Sitagliptin reduces FAP-activity and increases intact FGF21 levels in patients with newly detected glucose abnormalities

INTRODUCTION
Fibroblast growth factor 21 (FGF21), a hormone with pleiotropic metabolic effects, is inactivated by fibroblast activation protein (FAP), a member of the dipeptidyl peptidase-IV (DPP-IV) family. We investigate if sitagliptin (DPP-IV inhibitor) inhibits FAP-activity and increases intact FGF21.


METHODS
Patients with impaired glucose metabolism were randomized to 100 mg sitagliptin (n = 34) or placebo (n = 37) treatment for 12 weeks. Plasma samples obtained at study entry and at 12-weeks were analysed for FAP-activity, FAP, total FGF21 and intact FGF21.


RESULTS
Sitagliptin significantly inhibited FAP-activity (497 ± 553 vs. 48 ± 712 RFU/min, p < 0.01) and correspondingly increased intact FGF21 (253 ± 182 vs 141 ± 80 ng/L, p < 0.01) compared to placebo in plasma. Sitagliptin dose-dependently inhibited the FAP-activity in vitro. Intact FGF21 was higher in patients obtaining a normal glucose tolerance regardless of treatment (p = 0.03).


CONCLUSION
A sitagliptin-induced increase of intact FGF21 may contribute to an improved metabolic effect in patients with impaired glucose metabolism.


Introduction
The role of Fibroblast activation protein (FAP) in the metabolic regulation attract attention, due to its ability to cleave and inactivate fibroblast growth factor 21 (FGF21). FGF21 is a hormone with pleiotropic effects on glucose and fat metabolism and cardio protection (Angelin et al., 2012;Dunshee et al., 2016;Planavila et al., 2013Planavila et al., , 2015Staiger et al., 2017;Yan et al., 2015). Intact FGF21 is required for FGF21 signalling, which occurs through N-terminal binding to an FGF receptor (FGFR) and C-terminal binding to the co-receptor β-klotho (Micanovic et al., 2009), both highly expressed in adipose tissue (Petryszak et al., 2016) and in the central nervous system (Bookout et al., n.d.). Reduced levels of KLB in tissue-specific KLB-knock out mice impairs the FGF21 action (BonDurant et al., 2017;Owen et al., 2014). FAP shares more than 50% homology with the clinically validated drug-target dipeptidyl peptidase-IV (DPP-IV), and like DPP-IV FAP exists as a circulating serine protease (Lee et al., 2006;Zhen et al., 2016a). FAP exhibit both dipeptidyl peptidase activity and endopeptidase activity, capable of both N and C-terminal cleavage. The activity of FGF21 is regulated by FAP via proteolytic cleavage between Pro171 and Gly172 in the C-terminal part of FGF21 averting β-kslotho binding and thereby preventing FGF21 signalling through the FGFR (Sánchez-Garrido et al., 2016).
Total serum FGF21 levels are found to be elevated in patients with cardiovascular disease (Chow et al., 2013;Lin et al., 2010;Shen et al., 2013), type 2 diabetes (T2D) (Chavez et al., 2009;Keuper et al., 2020) and reported as a marker of disease severity in T2D patients with cardiovascular complications Lin et al., 2014). Additionally, the activity of FAP has been reported to correlate with patient Abbreviations: acute coronary syndrome, (ACS); acute myocardial infarction, (AMI); dipeptidyl peptidase IV, (DPP-IV); Fibroblast activation protein, (FAP); Fibroblast growth factor 21, (FGF21); impaired glucose tolerance, (IGT); oral glucose tolerance test, (OGTT); Relative fluorescent unit, (RFU); type 2 diabetes, (T2D). ☆ The protocol was registered at clinicaltrials.gov (NCT00627744). outcome, disease severity and susceptibility to treatment for diseases such as fibrosis and atherosclerosis (De Decker et al., 2019;Lay et al., 2019;Stein et al., 2021). If FAP-activity and total FGF21 increase with disease severity, FGF21 may be upregulated but cleaved and lose its function leading to decreased level of active FGF21. High levels of intact FGF21, either by administration of recombinant hFGF21 or by FAP inhibition, leads to decreased plasma glucose, lower triglyceride levels, less hepatic fat, and increased insulin sensitivity in animal models (Kharitonenkov et al., 2005;Kliewer and Mangelsdorf, 2019). Since FAP belongs to the same enzyme family as DPP-IV, we hypothesize that treatment with a DPP-IV-inhibitor may inhibit the FAP-activity and increase the circulating levels of intact, and therefore bioactive, FGF21. In this study, we investigate our hypothesis in patients with a recent hospitalization for acute coronary syndrome (ACS) and impaired glucose tolerance treated with or without sitagliptin.

