Acute Activation of GFRAL in the Area Postrema Contributes to Glucose Regulation Independent of Weight Free

Growth differentiation factor 15 (GDF15) is a hormone from the transforming growth factor-β superfamily (1). In rodents fed a high-fat diet (HFD), a pharmacological increase of plasma GDF15 levels decreases feeding and weight (2–5) while increasing energy expenditure (6,7), glucose tolerance (2,4), and insulin sensitivity (4,8). These effects (2–8) are mediated by the GDF15 receptor, GDNF family receptor α-like (GFRAL), which is expressed in the area postrema (AP) and nucleus of the solitary tract (NTS) of the hindbrain. However, there is an ongoing debate whether the GDF15–GFRAL signaling axis regulates glucose homeostasis without involving feeding and weight changes.

On one hand, GDF15 reduces plasma insulin levels and improves glucose tolerance in HFD rodents to a similar extent as seen in the pair-fed rodents (4). But on the other hand, despite having similar weight, GDF15 induces higher glucose tolerance and insulin sensitivity in HFD rodents compared with chow-fed rodents treated with vehicle (9). In fact, GDF15 enhances insulin sensitivity during hyperinsulinemia-euglycemic clamps by inhibiting glucose production and increasing adipose glucose uptake independent of changes in food intake and weight (8), and GDF15 treatment reduces the onset of diabetes by 50% in type 1 diabetic nonobese mice (10) and enhances glucose-stimulated insulin secretion in pancreatic β-cells (11). Taken together, these findings indicate GDF15 regulates glucose homeostasis independent and dependent of changes in feeding and weight.

In this study, we evaluated the potential feeding- and weight-independent effects of GDF15 on glucose regulation, the AP as a site of GDF15 action, and the necessary role of the GDF15–GFRAL signaling axis in the AP for metformin action in a 3-day high-fat diet (HFD) rat model. This 3-day HFD model is used to evaluate the pharmacological relevance of brain GDF15 action; as 3-day high-fat (HF) lard oil–enriched diet feeding in male Sprague-Dawley rats results in development of hyperphagia, hyperinsulinemia, and insulin resistance independent of weight gain (12,13). Our current working hypothesis is based on the fact that glucose-sensing mechanisms in the AP regulate glucose production independent of changes in food intake and weight (14), and GDF15 infusion activates GFRAL in the AP, but not the NTS, to substantially lower feeding and weight in rodents (15,16). In addition, metformin lowers feeding and weight via increasing plasma GDF15 levels as well as subsequent GFRAL activation in the AP (16–18). Thus, we postulate that activation of GFRAL in the AP contributes to glucose regulation of GDF15 and metformin in vivo.

Male Sprague-Dawley rats (Charles River Laboratories), weighing 250–270 g and 300–320 g, were used in the study. Rats were housed in individual cages at 23°C, subjected to a standard 12-h light–dark cycle, and had ad libitum access to drinking water and a regular chow diet (7012; Envigo) containing 17% fat, 25% protein, and 58% carbohydrate content (3.1 kcal/g total metabolizable energy). All animal protocols were reviewed and approved by the Institutional Animal Care and Use Committee at the University Health Network in accordance with the Canadian Council on Animal Care guidelines.

After 2 days of acclimation to the animal facility, the rats were anesthetized (ketamine, 60 mg/kg; xylazine, 8 mg/kg) and surgeries were performed. The rats receiving both AP and upper small intestinal (USI) and vascular (hereafter, USI/vascular) surgeries weighed 250–270 g, whereas the rats weighing 300–320 g received USI/vascular surgery only. For the AP surgery, a single 26-gauge, stainless steel guide cannula (C315G; Plastics One Inc.) was stereotaxically implanted via coordinates targeting the AP (0.3 mm posterior the occipital crest, 0 mm on the midline, 7.9 mm below the cranial surface), as described previously (14,16). For the AP-specific lentiviral infections, after AP cannulations while anesthetized, the rats immediately received lentivirus expressing shRNA targeting to GFRAL (shGFRAL; 1.0 × 10⁶ infectious units; sc-270688-V, Santa Cruz Biotechnology Inc.) or mismatch shRNA (shMM; 1.0 × 10⁶ infectious units; sc-108080, Santa Cruz Biotechnology Inc.) AP injection at 2 μL over 15 min, which has been shown to achieve the AP-specific gene knockdown (14,16). USI/vascular surgery was performed 7 days after the AP cannulation. A gut catheter was placed 6 cm distal to the pyloric sphincter for infusion purposes, and vascular cannulas were surgically implanted into the left carotid artery and right jugular vein for blood sampling and infusion, as previously described (13). The rats were assigned to an HFD with 10% lard oil (57IR; TestDiet) containing 34% fat, 22% protein, and 44% carbohydrate (3.9 kcal/g total metabolizable energy) after the surgeries for 3 days until the experimental day, and postsurgical food intake and body weight were monitored daily. A small subset of rats was kept on a regular chow diet after surgeries to serve as a control for HFD feeding. HFD rats that did not exhibit hyperphagia in 3-day cumulative food intake as compared with chow-fed rats and/or did not attain at least 90% of their pre–USI/vascular surgical body weight the day before the experiment were excluded.

