Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2020 Jul 1.
Published in final edited form as: J Cyst Fibros. 2019 Feb 7;18(4):557–565. doi: 10.1016/j.jcf.2019.01.010

Incretin dysfunction and hyperglycemia in cystic fibrosis: role of acyl-ghrelin

Xingshen Sun a,†, Yaling Yi a,†, Bo Liang a, Yu Yang a, Nan He a, Katie Larson Ode b,c, Aliye Uc b,c, Kai Wang d, Katherine N Gibson-Corley b,e, John F Engelhardt a,b,*, Andrew W Norris b,c,f,*
PMCID: PMC6591024  NIHMSID: NIHMS1521005  PMID: 30738804

Abstract

Background:

Insulin secretion is insufficient in cystic fibrosis (CF), even before diabetes is present, though the mechanisms involved remain unclear. Acyl-ghrelin (AG) can diminish insulin secretion and is elevated in humans with CF.

Methods:

We tested the hypothesis that elevated AG contributes to reduced insulin secretion and hyperglycemia in CF ferrets.

Results:

Fasting AG was elevated in CF versus non-CF ferrets. Similar to its effects in other species, AG administration in non-CF ferrets acutely reduced insulin, increased growth hormone, and induced hyperglycemia. During oral glucose tolerance testing, non-CF ferrets had responsive insulin, glucagon like peptide-1 (GLP-1) and gastric inhibitory polypeptide (GIP) levels and maintained normal glucose levels, whereas CF ferrets had insufficient responses and became hyperglycemic. Interestingly in wild-type ferrets, the acyl-ghrelin receptor antagonist [D-Lys3]-GHRP-6 impaired glucose tolerance, and abolished insulin, GLP-1, and GIP responses during glucose tolerance testing. By contrast, in CF ferrets [D-Lys3]-GHRP-6 improved glucose tolerance, enhanced the insulin-to-glucose ratio, but did not impact the already low GLP-1 and GIP levels.

Conclusions:

These results suggest a mechanism by which elevated AG contributes to CF hyperglycemia through inhibition of insulin secretion, an effect magnified by low GLP-1 and GIP. Interventions that lower ghrelin, ghrelin action, and/or raise GLP-1 or GIP might improve glycemia in CF.

Keywords: Cystic fibrosis related diabetes, growth hormone secretagogue receptor, glucagon-like peptide-1, glucose-dependent insulinotropic peptide, [D-Lys3]-GHRP-6

1. Introduction

By middle age, diabetes affects approximately 60–80% of individuals with cystic fibrosis (CF) who have the most common cystic fibrosis transmembrane conductance regulator (CFTR) mutations [1]. Furthermore, in childhood about half of patients with CF exhibit hyperglycemia during glucose tolerance testing [2,3]. Once present, diabetes worsens mortality rates for individuals with CF [1]. The pathophysiology of CF related diabetes (CFRD) is imperfectly understood, however insufficient insulin secretion is a key driver of hyperglycemia in CF [4].

Primary organ dysfunction in CF is most severe in the gastrointestinal tract, the exocrine pancreas, and the airway. Of these, gastrointestinal disease is typically the initial manifestation of CF in humans, mice, pigs, and ferrets [5]. CF gastrointestinal manifestations include inspissation of secretions, altered gut motility, and propensity for intestinal obstruction. Not surprisingly, secretion of various gut hormones is perturbed in CF. Among the gut hormones affected by CF are several potentially relevant to diabetes. Best studied is the impact of CF on the secretion of the incretins, which are hormones that enhance insulin secretion and are secreted by the gut in response to nutrient intake. Patients with CF sometimes have lower levels of the two major incretins, glucagon like peptide-1 (GLP-1) [6–8] and gastric inhibitory polypeptide (GIP) [7,8]. Since GLP-1 and GIP promote insulin secretion, it has been proposed that lower GLP-1 and GIP levels in CF may contribute to glycemic abnormalities [6,8]. However, less attention has been paid to acyl-ghrelin, which is a non-incretin gut hormone that inhibits insulin secretion and has been found to be elevated in patients with CF [9,10].

Ghrelin is a hormone secreted primarily by ghrelin cells in the stomach. Its classical actions are to increase growth hormone secretion and to increase appetite [11]. Circulating acyl-ghrelin levels are normally suppressed by feeding, as occurs in humans during standardized examination with mixed meal tolerance testing (MMTT) and oral glucose tolerance testing (OGTT) [12,13]. Ghrelin circulates in acylated and unacylated forms. The acylated form is created by ghrelin-O-acyltransferase (GOAT), an enzyme expressed in ghrelin expressing cells of the stomach. Only the acylated form of ghrelin is a high affinity ligand for the ghrelin receptor [14], officially named the growth hormone secretagogue receptor (GHSR). The classic actions of ghrelin are mediated primarily by its acylated form, which is the form elevated in CF [9,10]. Acyl-ghrelin has actions on islets that reduce insulin secretion [15–19]. Thus it is natural to postulate that the elevated acyl-ghrelin levels in CF contribute to hyperglycemia. However, the actions of acyl-ghrelin to reduce insulin secretion and raise glucose are partly offset by indirect acyl-ghrelin actions that enhance insulin secretion from beta cells, mediated by acyl-ghrelin action to stimulate GLP-1 secretion [20,21]. Given that GLP-1 secretion can be impaired in CF [6], it is possible that this indirect pathway of acyl-ghrelin to support insulin secretion is less active in CF. In such a case where GLP-1 secretion is impaired, elevated acyl-ghrelin in CF would be expected to have unopposed action to reduce insulin secretion.

