Preanalytical impact on the accuracy of measurements of glucagon, GLP-1 and GIP in clinical trials

Abstract Background Plasma concentrations of glucagon, GLP-1 and GIP are reported in numerous clinical trials as outcome measures but preanalytical guidelines are lacking. We addressed the impact of commonly used blood containers in metabolic research on measurements of glucagon, GLP-1 and GIP in humans. Methods Seventeen overweight individuals were subjected to an overnight fast followed by an intravenous infusion of amino acids to stimulate hormonal secretion. Blood was sampled into five containers: EDTA-coated tubes supplemented with DMSO (control), a neprilysin inhibitor, aprotinin (a kallikrein inhibitor) or a DPP-4 inhibitor, and P800 tubes. Plasma was kept on ice before and after centrifugation and stored at −80 Celsius until batch analysis using validated sandwich ELISAs or radioimmunoassays (RIA). Results Measures of fasting plasma glucagon did not depend on sampling containers, whether measured by ELISA or RIA. Amino acid-induced hyperglucagonemia was numerically higher when blood was collected into P800 tubes or tubes with aprotinin. The use of p800 tubes resulted in higher concentrations of GLP-1 by RIA compared to control tubes but not for measurements with sandwich ELISA. Plasma concentrations of GIP measured by ELISA were higher in control tubes and negatively affected by P800 and the addition of aprotinin. Conclusions The choice of blood containers impacts on measurements of plasma concentrations of glucagon, GLP-1 and GIP, and based on this study, we recommend using EDTA-coated tubes without protease inhibitors or P800 tubes for measurements of glucagon, GLP-1 and GIP in clinical trials.


Introduction
Glucagon, glucagon-like peptide-1 (GLP-1) and glucosedependent insulinotropic polypeptide (GIP) have important roles in the development and treatment of obesity and diabetes [1,2].According to clinicaltrials.gov,their plasma concentrations represent outcome measures in more than 500 clinical studies, either as biomarkers for metabolic health or in order to monitor treatment responses.As glucagon, GLP-1 and GIP all circulate in a low picomolar range, it is important to consider preanalytical considerations such as choice of blood containers and how their selection may influence the subsequent analytical analysis platform chosen.Glucagon, GLP-1 and GIP are all substrates of the protease dipeptidyl peptidase 4 (DPP-4), and glucagon and GLP-1 may also be degraded by the protease neprilysin [3][4][5].Measurements of glucagon have for decades been believed to be dependent on the addition of aprotinin (trasylol) to the blood container [6].This belief is based on measurements using single-site, processing-dependent radioimmunoassays (RIA), rather than currently used sandwich ELISAs.Development of commercially available blood containers has allowed standardization potentially reducing preanalytical variability although these containers are expensive and the claim of their superiority to simple inexpensive blood containers (plain EDTA tubes) plasma has not been independently demonstrated or is limited to company announcements.
Plasma concentrations of glucagon, GLP-1 and GIP at fasting have been reported in more than 4000 publications according to a PubMed search (search term: Glucagon OR GIP AND Plasma Levels AND Humans; July 2023), but few experiments have been done to assess the importance of preanalytical factors.We have previously evaluated commercially available assays and also the impact of freeze/thaw and storage temperature on plasma levels of glucagon and GLP-1 [7][8][9].Systematic studies evaluating the influence of the choice of blood containers on plasma levels of glucagon, GLP-1 and GIP have, however, to our knowledge not been performed.
Here, we aimed to establish guidelines regarding containers for blood sampling for measurements of glucagon, GLP-1 and GIP.To do so, we performed a clinical study using five different blood containers containing different enzyme inhibitors (Supplementary Figure 1).Samples were collected from healthy, but overweight volunteers after an overnight fast and during a 40-min intravenous amino acid infusion and plasma was analyzed with 1-3 different immunological methods allowing us to also assess whether the choice of sampling containers would differentially influence the results of the subsequent analytical approach.

Study approvals and ethical considerations
This was a substudy in the GLUSENTIC trial [10] approved by the scientific-ethical committee of the Capital region of Denmark (H-20023717) and registered with the Danish Data protection Agency (P-2021-39) and ClinicalTrials.gov(NCT04907721).Written and oral consent was obtained from all participants, and the study adheres to the principles of the Declaration of Helsinki.
Blood samples were obtained from a cubital vein after an overnight fast (~12 h), and following a 40-min intravenous infusion of amino acids (Vamin Electrolyte Free 14 g/L; Cat.No. B05ABA01; Fresenius Kabi, Copenhagen, Denmark; infusion rate 331 mg/min/kg body weight).Participants rested in a semirecumbent position prior to sampling, performed by experienced technicians.After collection, the (chilled) tubes were mixed gently and kept on ice until centrifugation (10 min at 2000×g at 4 °C).All samples were centrifuged within 30 min.Immediately after centrifugation, plasma was aliquoted into five different vials and subsequently frozen and stored at −80 °C for analysis.Study design is shown in Supplementary Figure 1.
All kits were stored and used as recommended by the manufacturer's instructions.
All samples were measured in duplicates with all five conditions represented on each plate.

