Knockdown of GLP-1 Receptors in Vagal Afferents Affects Normal Food Intake and Glycemia

Glucagon-like peptide 1 (GLP-1) is an incretin and satiating hormone that has provided new tools for the pharmacotherapy of obesity and diabetes (1,2). Yet, despite the clinical effectiveness of GLP-1–based drugs in ameliorating the symptoms of type 2 diabetes, the role of endogenous GLP-1 in the control of energy intake and glucose homeostasis is not fully understood. Vagal afferent neurons (VANs) express GLP-1 receptors (GLP-1Rs) (3,4) and terminate in the lamina propria of the intestinal mucosa as well as in the wall of the hepatic portal vein (HPV) (5). VANs may therefore relay the gut GLP-1–derived signals to the brain and, hence, mediate satiating and glucoregulatory responses. Previous studies using lesioning approaches have implicated the vagus nerve in the effects of peripherally administered GLP-1 on food intake and glycemia (reviewed in 6,7). In more recent studies, sudiaphragmatic vagal deafferentation (SDA) in rats clearly attenuated the acute eating-inhibitory effect of intraperitoneally (IP) infused GLP-1 (8) and exendin-4 (Ex-4), a GLP-1R agonist (9). Moreover, unlike Sham-operated rats, SDA rats failed to show a GLP-1R–mediated incretin response (10). Based on these findings, it is reasonable to hypothesize that endogenous gut-derived GLP-1 could activate GLP-1Rs on VANs in a paracrine-like fashion to reduce food intake, limit gastric emptying, and trigger a neural component of the incretin effect. Disruption of this endogenous GLP-1 signaling mechanism in the VANs due to genetic or environmental factors may contribute to the pathophysiology of obesity and diabetes. Hence, we examined the physiological role of VAN GLP-1Rs in the control of food intake and regulation of glucose homeostasis by generating a specific knockdown (kd) of VAN GLP-1R expression in rats. Our approach is based on the delivery of a short hairpin (sh)RNA construct targeting the GLP-1R mRNA transcript by injecting a lentiviral vector (LV) bilaterally into the nodose ganglia (NG) of rats. Using RNA interference to manipulate gene expression in a tissue-specific manner, we report that VAN GLP-1Rs 1) are required for the physiological control of meal size and gastric emptying but not for the regulation of long-term energy intake and body weight, 2) are necessary for the full effects of acute IP GLP-1 and Ex-4 administration on eating and gastric emptying, and 3) mediate a neural component of GLP-1’s incretin effect that is physiologically relevant for the postprandial regulation of blood glucose.

Collectively, our findings establish the VANs as a major mediator of endogenous GLP-1’s short-term effects on eating, gastric emptying, and glycemia. They also indicate, however, that the reduction of GLP-1R expression in VANs is not sufficient to promote obesity under normal eating conditions.

Male Sprague Dawley rats (Charles River) were individually housed (21 ± 1°C, 55 ± 5% relative humidity) with a 12/12 h dark/light cycle. Unless otherwise noted, animals had ad libitum access to water and standard chow (Kliba 3433, energy density 3.13 kcal/g). All experimental procedures were approved by the Zurich Cantonal Veterinary Office.

pLKO.1-puro vectors expressing turbo green fluorescent protein (GFP) and the U6 promoter-driven shRNA sequence targeting the rat GLP-1R mRNA or a nontarget shRNA sequence were obtained from Sigma-Aldrich. Efficiency of the GLP-1R–targeting shRNA construct was verified in vitro in INS-1E cells (a GLP-1R expressing rat pancreatic β-cell line, generously provided by P. Maechler and C.B. Wollheim, University of Geneva). GLP-1R–targeting (LV-shGLP-1R) or control (LV-shCTL) lentiviral particles were produced in human embryonic kidney 293T cells using the pMD2.G and psPAX2 plasmids (gifts from D. Trono, École Polytechnique Fédérale de Lausanne; cat. no. 12259 and 12260; Addgene) and concentrated to 10¹⁰ particles/mL using 8% PEG6000 (Millipore) and resuspended in PBS.