Participants
From the double-blinded randomized clinical trial, the BEta-cell function in Glucose abnormalities and Acute Myocardial Infarction (BEGAMI), post ASC patients in a stable phase 4-23 days (median 6 days) after hospital admission were screened by oral glucose tolerance test (OGTT). The patients were randomly allocated with a block size of four via a computer-generated randomization sequence and a 1:1 ratio to receive either 100 mg sitagliptin (n = 34; JanuviaTM; Merck Sharp & Dohme AB, USA) or placebo (n = 37) once daily for 12 weeks. A detailed description of the study has been presented elsewhere (C. . Based on OGTT results the patients were assigned as impaired glucose tolerant (IGT), type 2 diabetes (T2D) or normal glucose tolerant according to WHO recommendations (Camilla  at study entry and at follow-up. The study was approved by the ethics committee at Karolinska Institutet, Sweden, and performed in accordance with the Declaration of Helsinki. The protocol was registered at clinicaltrials.gov (NCT00627744). Written informed consent was obtained from all subjects before inclusion to the study. Compliance to study drug, estimated by pill counts, was 100% for the sitagliptin group and 99% for the placebo group.

Biochemical analyses
Fasting blood samples were collected at study entry and after 12 weeks. Samples for each individual were analysed in duplicates and within the same plate. Plasma glucose and insulin, HbA1c levels, cholesterol, HDL, LDL and triglycerides was analysed by routine (C.  and plasma insulin was measured by ELISA (DAKO, Cambridgeshire, UK). Intact GLP-1 was measured by a commercial ELISA kit (ab184857, Abcam) according to the instructions provided by the manufacturer. The intra-and interassay variation (%CV) were below 7 and 9%, respectively, and the LOD was 25 pg/ml.

Total FGF21 and intact FGF21
Total FGF21 levels were quantified using an in-house time-resolved immunofluorometric (TRIFMA) assay as previously described (Lauritzen et al., 2017). The limit of detection (LOD) was 1 pg/ml and intra-and interassay variations (%CV) were below 6 and 10%, respectively.
The physiological function of FGF21 relies on an active FGF21 protein, and the latter was measured by a commercial sandwich ELISA kit (F2131-K01, Eagle Bioscience) according to the instructions provided by the manufacturer; with this assay, one antibody specifically binds an epitope in the N-terminal part of human FGF21 while the second antibody is specific for the C-terminal human FGF21. The intra-and interassay (%CV) were below 6 and 7%, respectively, and the LOD was 1.7 pg/ml.

FAP and FAP-activity
FAP levels were measured using monoclonal antibodies (R&D System DY3715) modified into an in-house TRIFMA assay as previously described (Arlien-Søborg et al., 2020). The LOD was 50 pg/L and the intra-and interassay variation (%CV) were below 7% and 9%, respectively.
FAP-activity was determined using an in-house modified fluorescence resonance energy transfer (FRET) technique in which fluorescence signal is released when FAP cleaves the FGF21 specific substrate (Nterminal HyLite488-Val-D-Ala-Pro-Ser-Gln-Gly-C-terminal lysine conjugated QXL520, Kaneka Eurogentec S.A., Belgium) between the fluorescence donor HyLite488 and the quencher QXL520 (Bainbridge et al., 2017). Plasma samples were diluted 10-fold in assay buffer (HEPES; pH 7.2, 150 mM NaCl, 1 mM EDTA, 0.1 mg/mL BSA) and added in duplicates on a black 96-well plate (Thermo Scientific Nunc, #165305) followed by substrate (1 μM) in each well. The plate was placed on a shaker in 37 • C incubator for 5 min before the first measurement (t0). Fluorescence was measured at 37 • C using PerkinElmer Multimode reader EnVision (EnVision Manager version 1.13.3009 1401) at (t0) and after 1 h (t60). For blank correction relative fluorescence unit (RFU) values obtained at t0 was subtracted from RFU (t60) for each well, this corrected RFU was divided by 60 min for RFU/min. The LOD was 66 RFU/min and the intra-and interassay (%CV) were below 9% and 10%. Human EDTA plasma samples (anonymized samples for assay development with no clinical data available) were used for assay validation and quality controls. Recombinant human FAP, rhFAP, (RnD System) and EDTA plasma dilutions were used for dose-dependency and linearity. Additionally, paired EDTA-plasma and serum samples showed equal FAP-activity allowing for measurements in both serum and plasma, and the assay showed stability up to 10 freeze-thaw cycles.