For these reasons, we hypothesized that elevated acyl-ghrelin levels contribute to hyperglycemia in CF such that inhibition of GHSR signaling would improve glucose tolerance in CF. We have tested this hypothesis in the CF ferret, a model which exhibits histopathology in the gut, pancreas, and lung mirroring human disease, and furthermore develops spontaneous diabetic level hyperglycemia [22]. The aims of this study were to determine (i) if circulating AG is dysregulated in CF ferret in fasted and fed conditions; (ii) whether AG affects glucose homeostasis and related hormones in the ferret; and (iii) whether GHSR antagonism improves glucose tolerance in CF ferrets. To the best of our knowledge, metabolic aspects of the ghrelin axis have not been examined in ferrets. Thus an important ancillary aspect of this work was to examine whether the ghrelin axis proteins and physiology in normal ferrets are similar to that in humans.

2. Materials and Methods

2.1. Ferret Rearing

Wild-type (WT) and CFTR exon-10 disrupted (CF) ferrets [22] were housed in separate cages under controlled temperature (20–22°C) and long day light cycle (16 h light/8 h dark) with free access to water and food. Both male and female ferrets were studied. CF ferrets received extensive nutritional and antibiotic support as required for survival [23]. Non-CF ferrets were WT ferrets that served as controls by being co-reared with a paired CF ferret, receiving identical cares including CF nutritional and antibiotic support; these are referred to as “non-CF” in this manuscript. Some experiments were performed on WT ferrets reared conventionally (i.e. not co-reared with a CF ferret, receiving no CF nutritional or antibiotic support); these are referred to as WT in this manuscript. All animal experimentation was approved by the Institutional Animal Care and Use Committee.

2.2. Acyl-ghrelin and Ghrelin Receptor Antagonist (GRA) Administration

Ferrets were administered either acyl-ghrelin (AG) or ghrelin receptor antagonist (GRA). Rat acyl-ghrelin (RP10781, Genescript, Piscataway, NJ, USA) was prepared in saline and injected subcutaneously at a dose of 0.1 mg/kg [24] after 4 hour fast. The ghrelin receptor antagonist (GRA) [D-Lys3]-GHRP-6 (G4535, Sigma-Aldrich, St. Louis, MO, USA) was prepared freshly in saline and injected subcutaneously at a dose of 10 μmol/kg [20] after 4 hour fast.

2.3. Oral Glucose Tolerance Test (OGTT), and Mixed Meal Tolerance Test (MMTT)

OGTTs and MMTTs are used as standardized means to assess glycemic responses. OGTTs involve administration of only glucose and serve as the standard approach to assess glucose tolerance, including in humans with CF. MMTTs involve administration of carbohydrates, protein, and lipid, and thus may better model the impact of a typical meal. Because both tests employ delivery of enteral nutrients, they also assess gut hormone responses. We have previously standardized methods for OGTTs and MMTTs in ferrets fasted for 4 hours [22]. Briefly, an oral glucose dose of 1.75 g /kg body weight (glucose tolerance beverage, Thermo Scientific) was administered for OGTT. For MMTTs, mixed meals included Elecare (Abbott) and canned food (Fancy Feast). Each mixed meal contained a total of 0.026 g carbohydrate/cm2 estimated body surface area. Blood samples were collected by venipuncture during brief isoflurane anesthesia. Other analytes from the OGTT and MMTT data presented in Figure 1, including glucose and insulin, have been previously published [22]. At least one month’s recovery was allowed between tests for ferrets undergoing repeated measures.

Figure 1: Elevation of acyl-ghrelin in CF ferrets.

Figure 1:

Open in a new tab

Body weight (A) and acyl-ghrelin (B) in CF (CF) and non-CF ferrets aged 2–16 months. Acyl-ghrelin was measured in serum after 4 hours fast in co-reared CF:non-CF ferret pairs, some having measurements made at more than one age (A-D: N=10 ferret pairs; mean ± standard deviation shown by bars). The influence of CF status on acyl-ghrelin levels did not vary with weight (C) or age (D). Acyl-ghrelin levels during mixed meal tolerance tests “MMTT” (E-G), and oral glucose tolerance tests “OGTT” (H-J). Mixed meal (E) and glucose (H) administration occurred at time 0. The respective area under the curve “AUC” (F,I) and incremental area under the curve “iAUC” (G,J) are shown. (E-J: N=4–5 co-reared CF:non-CF ferret pairs aged 2–12 months; mean ± standard deviation shown; H-J data includes 2 repeated measures). Symbols summarizing statistical test results are conserved throughout all figures: * p<0.05 for difference between CF and non-CF; † p<0.05 for change in value compared to time 0; & p<0.05 for mean value differing from zero.