Rats (290–340 g on surgery day) were anesthetized by an IP injection of ketamine (88 mg/kg; Ketalar, Cantonal Pharmacy Zurich) and xylazine (5 mg/kg; Rompun 2%, Cantonal Pharmacy Zurich), and NG were exposed. A glass capillary (50 μm tip) was used to administer 1.5 μL viral solution into each NG with a Picospritzer III injector (Parker Hannifin). To ensure expression of the viral constructs, we allowed animals to recover for 20 days. IP and HPV catheters were implanted as previously described (complete description in 11). Intracerebroventricular (ICV) cannulas were implanted in the fourth ventricle (stereotaxic coordinates: 2.5 mm posterior to lambda, 0 mm lateral to mid-line, 5 mm below skull surface), and placement was verified functionally with infusion of 5-thioglucose (Sigma-Aldrich) using a 2.5-mm injector (210 μg/rat) and anatomically postmortem.

Animals received an IP injection of pentobarbital (100 mg/kg; Cantonal Pharmacy Zurich), and NG, brain, and pancreas were immediately collected. For gene and protein analysis, tissues were frozen in liquid nitrogen and stored at −80°C. For GFP visualization, NG were fixed for 2 h in 4% paraformaldehyde and 25% sucrose solution in PBS and cut at 10 μm in a cryostat and mounted on glass slides.

The nucleus tractus solitarii (NTS) and the hypothalamic dorsomedial, paraventricular, and arcuate nuclei were mircropunched using anatomical landmarks, and NG from the same animal were pooled before RNA and proteins were extracted using Trizol (Life Technologies). RT quantitative PCR was performed using SybR Green on a OneStep Plus instrument (Applied Biosystems), and results were analyzed using the 2ddCt method. A Western blot was performed to detect the GLP-1R protein (1:400; rabbit antibody 39072, Abcam) using β-actin as reference (1:3,000; mouse antibody AC-74, Sigma-Aldrich).

GLP-1(7-36)amide (Bachem H-6795), Ex-4 (Bachem H-8730), and cholecystokinin octapeptide (CCK) (Bachem H-2080) were resuspended in sterile PBS and administered at doses of 33 μg/kg IP (GLP-1), 0.3 μg/kg IP (Ex-4), and 4 μg/kg IP. (CCK) via catheters. Ex-4 was administered into the fourth ventricle at a dose of 0.3 μg/rat. Rats were habituated to IP or ICV injections with vehicle solutions on three occasions before experiments.

Food was available through a niche and placed on scales (XS4001S; Mettler-Toledo) for continuous measurement (described in 12). Meal patterns were analyzed with custom software (LabX meal analyzer 1.4, Mettler-Toledo). Data are presented as average of 3 days. For food intake experiments after IP GLP-1, rats were fasted overnight and received a 3-g premeal 1 h before dark onset to allow GLP-1R trafficking to the VAN membrane (4). For IP CCK and Ex-4 or ICV Ex-4, rats were fasted for 4 h before dark onset. In all cases, rats received IP or ICV injections right before dark onset and were brought immediately to their home cages.

One week prior to the experiment, rats were habituated to test meals and restricted feeding schedule (test meal at dark onset, ad libitum food access from 3 to 8 h after dark onset, and food deprivation otherwise). On experimental days, rats received a 4-g chow meal containing 1% (wt/wt) paracetamol (4-acetamido-phenol, Sigma-Aldrich) and 0.25% (wt/wt) saccharin (Sigma-Aldrich). IP or ICV injections were given 5 min prior to the test meal. Baseline tail vein blood was taken 30 min prior to test meal onset, and postmeal blood was collected according to the scheduled time points. Paracetamol concentrations were measured with a commercial kit (K8002; Cambridge Life Sciences).

Measurements were conducted in an open-circuit calorimetry Phenomaster system (TSE) after 5 days of habituation. Data are presented as 1 h time bins averaged over 3 days.

Blood was sampled from HPV catheters in unrestrained animals 30 min prior to (baseline) and according to the scheduled time points after the beginning of a 5-g chow test meal. Glucose was measured twice using AccuCheck (Roche), and 150 μL blood was immediately mixed with EDTA (Titriplex, Merck), aprotinin (Sigma-Aldrich), and DPP-IV inhibitor (Millipore) before centrifugation and storage of the plasma at −80°C. Total active GLP-1, insulin, and glucagon were measured simultaneously using an immunoassay (MesoScale Discovery multispot K15171C).