Statistical analysis
Data analysis was done using GraphPad Prism 9.4.1 (GraphPad Software, Inc., San Diego, CA).For statistical analysis, we used repeated measurements correcting for multiple testing using a one-way ANOVA.
In all tests, p ≤ 0.05 was considered significant.Data are presented as mean ± 95% CI, unless otherwise stated.

Glucagon
Plasma concentrations of glucagon were measured before and after a 40-min intravenous infusion of amino acids.When comparing plasma concentrations of glucagon after an overnight fast obtained using the sandwich ELISA, there were no significant differences between trasylol, DPP-4 inhibitor or p800 tubes compared to the vehicle (Table 1).In contrast, samples treated with a neprilysin inhibitor had significantly lower plasma levels of glucagon compared to the vehicle.Individual differences in plasma concentrations of glucagon are shown in Figure 1A, B.
During the AA-infusion, plasma concentrations of glucagon measured by the ELISA increased significantly compared to fasting as expected, and the increase did not depend on preanalytical condition.Plasma concentrations of glucagon were approximately 1.7-fold higher when measured with Trasylol as compared to vehicle and this was also the case for p800 tubes.The addition of a DPP-4 inhibitor numerically increased plasma levels of glucagon.No significant differences were observed when plasma contained the neprilysin inhibitor.
When comparing plasma concentrations of glucagon obtained using a RIA kit from Merck Millipore, there were no significant differences between neprilysin inhibitor and trasylol addition compared to the vehicle in the samples collected after ˜12 h of fasting (Table 1).However, using P800 tubes plasma levels of glucagon were 1.2-fold higher compared to the vehicle.The amino acid infusion increased glucagon levels and was 1.2-fold higher using p800 tubes (Figure 1C, D).Similarly elevated levels (1.2-fold) were seen with Trasylol addition.The addition of a neprilysin inhibitor did not result in any significant changes.
Neither neprilysin inhibitor, Trasylol, DPP-4 inhibitor nor the use of p800 tubes had a significant impact on the measurements of glucagon when blood was sampled after an overnight fast and measured using an in-house RIA (Table 1).Numerically, however, the addition of neprilysin inhibitors led to a 2.4-fold decrease compared to vehicles.After the 40-min intravenous infusion of amino acids the plasma concentrations of glucagon increased as expected.Increases were similar with the addition of a neprilysin inhibitor or a DPP-4 inhibitor compared to vehicle.Samples treated with Trasylol and P800 showed numerical increases compared with the vehicle.The same tendency is seen in Figure 1E, F).

Total GLP-1
Plasma levels of total GLP-1 were measured using a sandwich ELISA from Mercodia and did not show any statistical significance when comparing the vehicle with the four different blood container conditions (Table 2 and Figure 2A,B).As expected, the plasma levels of total GLP-1 were similar after ~12 h of fasting and during the 40-minute intravenous amino acid infusion.
When comparing plasma levels of total GLP-1 obtained using the in-house GLP-1 RIA after an overnight fast there was no significant difference when using a neprilysin inhibitor and trasylol compared to the vehicle (Table 2).Containers with DPP-4 inhibitor decreased 1.4-fold compared to the vehicle.Total GLP-1 levels were increased approximately 2.6-fold in P800 tubes compared to the vehicle.Similar patterns with neprilysin inhibitor, trasylol, DPP-4 inhibitor and p800 tubes were observed during the 40-min intravenous infusion of amino acids.Figure 2A-D shows the individual differences in plasma concentrations of total GLP-1 when using the in-house RIA and shows the same tendency as Table 2.

Total GIP
Plasma levels of total GIP in samples treated with Trasylol, DPP-4 inhibitor and p800 tubes were significantly lower approximately 1.5-fold, compared to the vehicle after ~12 h of fasting.Samples treated with neprilysin inhibitor had no significant difference.After the 40-min intravenous amino acid infusion the plasma levels of total GIP were the same, as expected.Similar patterns with Trasylol, DPP-4 inhibitor and p800 tubes were observed (Table 3 and Figure 3A,B).No significant differences were observed when plasma was obtained using neprilysin inhibitor.