In vitro FAP inhibition by sitagliptin
The bioavailability of sitagliptin has previously been determined to 87%, peaking 4 h after oral administration of a single dose of 100 mg, corresponding to a C max of 0.95 μM in plasma (Bergman et al., 2007).
Twelve different human plasma samples, serving as internal control for quality assurance and optimization of the FAP-activity assay, were mixed with sitagliptin phosphate (Biotechne, UK) at clinically relevant doses of 0 μM, 1 μM, 5 μM as well as supraphysiological doses of 15 μM and 25 μM sitagliptin. FAP-activity was determined as described above.

Insulin resistance and glucose tolerance
OGTT and frequently sampled intravenous glucose tolerance test (FSIGT) was performed at study entry and at the 12-week visit, as previously described (C. . The beta cell function was assessed by the insulinogenic index (IGI) calculated as Δinsulin 0-30 /Δglucose 0-30 obtained from the OGTT (C. . The Homeostatic Model Assessment (HOMA) was used to estimate insulin resistance HOMA-IR: (fasting serum insulin (mU/ml) × fasting plasma glucose (mmol/L)/22.5). Acute insulin response to glucose (AIRg) was calculated from the FSIGT as the incremental area under the curve from 0 to 10 min and the glucose disappearance constant (K g ) as the slope of the natural logarithm of the difference between the glucose samples at 10-and 20-min K g = (Δ ln plasma glucose/Δ min) ⋅100.

FGF21 cleavage site prediction
A cleavage site prediction analysis for FGF21 was perform using the PeptideCutter tool available at expacy.org (Godfrey et al., 2005). This tool searches a protein sequence and predicts potential cleavage sites for a list of proteases and chemicals. The sequence of the peptide in interest, in this case the 180aa mature circulating FGF21 sequence, was entered into the tool and the potential cleavage sites were identified along with a list of the enzymes or chemicals responsible for the theoretical cleavage.

Statistical analysis
Normal distributions were evaluated by QQ-plots and histograms. Data with normal distribution are expressed as means ± SD, non-normal distribution as medians [interquartile range], and categorical variables are expressed as numbers and percentages. Paired t-tests were used to evaluate normal distributed baseline and follow-up data, whereas paired samples Wilcoxon tests were used for non-normal distributed data. Unpaired t-tests were used to compare the treatment effects (Δ-values) between the groups, the treatment effects (Δ-values) are reflected by the changes in variables (V) of interest, calculated as delta (Δ) = V (12-weeks) -V (study entry) . One patient in the sitagliptin group had total FGF21 levels above 14500 ng/L at both timepoints and were excluded as an outlier (sitagliptin (n = 33)). The FSIGT analysis was available in 66 patients (sitagliptin (n = 31) and placebo (n = 35)) and paired intact GLP-1 levels were available in 49 patients (sitagliptin (n = 25) and placebo (n = 24)). Correlations are presented as Pearson R 2 for parametric variables and Spearman rho (R) for non-parametric variables and p-values. Intact FGF21 delta-values, stratified into groups based on OGTT results at baseline/follow-up (IGT/IGT, IGT/normal, IGT/T2D, T2D/T2D, T2D/ IGT and T2D/normal). These groups and the in vitro data were examined using a Dunnett's test. P < 0.05 was considered statistically significant. All statistics were conducted using RStudio version 1.2.5019.

Results
Patient characteristics at study entry and after 12 weeks treatment with sitagliptin or placebo are shown in Table 1. The groups were matched for age and BMI and did not differ in metabolic or hormonal analyses at baseline (Arnetz et al., 2015).