Rats food-deprived for 16 h and adapted to gavage received an oral bolus of 40% glucose solution (2 g/kg). Blood samples for glucose and insulin were taken from tail vein at baseline and 15, 30, 60, 90, and 120 min after the oral glucose bolus. Insulin was measured using an immunoassay (MesoScale Discovery single spot for mouse/rat K152BZC).

Data normality was verified using the Shapiro-Wilk (when n ≥ 7) and the Kolmogorov-Smirnov (when n ≤ 6) tests, and homoscedasticity was checked by visualizing the distribution of residuals. Nonparametric tests were used otherwise. When data distribution was compatible with normality, outliers were detected using the Grubb test. Differences were analyzed by a Student t test for unpaired normally distributed values of equal variance (Figs. 1C, 1E–G, 2D, 3A–I, 3K–L, and 5A and B), or a Mann-Whitney U test for unpaired comparison of non–normally distributed data (Figs. 1D and 3L) using GraphPad Prism (version 6.05). Where the dependent variable was affected by two factors—one within-subject factor (time or injection) and one between-subject factor (surgery group)—the data were analyzed with a mixed ANOVA (Figs. 2A–C, 3J, 4A–F, and 5C–H) using SAS (version 9.3). When the main effect or interaction terms were significant, post hoc analyses using the Bonferroni correction were performed. Data are presented as mean ± SEM. P values <0.05 were considered significant. ns indicates that statistical significance was not reached. All graphs were generated using GraphPad Prism (version 6.05).

Three weeks after bilateral NG injection of a lentivirus containing a GLP-1R–targeting shRNA construct (Fig. 1A), infection of VANs was confirmed by visualizing GFP expression in NG sections (Fig. 1B). GLP-1R mRNA expression in NG was reduced by 52.5% in GLP-1R kd rats compared with control rats injected with control lentiviral particles containing nonspecific target shRNA (Fig. 1C). In addition, reduction of GLP-1R protein was confirmed using Western blot from NG protein extracts of control and GLP-1R kd rats (Fig. 1D). GLP-1R expression was unchanged in the pancreas, where GLP-1R activation improves insulin secretion, as well as in the key GLP-1R–expressing regions in the brain involved in the control of food intake and glucose homeostasis (Fig. 1E), indicating tissue specificity of the LV-mediated gene kd approach. Moreover, VAN genes involved in the control of food intake, such as the CCK A receptor (CCKaR), leptin receptor (LepR), and peroxisome proliferator–activated receptor γ (PPARg), were similarly expressed in the NG of control and GLP-1R kd rats (Fig. 1F), suggesting that the GLP-1R shRNA construct used was specific. Finally, we measured food intake 30 min after an IP injection of CCK and found a 25–40% reduction (9,13) in both groups, demonstrating the preservation of VAN functional integrity (Fig. 1G). Together, these results indicate that the LV-mediated delivery of an shRNA construct by bilateral NG injection is tissue specific, target-specific, and does not impair vagal afferent function.

Body weights of control and GLP-1R kd rats remained similar during the entire course of the experiments when fed ad libitum (Fig. 2A). Moreover, energy balance remained undisturbed as documented by similar daily chow intake (Fig. 2B) and daily energy expenditure (Fig. 2C and D).

Measurements of undisturbed meal patterns showed that GLP-1R kd induced increases in meal size (Fig. 3A) and meal duration (Fig. 3D). Although these changes were significant over 24 h, the increase in meal size and duration was only evident during the dark phase with no significant differences in the light phase (Fig. 3B, C, E, and F). Consistent with the long-term daily food intake measurements (Fig. 2A), 24-h food intake was not affected by the kd during the period of meal pattern measurements (data not shown): this was mainly due to a compensatory decrease in the number of meals in the GLP-1R kd rats (Fig. 3G), which was also evident only during the dark phase (Fig. 3H and I). Moreover, the rate of gastric emptying after a meal, as measured by the appearance of paracetamol in the plasma after a test meal, was enhanced in GLP-1R kd rats compared with controls (Fig. 3J). Finally, in line with their ad libitum meal pattern, GLP-1R kd rats showed an increase in food intake during the 1-h refeeding period after a 16-h fast (Fig. 3K), which was associated with an amplified peak of energy expenditure during this period (Fig. 3L).