Discussion
As glucagon, GLP-1 and GIP circulate in low picomolar range it has been troublesome to accurately measure their plasma concentrations [2,6].We here investigated the importance of the preanalytical sampling conditions.Since the success of the analytical approach may also be influenced by the sampling conditions, we employed both single-antibody assays and sandwich ELISA to look for assay-specific differences.
The main findings of this study are that accurate measurements of glucagon, total GLP-1 and total GIP indeed depend on the choice of blood containers and that this effect will also influence the results of the analytical approach.Theoretically, single-site immunoassays for these peptide hormones would be expected to be relatively independent of enzymatic cleavage, given that this results in measurable fragments of the analyte.In contrast, the sandwich ELISA, with its requirement for simultaneous interaction with two separate antigenic epitopes, would lose reactivity if the two regions were separated by enzymatic cleavage.Most often, the binding regions of the two antibodies in an ELISA are not well characterized or revealed by the manufacturer, and if the enzymatic cleavage does not lead to a separation of the two regions, degradation would not influence the result.
The Mercodia assay for glucagon is thought to involve "terminal wrapping antibodies" in the ELISA because "side-viewing" antibodies (reacting with regions not strictly  note: results are represented as mean ± Sd. all samples were analyzed in duplicate.comparisons were made using one-way anova correcting for multiple comparison using statistical hypothesis testing (dunnett).Sample size for plasma levels of glP-1 analyzed with Mercodia (fasted and after amino acid infusion): N = 17 for vehicle, neprilysin inhibitor, trasylol and p800.N = 6 for dPP-4 inhibitor.Sample size for plasma levels of glP-1 analyzed with 390 (faste): N = 17 for vehicle, neprilysin inhibitor and trasylol.N = 6 for dPP-4 inhibitor and n = 14 for p800.Sample size for plasma levels of glP-1 analyzed with 390 (after amino acid infusion): N = 17 for vehicle, neprilysin inhibitor and trasylol.N = 6 for dPP-4 inhibitor and n = 15 for p800.numbers of samples reflect the numbers of plasma samples that were available for analysis.dPP-4 samples were collected by the end of the experiment, therefore samples from only 6 healthy individuals were available.terminal) would react with several proglucagon-derived products including glicentin, oxyntomodulin, proglucagon 1-61 and others and thereby lose specificity for glucagon.The results of this assay would therefore be expected to be highly sensitive to enzymatic activity in plasma after sampling.In our study, this was not particularly evident for fasting samples, whereas samples obtained after stimulation showed much higher values after the addition of aprotinin (Trasylol), a DPP-4 inhibitor, and after sampling in p800 tubes, whereas neprilysin inhibition was relatively ineffective.In vivo, neprilysin inhibition clearly elevates glucagon levels, perhaps suggesting that the neprilysin activity is membrane-bound rather than present in the circulating fluids.As expected, the inhibitors had little influence on levels when measured with the single-site RIAs, which are directed against the free C-terminus of the molecule because this region is liberated during the biosynthetic processing of proglucagon in the alpha cells (but not in the L-cells of the gut).Thereby these assays become specific for glucagon from the pancreas but may react with both N-terminally extended and truncated forms.The results are only weakly affected by enzymatic degradation, even when it does occur as indicated by the results of the Mercodia assay.
GLP-1 is an extremely sensitive substrate for DPP-4 and circulating levels of intact GLP-1 are therefore very low (often immeasurable), and not very helpful for studies of secretion.Therefore, secretion is best measured by assays directed towards the amidated C-terminus, because this region is mainly exposed after the intestinal processing of proglucagon to GLP-1 7-36amide (and glicentin and GLP-2).Using single antibody immunoassays against this region, one will fully capture the metabolite formed after DPP-4 degradation, GLP-1 9-36amide, which is by far the predominating circulating form, but also intact GLP-1 if present.Small amounts of GLP-1 1-36amide may be formed from the pancreas during proglucagon processing and will be included in such measurements of "total GLP-1" but are generally negligible.The epitope of GLP-1 to which C-terminal antibodies bind may be quite small, and therefore the assay should not be influenced by further enzymatic degradation, given that this epitope remains intact.A sandwich ELISA for total GLP-1 is designed in a completely different way.Here it is important that neither of the antibodies are "terminal wrapping" since this would prevent reactivity with molecular forms included in "total GLP-1, " and the assay therefore rendered useless.Indeed, even the C-terminal antibody in those Table 3. Plasma levels of giP were measured using an immunological approach, a sandwich eliSa from Mercodia (cat.no.10-1258-01; lot.no.34166; uppsala, Sweden), after an overnight fast and after a 40-minute intravenous amino acid infusion (vamin electrolyte free 14 g/l; cat.no.b05aba01; fresenius Kabi, copenhagen, denmark; final concentration 331 mg/min/kg body weight).assays is not terminal wrapping to allow reactivity with also non-amidated (glycine-extended) molecular forms whichalthough minor in humans -are formed in some animals (rats, pigs).This also means that the influence of further enzymatic activity on such assays is unpredictable.In the present study, the Mercodia assay did not seem to be influenced by the addition of the inhibitors.The dramatic increase in the results of the C-terminal RIA after sampling in p800 tubes, on the other hand, almost certainly must be due to cross-reacting ingredients in the coating of the tubes.These findings clearly will require further investigation.As expected, GLP-1 secretion was not significantly affected by the intravenous infusion of amino acids.GIP is a much weaker substrate for DPP-4, and circulating GIP is composed of a mixture of intact GIP 1-42 and the DPP-4 metabolite GIP 3-42.GIP is most often measured with the use RIAs based on C-terminal antibodies, resulting in "total GIP." The sandwich ELISAs are not characterized in detail, but the results are generally in good agreement with the C-terminal RIAs, suggesting that neither is particularly sensitive to further enzymatic degradation.The results observed here did not reveal a consistent effect of the inhibitors.The tendency to lower concentrations after inhibitor addition is difficult to explain but may represent some instability of the assay.As expected, GIP secretion was not influenced by the intravenous infusion of amino acids.
Based on these findings, we recommend the following blood containers for clinical trials aiming to measure endogenous secretion of glucagon, GLP-1 and GIP.
1. Glucagon: Use EDTA containers or p800 tubes independent of assay used for measurement 2. Total GLP-1: For Sandwich ELISA, EDTA containers appear sufficient.For RIA, p800 tubes must be evaluated for interaction with the assay before use.
Otherwise EDTA tubes are adequate.3. Total GIP: Use EDTA containers Limitations: The choice of enzyme inhibitors in this trial was made based on previous trials and literature [7].The sample size was chosen to provide robust results in spite of analytical variabilities of up to 10%.A mixed meal stimulation would provide greater physiological variation in the plasma levels of GLP-1 and GIP.
In conclusion, when comparing studies and designing studies the addition of and choice of enzyme inhibitor must be recognized as a source of variation, and this also has important implications for the assays employed.Addition of enzyme inhibitors may be important, particularly for assays of glucagon with specific sandwich ELISAs, whereas RIAs are less influenced.For total GIP and GLP-1 addition of inhibitors seems superfluous (and p800 tubes should be checked for unspecific interaction with the assay).
are represented as mean ± Sd. all samples were analyzed in duplicate and compared using one-way anova correcting for multiple comparison using statistical hypothesis testing (dunnett).Sample size for plasma levels of giP analyzed with Mercodia (fasted): N = 16 for vehicle, neprilysin inhibitor and trasylol.N = 5 for dPP-4 inhibitor and n = 15 for p800.Sample size for plasma levels of giP analyzed with Mercodia (after amino acid infusion): N = 16 for vehicle, trasylol and p800.N = 15 for neprilysin inhibitor and n = 5 for dPP-4 inhibitor.