GDF15 (9279-GD; R&D Systems) was dissolved in 0.9% saline to 4 μmol/L, and GDF15 or 0.9% saline was infused into the AP at the rate of 0.006 μL/min from 60 min before until the end of an intravenous glucose tolerance test (IVGTT; 2 pmol of GDF15 over 95 min). The total amount of GDF15 delivered was the same as reported previously to reduce feeding and weight in the AP via activating GFRAL (16), and the 60-min preinfusion time was based on the finding that subcutaneous GDF15 injection twice, with the last injection given 1 h before an insulin tolerance test, effectively increased systemic insulin sensitivity in HFD-fed rats (8). Metformin (200 mg/kg; PHR-1084, Sigma-Aldrich) or 0.9% saline was infused via the USI catheter at the rate of 0.4 mL/min for 5 min (2 mL in total) at 2 h before the starting of the IVGTT. The dose of metformin and the infusion protocol were decided on the basis of the previous metformin study that showed metformin delivered at this dose and via the current protocol increased plasma GDF15 levels while lowering feeding and weight (16).

The IVGTT was performed 4 days after USI/vascular surgery, and the rats were fasted overnight (from 4:00 p.m. until 10:00 a.m.) prior to IVGTT. The pre-experimental body weight was measured first, and basal blood samples were obtained from conscious and unrestrained rats immediately before the start of AP GDF15 or USI metformin administration. After obtaining additional blood samples at the 0-min time point of IVGTT, a bolus of d-(+)-glucose (45%; G8769, Sigma-Aldrich) at a dose of 0.25 g/kg was administered intravenously, followed by a flush with 0.9% saline. Subsequent blood samples were collected at 2-, 5-, 10-, 15-, 20-, 25-, 30-, and 35-min intervals after the glucose injection, in accordance with the described protocol (19). Blood samples were collected in heparinized tubes and centrifuged.

The clamp procedure was conducted 5 days after the vascular surgery. Rats were fasted 4–6 h, and their pre-experimental body weight was measured right before the clamp. The clamp procedure was performed in conscious, unrestrained, and nonstressed rats, as described elsewhere (13). A primed intravenous infusion of [3-³H]glucose (40 μCi bolus, 0.4 μCi/min; Perkin Elmer) was started and continued until the end of the clamp (0–210 min) to evaluate steady-state glucose kinetics, using tracer-dilution methodology. At 90 min after [3-³H]glucose infusion, the pancreatic-euglycemic clamp was initiated. Somatostatin (3 μg/kg/min) was infused to suppress endogenous insulin secretion and moderately reduce glucagon secretion, and insulin (2 mU/kg/min) was co-infused to maintain hyperinsulinemia. An adjustable 25% glucose infusion was periodically adjusted to maintain euglycemia until the end of the clamp. Blood samples were collected (in heparinized tubes and centrifuged) every 10 min throughout the studies for plasma glucose measurement as well as during the basal (60–90 min) and clamp (180–210 min) periods for [3-³H]glucose radioactivity and insulin level determination. Glucose turnover was calculated using [3-³H] and steady-state formulas, in which the rate of glucose appearance is equivalent to disappearance. The steady state for the clamp period was determined between 180–210 min. At 60 min after [3-³H]glucose infusion, GDF15 (4 μmol/L) or 0.9% saline was infused into the AP at the rate of 0.006 μL/min until the end of the clamp (3 pmol of GDF15 over 150 min).