According to previous studies (for review, see 6,7), GLP-1 and low-dose Ex-4 require intact VAN to exert their full inhibitory effects on short-term food intake. Hence, we further tested whether the satiating effects of exogenous GLP-1 and Ex-4 were attenuated in GLP-1R kd rats. Indeed, IP injections of GLP-1—at a dose that elevates intestinal lymph GLP-1 similar as a meal (14)—or low-dose Ex-4 failed to significantly reduce 1-h food intake in GLP-1R kd rats (Fig. 4A and B). In a similar design, we tested whether GLP-1Rs in the VAN mediate the GLP-1– or Ex-4–induced inhibition of gastric emptying using the paracetamol test. IP GLP-1 and Ex-4 failed to inhibit gastric emptying in the GLP-1R kd group as shown by early appearance of paracetamol in the plasma 20 min after a test meal (Fig. 4D and E). To test whether GLP-1R expressed on peripheral or central terminals of vagal afferents mediate these effects, we performed a fourth ICV injection of Ex-4. In contrast to the attenuated effects of IP Ex-4, the fourth ICV injection of Ex-4 showed the full expression of the inhibitory effects on eating and gastric emptying in kd rats (Fig. 4C and F). These observations are consistent with the idea that peripheral but not central GLP-1Rs expressed on VAN terminals mediate the effects of IP-infused GLP-1 on food intake and gastric emptying. Together, these results indicate that activation of GLP-1Rs on the VANs in the gut mediates the inhibitions of eating and gastric emptying induced by IP GLP-1 or low-dose Ex-4.

GLP-1Rs on vagal afferents have been implicated in the neuroincretin effects of endogenous GLP-1. We measured HPV blood glucose in the fasted and fed conditions in control and GLP-1R kd rats. After an overnight fast (16 h), HPV blood glucose was not different between GLP-1R kd and control rats (Fig. 5A), whereas kd rats showed a higher blood glucose than controls in the fed state (2 h of fasting after ad libitum food access 5 h into the dark phase) (Fig. 5B). This indicated that GLP-1Rs in the VAN are necessary for the full incretin effect of meal-induced GLP-1. To test this hypothesis, we measured HPV blood glucose (Fig. 5C), plasma insulin (Fig. 5D), GLP-1 (Fig. 5E), and glucagon (Fig. 5F) after a 16-h fast followed by a 5-g chow test meal. HPV blood sampling was chosen to allow for the concomitant measurement of the meal-induced increase in HPV GLP-1 levels (which is subject to a rapid degradation by dipeptidyl peptidase-4 in the systemic circulation and in the liver). The test meal elevated HPV blood glucose in both groups but resulted in a higher postmeal blood glucose level in the GLP-1R kd rats. Interestingly, postmeal HPV GLP-1 and glucagon were similar in both groups, but insulin appearance in the HPV was blunted in the GLP-1R kd group. An oral glucose tolerance test, however, did not reveal differences in plasma glucose and insulin levels between control and kd rats (Fig. 5G and H)

It has long been hypothesized that VANs could control eating behavior by serving as a key mediator of nutritional cues from the intestine to the brain (6,7). Vagal lesioning methods (including SDA, the most specific method for the disconnection of subdiaphragmatic vagal afferents [15]) provided initial evidence for the role of the VANs in mediating peripheral exogenous GLP-1 effects on food intake and glycemia (9,16). These methods, however, resulted in a complete impairment of VAN signaling and function, and they did not specifically test the role of VANs in the effects of endogenous GLP-1. Therefore, the role of VAN GLP-1R signaling in mediating the effects of endogenous GLP-1 on energy homeostasis has been difficult to elucidate. Lately, vagal-specific genetic deletions of receptors involved in nutrient sensing have been attempted in mouse models using Phox2b or Nav1.8 genes, whose promoters drive cre-recombinase expression (17–19). The cre expression in these mouse models is, however, not limited to the VANs and extends to the spinal afferents and brainstem (20,21). Moreover, for a tightly controlled system such as eating behavior, gene deletion approaches are suspected to yield compensatory mechanisms during development to maintain overall energy balance (18,22,23). To overcome these obstacles, we used the bilateral delivery of an shRNA-expressing viral vector into the NG to accomplish an inducible molecular manipulation of VAN function in adult rats. To our knowledge, NG injection has so far been limited to nonsurvival administration of compounds for electrophysiological recordings (24) and, more recently, to the unilateral delivery of viral tracers and optogenetics-related tools (25). Here, we demonstrate that bilateral NG injection of a viral-mediated shRNA yields a specific and long-lasting reduction of GLP-1R expression in the VANs.