Table 1 .
Plasma levels of glucagon were measured using three different immunological approaches, a sandwich eliSa from Mercodia (cat.no.10-1271-01; lot.no.32837 and 33191; uppsala, Sweden), a c-terminal in-house ria (code name 4305) and a ria kit from Merck Millipore (cat.no.gl-32K; lot.no.37695501; billerica, Ma uSa), after an overnight fast and after a 40-minute intravenous amino acid infusion (vamin electrolyte free 14 g/l; cat.no.b05aba01; fresenius Kabi, copenhagen, denmark; infusion rate 331 mg/min/kg body weight).Sd. all samples were analyzed in duplicate using one-way anova correcting for multiple comparison using statistical hypothesis testing (dunnett).Sample size for plasma levels of glucagon analyzed with Mercodia (faste and after amino acid infusion): N = 17 for vehicle, neprilysin inhibitor, trasylol and p800.N = 6 for dPP-4 inhibitor.Sample size for plasma levels of glucagon analyzed with ria Millipore (faste and after amino acid infusion): N = 11 for vehicle, neprilysin inhibitor, trasylol and p800.Sample size for plasma levels of glucagon analyzed with 4305 (at fasting): N = 17 for vehicle.N = 17 for neprilysin inhibitor and trasylol.N = 6 for dPP-4 inhibitor and n = 14 for p800.Sample size for plasma levels of glucagon analyzed 4305 (after aa infusion): N = 17 for vehicle, trasylol and neprilysin inhibitor.N = 6 for dPP-4 inhibitor and n = 15 for p800.