At the end of the IVGTT or clamps, the rats were anesthetized, and the rats with AP cannulation were injected with 3 μL of bromophenol blue over 30 s through the AP cannula to verify the correct placement of the cannula. The rats that showed injection of dye beyond the correct AP location were excluded. The AP and NTS tissue samples were collected separately, as previously described (14). In brief, the rats were immediately decapitated, and the cerebellum was lifted to expose the caudal part of the brain. Then, the whole brain was isolated from the skull, bathed in PBS, and placed on a metal plate cooled by dry ice. The dorsal vagal complex was obtained by cutting from the point of the obex to the end of the vagal triangle, and a micro knife was used to isolate the AP and NTS tissues, guided by a rat brain atlas as well as with visual assistance from markers such as the cuneate nucleus and central canal. Liver tissue was freeze-clamped in situ with steel tongs that had been precooled in liquid nitrogen and was cut off from the liver. All tissues were immediately frozen in liquid nitrogen and stored at −80°C for future analysis.

AP (1 mg), NTS (5–8 mg), and liver (10 mg) tissue samples were weighed, and lysis buffer (20 mL/mg) was added. The tissue samples were homogenized for 3 min in lysis buffer using a tissue homogenizer (Bullet Blender Storm Pro; Next Advance Inc.). RNA was isolated using the PureLink RNA Mini Kit (catalog 12183025; Thermo Fisher Scientific) and was subjected to DNase I digestion. Then, RNA was quantified by measuring the absorbance at 260 and 280 nm (Cytation 5; BioTek Instruments), and cDNA was generated using 2 μg of RNA and the SuperScript Vilo cDNA Synthesis Kit (catalog 11754050; Thermo Fisher Scientific). Quantitative PCR was performed using TaqMan Gene Expression master mix (catalog 4369016; Thermo Fisher Scientific) and TaqMan primers (Thermo Fisher Scientific) for rat ribosomal protein 18s (assay identifier Rn01428913_gH), rat Gfral (assay identifier Rn04244814_m1), rat glucose-6-phosphatase (G6pc; assay identifier Rn00689876_m1), and rat phosphoenolpyruvate carboxykinase (Pepck; assay identifier Rn01529014_m1) using a quantitative PCR machine (QuantStudio 7 Flex; Applied Biosystems). Relative gene expression was normalized to rat ribosomal protein 18s as the reference gene using the comparative cycle threshold method.

Plasma glucose levels were determined by the glucose oxidase method using a glucose analyzer (GM9; Analox Instruments). Plasma insulin levels were measured by an ELISA kit (catalog 90060; Crystal Chem). Plasma GDF15 levels were also measured by an ELISA kit (catalog MGD150; R&D Systems).

Statistical analyses were performed using Prism 8.0 (GraphPad Software). Unpaired Student t test (two-sided with 95% confidence level) was used to compare two groups. One-way ANOVA with Tukey post hoc test was performed for more than two groups with single variable, and two-way ANOVA with Tukey post hoc test was performed when there were more than two groups with two variables. Differences were considered significant at P < 0.05, and all results were presented as mean + SEM.

The data sets generated during this study are available from the corresponding author upon reasonable request. All the resources used during this study are commercially available.

To investigate the acute AP-specific action of GDF15 on glucose tolerance that is independent of feeding and weight changes, we implanted a single brain cannula targeting the AP of male Sprague-Dawley rats (14,16). After 7 days of recovery, the rats underwent carotid artery and jugular vein cannulation for blood sampling and glucose infusion during an IVGTT (19). Thus, it is important to note that all IVGTTs were performed in conscious, unconstrained, and nonstressed rats. Right after the vascular cannulations, the rats were placed on an HFD for 3 days and, after an overnight fast, the rats received a continuous infusion of either GDF15 (4 μmol/L) or 0.9% saline via the AP cannula at a rate of 0.006 μL/min (total 2 pmol in 95 min), starting 60 min before and lasting until the end of the IVGTT (Fig. 1A). The same dose of GDF15 when delivered into the AP was found to lower feeding and weight (16), and the same AP GDF15 infusion rate was used for AP glucose infusion (with an even larger total delivered volume than the current GDF15 studies) that selectively activated the glucose-sensing mechanism in the AP but not the NTS (14). Because all rats had no access to food during the acute 95-min infusion experiments and had comparable body weight prior to the experiment (Table 1), all effects incurred by GDF15 would be independent of changes in food intake and weight.