When fed ad libitum, control and GLP-1R kd rats had similar body weights over the entire course of the experiment. Also, daily food intake was unchanged, which, together with the unchanged energy expenditure, demonstrates that a reduction of GLP-1R expression in VANs does not chronically disturb energy balance. Overnight fasting and refeeding, however, led to a much larger food intake and peak of energy expenditure in the GLP-1R kd rats, presumably due to an increased meal-induced thermogenesis. A major caveat of the viral-mediated GLP-1R kd approach is that the reduction of GLP-1R expression in the VAN is partial. Therefore, it cannot be excluded that the remaining expression of GLP-1R in the VAN accounted for the absence of chronic changes in energy homeostasis. Nevertheless, this negative phenotype in the body weight and daily food intake in our kd model is in accordance with the GLP-1R kd^ΔPhox2b mouse model (17).

Previous data using surgical lesions concluded that vagal afferents are needed for the full expression of IP-injected GLP-1 or acute Ex-4 effects on food intake (8,9,16,26). Given that GLP-1 is cleared from plasma within minutes (27,28), these results suggested that endogenous GLP-1 released from intestinal L cells activates GLP-1R located on intestinal VANs in a paracrine-like fashion. Here, we present strong evidence that GLP-1Rs in the VANs in fact mediate the satiating and gastric emptying–inhibiting effects of endogenous GLP-1, as well as IP-infused GLP-1 and Ex-4.

Our data specifically confirm the importance of VAN GLP-1R signaling in the short-term control of eating by endogenous GLP-1. GLP-1R kd specifically delayed nocturnal meal termination (satiation) without affecting postmeal satiety, consistent with a paracrine effect of endogenous GLP-1 on VAN GLP-1Rs. It stands to reason that the effects are mostly nocturnal, when rats consume most of their calories, because intestinal GLP-1 is being secreted via luminal nutrient stimulation. Moreover, recent data demonstrated a circadian rhythm for GLP-1 (29), with a maximal GLP-1 release upon glucose stimulation before dark onset. The effects of the GLP-1R kd may therefore be magnified during the early dark phase when circulating GLP-1 levels are high. Finally, GLP-1Rs in the VAN are internalized during fasting and translocated to the membrane in the fed state (4). The absence of an effect on meal size and duration during the light phase may therefore be due to the internalization of GLP-1Rs in the VAN when food intake is low and intermeal intervals are prolonged.

In addition, we demonstrated that VAN GLP-1R expression is necessary for the normal gastric emptying of a meal, as hypothesized from earlier studies (27,30,31), consistent with a paracrine effect of endogenous GLP-1 on VAN GLP-1Rs. Whether the GLP-1–induced reduction in gastric emptying is a major mechanism by which endogenous GLP-1 controls meal size is unclear. Further experiments should test whether the eating-inhibitory effect of GLP-1 on meal size is still present in sham-fed animals in which normal gastric emptying is prevented by a gastric fistula.

Surprisingly, despite the consistent and lasting increase in meal size, daily food intake was not altered by VAN GLP-1R kd, which was the result of a compensatory decrease in meal number. Given the fact that IP administration of GLP-1 induces short-term satiation by a reduction in meal size (8,32), it appears that the increase in meal size is the primary effect of VAN GLP-1R kd, reflecting a specific satiating effect of endogenous GLP-1 via a paracrine action. The decrease in meal number is likely a secondary, compensatory mechanism. It is also possible to speculate that bigger meals in VAN GLP-1R kd rats trigger the release of other eating-inhibitory hormones (e.g., CCK, peptide tyrosine tyrosine) whose signaling remains intact and delay the rise in ghrelin. Moreover, VAN GLP-1R kd may lead to compensatory changes in neuronal activity of the dorsal vagal complex (N. Gafner, J.-P. Krieger, S.J. Lee, unpublished observations), that, in turn, may prolong the intermeal interval and hence decrease meal number. Together, this strengthens the classical view that VAN GLP-1Rs mainly mediate short-term satiation and may be one of several redundant eating control mechanisms comprising the gut-brain axis (7,9,16,33).