Intact FGF21 levels are higher in patients with improved glucose tolerance
After 12 weeks, 59% of the patients (26 patients from the sitagliptin group and 15 from the placebo group), who were at study entry stratified by OGTT as either IGT or T2D, had obtained normal glucose tolerance and these patients presented with increased levels of intact FGF21 (Fig. 2). A significant increase in intact FGF21 was observed in the patients who improved their glucose tolerance from T2D to normal (T2D/normal) at 12 weeks as compared with those who showed no improvement (T2D/T2D). For post-hoc testing, the groups were stratified according to the OGTT assigned group at study entry; IGT or T2D, allowing a true reference group (the group shifting to normal). No significant differences vs. the reference group were observed in the group assigned IGT at study entry. No change in either total FGF21 nor FAP-activity were found when stratifying by group shift (data not shown).

Levels of intact FGF21 correlated with FAP-activity but not with metabolic parameters
The FAP-activity was negatively correlated with intact FGF21 after  Fig. 2. Treatment effect on intact FGF21 (Δ = V(12 weeks) -V(study entry)) stratified by group shift. The group shift was based on changes in OGTT result from study entry (IGT or T2D) to 12-week follow-up (normal, IGT or T2D). A significant increase in intact FGF21 was found for the group with the largest OGTT improvement ("T2D to normal") compared to the group that did not improve in OGTT ("T2D to T2D") (Dunnett's test p = 0.03).
12 weeks in the placebo group (R 2 = − 0.48, p = 0.0025) but not in the sitagliptin treated group (Fig. 3a). We found no correlation between treatment effects of intact FGF21 (Δ-values) with any of the metabolic improvements found after sitagliptin treatment ( Table 2). The change in adjusted FAP-activity/FAP level (Δ-values) correlated positively with the difference in HbA1c (R = 0.39, p = 0.03) in the placebo group. No other correlations between treatment effects and FAP-activity/FAP level were found (Table 2). Intact GLP-1 correlated positively with intact FGF21 after treatment with sitagliptin (Fig. 3b), however not significantly, most likely due to the reduced sample size, as we had insufficient amount of sample-volume for the GLP-1 analysis (placebo group n = 24 and sitagliptin group n = 25. No correlations were found with the OGTT results or HOMA-IR.

FAP-activity assay
At study entry total FAP levels and the FAP-activity correlate positively (R 2 = 0.41, p < 0.001) (Fig. 4a). FAP-activity increased dosedependently with addition of rhFAP or with higher plasma concentration (Fig. 4b). However, the rhFAP was less enzymatically active than the endogenous plasma FAP, and supraphysiological levels of rhFAP, 10fold higher, were needed to reveal this effect (Fig. 4b). Prediction of theoretical cleavage sites for the mature human FGF21 (180 aa) was performed using the database expasy.org (https://web.expasy. org/peptide_cutter/). The intracellular oligopeptidase prolyl endopeptidase PREP, the DPPIV-enzyme most similar to FAP, was the only other enzyme on the list that could in theory cleave the FGF21 sequence. The database also generated a list of enzymes not able to cleave human FGF21, including caspases 1-10, Factor Xa, Granzyme B (GRAB), proline endo peptidase (PPCE), Thrombin and Trypsin.
A similar effect was found with a fixed amount of sitagliptin in different plasma dilutions, or with addition of rhFAP, however the latter was only found in high concentrations (data not shown).