Rats subjected to the HFD had higher amounts of pre-experimental 3-day cumulative food intake (131.6 + 6.1 vs. 89.4 + 4.0 kcal; n = 12 and 6, respectively; P < 0.01) but similar pre-experimental weight (288.3 + 4.0 vs. 283.3 + 2.9 g; n = 12 and 6, respectively) as compared with rats fed regular chow and that underwent the same surgical procedures, confirming that this short-term HF feeding model induces hyperphagia without causing obesity, as seen in previous studies (13). Importantly, with comparable pre-experimental 3-day cumulative HF food intake and weight (Table 1), GDF15 versus saline infusion into the AP of HFD rats increased glucose tolerance from 10 to 30 min after an intravenous glucose bolus injection (Fig. 1B), in parallel with a decrease in the area under the curve of plasma glucose levels (Fig. 1B). Plasma insulin levels during the IVGTT were comparable in AP GDF15 rats compared with saline-infused rats (Fig. 1C); the location of the AP cannula placement was verified after the studies via bromophenol blue infusion (Fig. 1D). These results indicate that selective GDF15 acute infusion into the AP enhances intravenous glucose tolerance independent of food intake, weight, and insulin secretion.

Given that AP GDF15 may influence glucose metabolism through changes in insulin sensitivity, we assessed the role of AP GDF15 in insulin sensitivity by repeating the GDF15 AP infusion studies in HFD rats fasted for 4–6 h that underwent the pancreatic hyperinsulinemic-euglycemic clamps with tracer glucose methodology (Fig. 1E). The pre-experimental cumulative HF food intake (230.1 + 31.2 vs. 227.5 + 20.8 kcal; n = 6/group) and weight (316.8 + 10.9 vs. 321.5 + 11.5 g; n = 6/group) was comparable between the rats infused with AP GDF15 and those infused with saline in the clamp experiments, but food intake and weight were higher compared with the rats undergoing the same surgeries for IVGTT (Table 1). This difference can be attributed to the additional recovery day and shortened pre-experimental fasting time in the clamp versus IVGTT protocol.

With comparable pre-experimental HF food intake, weight, and plasma insulin levels (Fig. 1F) during the clamps, the HFD rats receiving GDF15 versus saline infusion into the AP required an increase in exogenous glucose infusion rate (Fig. 1G) to maintain comparable euglycemia (Fig. 1H), indicating GDF15 infusion into the AP increases insulin sensitivity. This higher insulin sensitivity incurred by AP GDF15 versus saline infusion was due to the suppression of hepatic glucose production (Fig. 1I) rather than an increase in whole-body glucose uptake (Fig. 1J), although tissue-specific glucose uptake remains to be assessed. The expression of the gluconeogenic gene G6pc was decreased in the liver of rats after the clamps with AP GDF15 versus saline infusion (Fig. 1K) as well, whereas the expression of Pepck remained comparable (Fig. 1L). Taken together, GDF15 signaling in the AP of HFD rats increases glucose tolerance and enhances insulin sensitivity to lower hepatic glucose production independent of changes in food intake and weight.

To assess the necessary role of GFRAL in the AP, we conducted a knockdown experiment by injecting lentivirus expressing shGFRAL or shMM into the AP via the AP cannula (2 μL in 15 min), immediately after the brain cannulation. IVGTT was performed 11 days after viral injection with either AP GDF15 or saline infusion into HFD rats (Fig. 1A). With comparable pre-experimental, cumulative, 3-day HFD intake and weight (Table 1), GDF15, compared with saline infusion, into the AP of AP shGFRAL-injected HFD rats did not increase glucose tolerance (Fig. 1M), whereas infusion of GDF15 into the AP of shMM- compared with shGFRAL-injected rats increased intravenous glucose tolerance and reduced the area under the curve of plasma glucose levels (Fig. 1M) without affecting plasma insulin levels during the IVGTT (Fig. 1N).

