and bone remodeling in the regulation of glucose homeostasis in the context of bariatric surgery, weight loss, obesity, and diabetes.