Discussion
12 weeks of sitagliptin treatment inhibits plasma FAP-activity in patients with a recent ACS and impaired glucose regulation. Correspondingly, increased level of circulating intact, and thus bioactive, FGF21 was available as compared with patients treated with placebo. The level of circulating FAP protein was not affected by sitagliptin treatment compared to placebo, revealing that only the FAP-activity was altered by sitagliptin treatment. The fact that an increased FAP-activity was precent in the sitagliptin group at study entry may even underestimate the reported inhibitory effect of FAP by sitagliptin. Additionally, total FGF21 levels were similar between the groups, highlighting the necessity of including measurements of intact FGF21.
FAP-activity is reported to correlate with the concentration of the FAP protein (Zhen et al., 2016b), which our data supports. DPP-IV and FAP exhibit overlapping dipeptidyl peptidase activity and cleave a subset of similar substrates, e.g. neuropeptide Y, substance P and brain natriuretic peptide (Keane et al., 2011). The endopeptidase activity of FAP results in cleavage of denatured collagen type I and III, α2 antiplasmin and FGF21 (Lee et al., 2006;Sánchez-Garrido et al., 2016;Zhen et al., 2016a). Despite their similarities, FAP does not inactivate GLP-1 and DPP-IV does not cleave and inactivate FGF21, supporting that FAP is the only enzyme in the circulation that cleaves FGF21 and that FGF21 is not a substrate for DPP-IV (Dunshee et al., 2016). In addition, we performed a cleavage site prediction analysis using expasy.org, which revealed a theoretical cleavage site for the intracellular oligopeptidase prolyl endopeptidase PREP, the DPP-IV-enzyme most similar to FAP, but no other enzymes found in blood circulation (included in the database) were capable of cleaving the 180 aa human FGF21 sequence. However (Coppage et al., 2016), showed that human FGF21 is only digested by FAP but not PREP. We use a highly FAP specific substrate, which is not cleaved by PREP and thus our results support that FAP is responsible for the of cleaves FGF21 as shown in Fig. 3a and b. Like GLP-1, FGF21 is highly susceptible to cleavage and inactivation. Intact FGF21 is required for simultaneous binding to FGFR1 and β-klotho (Micanovic et al., 2009) to obtain FGF21 cell-signalling (Sánchez-Garrido et al., 2016), thus elevation of intact FGF21 is an attractive new treatment strategy.
Assessment of specific FAP-activity is challenging as many of the activity-based probes lack selectivity with respect to FAP related peptidases, especially the endopeptidase prolyl oligopeptidase (PREP) (De  Decker et al., 2019). In order to determine the FGF21 specific FAP-activity we have taken advantage of the FRET technology using an FGF21 peptide including a slightly modified FAP cleaving site as a substrate (Bainbridge et al., 2017). Bainbrigde et al. elegantly showed, that substitution of the P2 Gly with D-Ala in the FGF21 peptide, revealed a FAP-specific substrate, not cleaved by any of the other DPP-family Table 2 Correlations between the treatment effect for intact FGF21 or FAP-activity/FAP level and selected variables. Treatment effects are evaluated as Δ = V(12 weeks) -V (study entry). Correlations are presented with Pearsons R 2 or Spearman rho (R) and p-values. Spearman correlations are indicated with s. Oral Glucose Tolerance Test (OGTT) was used for area under the curve (AUC), Insulinogenic Index (IGI) and Homeostatic Model Assessment for Insulin Resistance (HOMA-IR), whereas Frequently Sampled Intravenous Glucose Tolerance test (FSIGT) was used to calculate the Acute Insulin Response to glucose (AIRg) and the glucose disappearance constant (Kg). Body Mass Index (BMI), glycohemoglobin (HbA1c). *P < 0.05.  . 4. A) circulating FAP levels correlated to FAP activity at study entry (R = 0.41, p < 0.001). B) Dose-dependent increase of FAP-activity in plasma (a two-fold dilution of human plasma (100 μg/L) starting at 50% plasma) and in rhFAP (a two-fold dilution starting at a concentration of 440 μg/L).