By the end of the IVGTT, the AP and NTS tissue samples were isolated separately from the AP shGFRAL and shMM GDF15-infused rats to measure Gfral mRNA expression. Gfral expression was higher in the AP versus NTS in the AP shMM-injected rats (Fig. 1O), as previously reported (16). Importantly, AP shGFRAL lentiviral injection knocked down Gfral expression by ∼45% in the AP while not affecting the Gfral expression in the NTS, as compared with AP shMM-injected HFD rats (Fig. 1O). These findings indicate that activation of GFRAL in the AP by GDF15 is sufficient to increase glucose tolerance independently of changes in feeding, weight, and insulin secretion.

To examine whether the GDF15-GFRAL axis in the AP is necessary for metformin to increase glucose tolerance, we first explored whether metformin enhances intravenous glucose tolerance in association with increased plasma GDF15 levels but independently of changes in feeding and weight. In overnight-fasted 3-day HFD rats with comparable pre-experimental, 3-day cumulative food intake (145.0 + 13.2 vs. 148.9 + 14.4 kcal; n = 6/group) and weight (302.2 + 4.3 vs. 303.2 + 5.6 g; n = 6/group), we infused a single dose of metformin (200 mg/kg, 0.4 mL/min for 5 min) or 0.9% saline into the lumen of the USI via a gut catheter implanted toward the lower duodenal lumen (Fig. 2A). When metformin is administered using this USI protocol, plasma GDF15 levels are elevated at 6 h after metformin administration in HFD rats under refeeding conditions (16). Herein, we first found that metformin versus saline USI administration elevated plasma GDF15 levels as early as 2 h, and the GDF15 elevation was maintained at least 6 h after USI metformin infusion (Fig. 2B). On the basis of this finding, we performed IVGTT at 2 h after USI metformin versus saline infusion in overnight-fasted HFD rats with comparable pre-experimental, cumulative, HF food intake and body weight (Table 1). USI metformin versus saline infusion increased glucose tolerance (Fig. 2C) in parallel with a decreased area under the curve of plasma glucose levels (Fig. 2C) but did not significantly affect plasma insulin levels during the IVGTT (Fig. 2D). Importantly, the metformin- versus saline-induced increase in plasma GDF15 levels were maintained throughout the IVGTT (Fig. 2E). Thus, metformin increases glucose tolerance and plasma GDF15 levels independent of changes in food intake, weight, and insulin secretion.

Finally, to investigate whether metformin increases plasma GDF15 to trigger the GFRAL expressed in the AP to increase glucose tolerance, we injected shGFRAL or shMM lentivirus into the AP of rats to knock down GFRAL expression within the AP that is sufficient to negate the glucoregulatory effect of direct GDF15 infusion, as described in Fig. 1M. On the 11th day after viral injection, we performed the IVGTT at 2 h after USI metformin or saline infusion (Fig. 3A). With comparable pre-experimental, 3-day, cumulative HF food intake and body weight (Table 1), we found that metformin versus saline USI administration did not increase glucose tolerance in AP shGFRAL-injected HFD rats (Fig. 3B), whereas in metformin-treated, AP shMM-injected HFD rats, glucose tolerance was increased as compared with metformin-treated, AP shGFRAL-injected HFD rats, in parallel with a respective change in the area under the curve of plasma glucose levels (Fig. 3B) without affecting plasma insulin levels during the IVGTT (Fig. 3C). By the end of the IVGTT, AP and NTS tissue samples were collected separately from the AP shMM and shGFRAL rats receiving USI metformin administration, and we confirmed that the AP shGFRAL- versus shMM-injected rats had a reduction of Gfral expression in the AP but not the NTS (Fig. 3D). These findings collectively indicate that the GDF15–GFRAL signaling axis in the AP is necessary for metformin to increase glucose tolerance in HFD rats.

In this study, we found that the GDF15–GFRAL axis contributes to metformin action in increasing glucose tolerance. However, we were unable to determine the origin of the plasma GDF15 after USI metformin infusion. Notably, metformin delivered via the current USI infusion method reaches the USI and ileum, and is absorbed into the liver but likely not the kidney yet, as indicated by the increased GDF15 expression in the USI, ileum, and liver, but not the kidney, at 6 h after USI metformin infusion (16). Considering that metformin-induced upregulation of GDF15 in the USI or ileum did not contribute to the increase in plasma GDF15 levels (16), we postulate that the liver might be the source of plasma GDF15 responsible for mediating the effect of acute USI metformin infusion on glucose tolerance that clearly warrant future investigation.