Activation of pancreatic GLP-1Rs by gut-derived endogenous GLP-1 is considered to be the classical mechanism of GLP-1’s incretin effect. Several findings in rodents and humans, however, suggest an additional involvement of a neural pathway in the GLP-1R–dependent release of insulin (34–38). Our findings demonstrate that vagal GLP-1R signaling is necessary for the normal control of glycemia after eating. Interestingly, the elevated postmeal glycemia in GLP-1R kd rats was concomitant with lower levels of HPV insulin 15 min after the beginning of the meal without changes in GLP-1 or glucagon levels. These results indicate that endogenous meal-induced GLP-1 acts on VAN GLP-1Rs to control postmeal glycemia via a neural component of the incretin effect. Moreover, recent data showed that IP infusion of atropin, a blocker of muscarinic receptors, reduces the insulin response after intravenous coinfusion of glucose and GLP-1 (39). Together with our findings, this supports the idea that ascending VANs and descending pancreatic efferents form a “gut-brain-pancreas” axis mediating some of the effects of intestinal GLP-1 on insulin secretion. Intriguingly, after an oral glucose tolerance test (OGTT), no significant differences in blood glucose or plasma insulin were seen between the two groups. Several differences between the use of a test meal or a glucose bolus could explain this discrepancy. First, an OGTT is thought to induce the release of GLP-1 with a different amplitude and/or time course than a solid meal ingested over several minutes (40). As vagal lesion studies indicated that high levels of circulating GLP-1 can exert a VAN-independent effect (8,41), it is plausible that higher GLP-1 levels after the OGTT mask the role of VAN GLP-1Rs. Second, glucose measurements from HPV blood sampling may not capture the hepatic effect on glycemia. The contribution of the liver to the glucoregulatory effect of GLP-1 via both insulin-dependent and insulin-independent mechanisms is indeed receiving increasing support (42–44), and it is consequently possible that HPV blood glucose and insulin levels differ from systemic blood values. Finally, the stress caused by the OGTT procedure (gavage and tail vein sampling) may have masked the difference in glucose levels between GLP-1R kd and controls. Corticosterone has a powerful effect on glucose levels because it inhibits insulin secretion and increases hepatic gluconeogenesis. Therefore, a stress-free HPV sampling (voluntary meal followed by unrestrained blood sampling through HPV catheter) may be a more accurate way to differentiate VAN GLP-1R kd effects on glucose/hormonal changes than tail vein sampling.

Together, our findings demonstrate a crucial role for a vagal pathway in the maintenance of normal eating behavior and postprandial glycemia by endogenous GLP-1. Recent studies, however, have shown that GLP-1R agonists, such as liraglutide, exert their body weight and glucose-lowering effects independent of vagal afferents (45) and do not require VAN GLP-1Rs (17). Most likely, GLP-1R agonists do not access VAN GLP-1Rs when administered subcutaneously. Instead, GLP-1R agonist effects on body weight may be mediated by the activation of central GLP-1Rs (45). Based on our data, it is, however, possible to consider the vagus nerve as a target organ to modulate satiation and glycemia.

Acknowledgments. The authors thank M. Hayes (University of Pennsylvania), R. Burcelin (INSERM Toulouse), and T. Lutz (University of Zurich) for scientific advice during study preparation. R. Clara, S. Fedele, N. Jejelava, S. Kaufman, R. Kästli, M. Klarer, and M. Labouesse (ETH Zurich) are acknowledged for their precious help during animal experiments and data analysis. The authors thank C. Boyle (University of Zurich) for support and advice related to indirect calorimetry measurements.

Funding. This study received funding from ETH Zurich (4712-2).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. J.-P.K. participated in the study conception and design, conducted the experiments, analyzed data, and wrote the manuscript. M.A. set up and performed the nodose ganglion injections. K.G.P. and P.L. participated in the experiments and data interpretation. W.L. participated in the study conception and design, participated in data interpretation, and edited and reviewed the manuscript. S.J.L. participated in the study conception and design, participated in the experiments/data interpretation, and edited and reviewed the manuscript. S.J.L. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.