Fig. 5.
A dose-dependent inhibition of FAP-activity in vitro from control dose of 0 μM sitagliptin to dose of 1 μM or 5 μM was not significant. However, a significant decrease in FAP-activity between control and 15 μM sitagliptin (Dunnett test p = 0.01) was identified and a further decrease in FAP-activity between control and 25 μM sitagliptin was also significant (Dunnett test p < 0.00001).
proteases, including DDP-IV and PREP (Bainbridge et al., 2017;De Decker et al., 2019). We observed a dose-dependent effect of sitagliptin added to a plasma sample from healthy individuals. However, the concentrations of sitagliptin needed for FAP inhibition in vitro were higher than the reported plasma bioavailable concentration (Herman et al., 2005). This might reflect a difference in a local short-term (plasma samples were pre-incubated with sitagliptin for 1h prior to FAP-activity analysis) or a systemic long-term (12-weeks oral treatment) exposure to sitagliptin. We found a negative correlation between FAP-activity and intact FGF21 in the placebo treated group supporting earlier findings that higher FAP-activity leads to decreased functional FGF21 . This was not observed in the sitagliptin group possibly due to an increased FAP-activity in this group at study entry and therefore lack of complete inhibition in some patients.
FAP levels increased, or in fact normalized in all BEGAMI patients after 12 weeks as compared to study entry. This is in alignment with other studies and may be explained by recovery from ACS, since decreased FAP levels has been associated with myocardial damage, followed by increased FAP levels 1-4 months after hospitalization for ACS, which indicate a dynamic change of FAP in the acute event (De Willige et al., 2017;Tillmanns et al., 2013Tillmanns et al., , 2017. Interestingly, FGF21 may be secreted by cardiomyocytes as an autocrine factor and potentially protect the heart from adverse cardiac events and prevent the heart from pathological remodelling after a myocardial infarction (Joki et al., 2015;Planavila et al., 2015). Higher FAP and FGF21 levels are reported in obese as compared to lean individuals, but no correlation with BMI was reported (Zhen et al., 2016a), which our data support. Despite the reported beneficial effect of FGF21 on the glucose and lipid metabolism, circulating FGF21 are elevated in obesity (Chavez et al., 2009;Zhang et al., 2008), T2D (Chavez et al., 2009;Keuper et al., 2020) and in patients with non-alcoholic fatty liver disease (NAFLD) (Barb et al., 2019;Dushay et al., 2010) compared to healthy individuals. The role of FGF21 has been intensively studied, however, due to the pleiotropic effects of FGF21 a precise mechanism is yet to be established. Most studies measure the total level of FGF21 and therefore overestimate the levels of intact FGF21, although the intact FGF21 is reported to follow a similar pattern as total FGF21 (Samms et al., 2017). Our results strongly indicate the necessity of measuring intact FGF21 and/or FAP-activity when assessing FGF21 signalling.
Our observations in clinical samples are supported by pre-clinical data. Sanchez-Garrido et al. showed that talabostat, a potent DPP-IV inhibitor, inhibited FAP activation in mice (Sánchez-Garrido et al., 2016). The talabostat-treated mice had improved metabolic conditions and elevated total and intact plasma FGF21 in diet-induced obese mice, however not in lean mice. Additionally, a pharmacological inhibition of FAP were found to increase FGF21 levels 3-fold in monkeys (Dunshee et al., 2016;Zhen et al., 2016a). Also, mice with diet-induced NAFLD treated with sitagliptin had increased FGF21 levels and reduced circulating lipids as compared to the non-treated group (Bilin et al., 2017). In contrast, in vitro studies found no inhibition of FAP-activity by sitagliptin, however this was analysed with rhFAP expressed in baculovirus and by cleavages of Ala-Pro− /− amido-4-trifluromethylcoumarin (Lee et al., 2008). Of note, the FAP monomer has five potential N-linked glycosylation sites, which are necessary for FAP endopeptidase activity (Sun et al., 2002). Additionally, a dimerization of FAP is required for the enzymatic action, whereas the FAP monomer is inactive (Aertgeerts et al., 2005), which may also explain the lower FAP-activity we found in rhFAP as compared to the endogenous FAP. Together these models support our findings in humans and reveals future treatment strategies.
Exogeneous administration of FGF21 analogues have shown to improve lipid profiles and glycaemic control, reduce insulin levels and lead to weight loss in rodents (Staiger et al., 2017), however, clinical studies have shown inconsistent effects in humans (reviewed by (Kliewer and Mangelsdorf, 2019)). The short half-life of the exogeneous FGF21 may result from proteolytic degradation mediated by FAP.
Recently, a modified FAP-resistant FGF21 analogue (AKR-001/AMG876) have shown promising metabolic effects in humans (Kaufman et al., 2020), indicating that degradation of FGF21 by FAP needs to be in focus. A different approach to increase the endogenous FGF21 activity may thus be to inhibit FAP-activity, e.g., by sitagliptin as reported here. Two novel FAP inhibitors, BR102910, a chemical compound with high FAP inhibitory effect and blood glucose reduction in mice (Jung et al., 2021), and BR103354 with increased intact FGF21 level in ob./ob. mice and in cynomolgus monkey (Cho et al., 2020), supports this strategy.