Long-term 11-day daily oral metformin administration in HFD mice increased glucose tolerance and insulin sensitivity and plasma GDF15 levels (18). However, the increase in plasma GDF15 level after 11 days of metformin administration may be dispensable for the metformin action on glucose tolerance and insulin sensitivity, because these metformin effects are preserved in GDF15 knockout mice (18). On the other hand, HFD mice receiving metformin via drinking water for 8–10 weeks increased glucose tolerance and decreased plasma insulin levels via GDF15 action (17). Thus, studies are needed to clarify the role of GDF15 and its receptor, GFRAL, in the AP in the long-term glucoregulatory effect of metformin and to investigate the potential contribution of kidney GDF15 in glucose regulation in parallel with feeding regulation (16).

Sjøberg et al. (8) reported that subcutaneous GDF15 injection in female mice fed a chow diet enhanced insulin sensitivity independent of changes in feeding and weight as assessed by a hyperinsulinemic-euglycemic clamps (8), consistent with the AP GDF15 effect on insulin sensitivity in HFD male rats reported in the present study (Fig. 1G and I). Although both studies report an enhancement of insulin sensitivity in lowering hepatic glucose production by GDF15, Sjøberg et al. additionally reported an increase of glucose uptake in the adipose tissue (8). The difference in observation of insulin-induced changes in glucose uptake may be due to differences in route of GDF15 delivery (AP versus subcutaneous), nutritional status (HF versus chow), species (rats versus mice), and/or sex; in the present study, we only used male rats. It would be important to continue to study the sufficient and necessary glucoregulatory effects of GDF15 in the AP of female and male rats as well as in chow- and HFD-fed conditions to appreciate the sex-dependent physiological and pharmacological relevance of brain GDF15 action.

In addition to the hyperinsulinemic clamp results, in contrast to our findings in the IVGTTs (Fig. 1B), Sjøberg et al. (8) did not observe an increase in glucose tolerance as assessed by an intraperitoneal glucose tolerance test (IPGTT). The discrepancy in observations might be explained by the minimal increase in plasma insulin levels in response to IPGTT versus the striking increase in plasma insulin levels during an IVGTT protocol (Fig. 1C) (8,19,20), because the minimal increase in plasma insulin levels incurred by IPGTT may limit the impact GDF15 has on insulin action. Additionally, the stress incurred by IPGTT and tail blood sampling could also potentially obscure the glucoregulatory effect of GDF15 (8), whereas in our IVGTT protocol, intravenous glucose administration and blood sampling were conducted in conscious, unconstrained, and nonstressed rats. Finally, Zhang et al. (11) reported that GDF15 enhances glucose-stimulated insulin secretion in islets, whereas we report here that acute GDF15 action in the AP did not alter insulin levels during the IVGTT. The difference in observation is likely due to the GDF15 glucoregulatory effect that we report in the AP is GFRAL dependent, whereas the effect of GDF15 on islet insulin secretion as shown by Zhang et al. is GFRAL independent.

In conclusion, our findings indicate that the activation of GFRAL in the AP by GDF15 increased glucose tolerance and insulin sensitivity, whereas genetic inhibition of GFRAL in the AP negated metformin glucoregulatory action, independent of feeding, weight, and insulin secretion (Fig. 3E). Thus, acute GFRAL activation in the AP contributes to GDF15 or metformin action to enhance glucose tolerance.

Funding. S.-Y.Z. is supported by a Canadian Institutes of Health Research (CIHR) postdoctoral fellowship. Z.D. and K.B. are supported by the Ontario Graduate Scholarship. K.B. and J.F.M.C. are supported by the Banting and Best Diabetes Centre graduate studentship. T.K.T.L. is supported by a CIHR grant (PJT-189957) for this project and holds a Tier 1 Canada Research Chair in Diabetes and Obesity at the Toronto General Hospital Research Institute and the University of Toronto.

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

Author Contributions. S.-Y.Z. and Z.D. conducted and designed experiments, performed data analyses, and wrote the manuscript. K.B. and J.F.M.C. assisted with the experiments. T.K.T.L. supervised the project, contributed to designing experiments, and edited the manuscript. T.K.T.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.