Importantly, we showed that patients who improved their glucose tolerance had higher levels of intact FGF21. This finding is supported by in vitro studies in adipocytes and liver cells, which reported that FGF21 signalling stimulate glucose uptake through upregulation of GLUT-1 (Ge et al., 2011;Liu et al., 2018). We did not find any association between intact FGF21 and the rate of disappearance for glucose, insulinogenic index or beta cell function, although these metabolic parameters were significantly improved in the BEGAMI patients after treatment with sitagliptin (C. . However, we found a positive correlation between delta HbA1c and the adjusted FAP-activity/FAP level in the placebo group. This indicates that FAP-activity is associated with HbA1c, supporting the hypothesis that FAP-activity and reduced active FGF21 is involved in the pathogenesis of impaired glucose tolerance. Interplay of FGF21 and GLP-1 in hepatic glucose metabolism has been reported (Liu et al., 2019). The GLP-1 analogue, Liraglutide, were found to stimulate FGF21 mRNA expression in both liver and adipose tissue, as well as increased circulating FGF21 plasma levels, in adiponectin deficient/ApoE knockout mice on a high-fat-diet (Yang et al., 2012) whereas the GLP-1R antagonist, Exenatide, reduced the total FGF21 levels, which correlated negatively with fasting insulin (Hu et al., 2016). Interestingly, the two large randomized clinical trials; LEADER (Liraglutide) and EXSCEL (exenatide) showed different effects of GLP-1 receptor agonist treatment on cardiovascular protection. The LEADER study showed that Liraglutide reduced the risk of cardiovascular events compared to placebo, whereas the EXSCEL study showed no difference in cardiovascular outcomes between exenatide and placebo (Holman et al., 2017;Marso et al., 2016). One could speculate if this was associated with the ability of liraglutide to increase FGF21 and exenatide to reduce the levels of FGF21. Our data support, that in addition to inhibition of DPP-IV and thus improving endogenous GLP-1 function, treatment with sitagliptin also increase the levels of intact FGF21 (Table 1), through inhibition of FAP-activity, which may in combination lead to improved OGTT. In mice, the increased levels of intact GLP-1 and FGF21 in combination were reported to improve insulin sensitivity and thereby improve oral glucose tolerance (Yang et al., 2012). Recently (Liu et al., 2019), reported that GLP-1 stimulate hepatic FGF21 production in vivo in db/db mice and in vitro in human HepG2 cells and showed that FGF21 and GLP-1 participate in hepatic glucose metabolism, thus suggesting a new glucose-lowering mechanism of GLP-1. The complementary functions of GLP-1 and FGF21 are highly beneficial for T2D patients and dual-treatment have recently been suggested (Gilroy et al., 2020;Pan et al., 2021) with beneficial effect in mice and a treatment effect superior to GLP-1 or FGF21 alone. Accumulating knowledge suggest that FGF21 and GLP-1 may cross-talk and provide synergism, at least in rodents and in vitro. We do see a trend supporting a positive correlation between intact GLP-1 and intact FGF21 after treatment with sitagliptin, but we do not find a combined effect on glucose tolerance or HOMA-IR. Thus, additional studies including more patients are needed to determine an additive effect of increasing both intact GLP-1 and intact FGF21.
The BEGAMI study was not designed to investigate FAP-activity nor changes in FGF21 levels and taking the well-known inter-individual difference in human FGF21 levels into account (Gälman et al., 2008), a clinical translation of our findings may have benefitted from a larger sample size. A longer treatment period with sitagliptin may have reduced the FAP-activity even further and stimulated an improvement in the metabolic endpoints. The patients in this study were recovering from ACS and we do not know how this affects FAP-activity and FGF21. An improvement in glucose tolerance (evaluated by OGTT) was observed in both groups, possibly affected by similar lifestyle counselling. However, the improvement was considerably higher in the sitagliptin group as compared to the placebo group (Table 1).

Conclusion
Sitagliptin treatment prevented FAP-activity increase in response to ACS and increased levels of intact FGF21 in a placebo-controlled design. Higher levels of intact FGF21 were found in patients with improved glucose tolerance. Inhibition of FAP may be an alternative approach to increase endogenous intact FGF21 levels to obtain better metabolic regulation. Metabolic improvements achieved by DPP-IV inhibitor, e.g., sitagliptin treatment, may arise from a dual inhibition of DPP-IV and FAP, resulting in increased levels of both intact GLP-1 and intact FGF21 in circulation.

Declaration of interest
CH: consulting fees from Novartis and Roche Diagnostics, research grants from Roche Diagnostics and speaker honoraria from MSD and Novartis.
AKNP, NJ and MB have no conflict of interest.

Funding
The study was supported by a grant from the Novo Nordisk Foundation (NNF17OC0029532), A.P. Møller Fonden and Aarhus University.

Data availability
The data that has been used is confidential.