2.4. Measurement of Analytes

Blood glucose was measured by portable meter (One Touch Lifescan, Milpitas, CA, USA). Blood was collected in sodium heparin tubes on ice, stored at −80°C, and thawed only once for assay. Plasma acyl-ghrelin, glucagon, active glucagon-like peptide 1, gastric inhibitory polypeptide, and insulin were measured using a Milliplex Map Human Metabolic Hormone Magnetic Bead Panel (Millipore, Billerica, MA, USA) [22]. Plasma growth hormone was measured using a Rat/Mouse Growth Hormone ELISA kit (Millipore).

2.5. Immunohistochemistry

At necropsy, stomachs were fixed in 10% neutral buffered formalin, embedded in paraffin, and sectioned at 5 μm. Tissue sections were subsequently deparaffinized, rehydrated, boiled 20 min in 0.1 M sodium citrate (pH 6.0) for antigen retrieval, and blocked with 100 −120 μl per slide of PBS with 10% donkey serum + 0.5% Triton-X-100 for 1 hour at room temperature. Sections were incubated overnight at 4°C with anti-ghrelin (Abcam, Cambridge, MA, #ab129383) or anti-ghrelin-O-acyltransferase (anti-GOAT) (Abcam, #ab99449) antibodies. HRP conjugate secondary antibody (Abcam) was applied for 1 hour at room temperature. Slides were developed with ImmPACT Peroxidase Substrate (Vector Laboratories, Burlingame, CA). Slides were counterstained in hematoxylin and mounted with Ultramount Aqueous Permanent Mounting Medium (S1964, Dako, Santa Clara, CA). Digital whole image analysis was performed using a color deconvolution algorithm on Aperio slide scanning technology and software. Quantification is presented as percent immunopositive based on this Aperio algorithm and was completed by a board-certified veterinary pathologist.

2.6. RNA Quantification

A bead-based oligonucleotide probe set specific for ferret ghrelin mRNA was developed by Affymetrix (Santa Clara, CA). See Supplementary Material for RNA probe map. Stomach homogenates were prepared using the QuantiGene Sample Processing Kit (Affymetrix) and quantified by the QuantiGene Plex Assay Kit (Affymetrix), using peptidylprolyl isomerase B (PPIB) as the housekeeping gene by which to normalize expression for each sample.

2.7. Statistical Analyses

Data were analyzed using R 3.3.3 (The R Foundation for Statistical Computing). Graphed summary statistics are mean ± standard deviation. Area under the curve (AUC) and incremental area under the curve (iAUC) were calculated using the trapezoidal rule, subtracting the baseline area in the latter case. The normality of analytes was examined on the residuals of presented control data using the Shapiro-Wilk test, which rejected normality (p<0.05) only for ghrelin mRNA and growth hormone levels, but none of the other measures. Differences between groups was assessed by linear mixed modeling for datasets containing pairing and/or repeated measures using the lmerTest R package, and by linear modeling for datasets without pairing and repeated measures. The exception was assessment of differences in ghrelin mRNA, which were examined by the Wilcoxon-Mann-Whitney test, given the non-normal distribution of this analyte. For data undergoing multiple tests, false discovery rates, reported as pfdr, were determined using the Benjamini-Hochberg procedure. Changes over time were assessed by ANOVA-on-ranks using the nlme R package, followed by Westfall posthoc analysis using the multcomp R package assessing each time point versus time 0. The influence of weight or age to influence ghrelin and its differences between CF and non-CF ferrets was assessed by linear mixed modeling. The potential difference of population means from zero was assessed by the one-sample t-test. Protein sequence similarity to that of humans was calculated using the BLOSUM80 substitution matrix employing the protr R package [25]. Sequences were retrieved from NCBI.nih.gov on April 4–5, 2018.

3. Results

3.1. Ferrets with CF have Elevated Acyl-ghrelin

We examined body weight and plasma acyl-ghrelin levels in CFTR knockout (CF) and non-CF ferrets aged 2 to 16 months. As expected from published studies in CF ferrets [23], the average weight in CF ferrets was lower (CF 679±230 grams, non-CF 828±227 grams, mean±sd, p=0.018) than in non-CF ferrets (Figure 1A). Circulating acyl-ghrelin was elevated in CF compared to non-CF ferrets (Figure 1B, CF 178.7±110, non-CF 95.9±67, p=0.007). The higher levels of acyl-ghrelin in CF versus non-CF animals were not affected by body weight (Figure 1C, p=0.58) or by age (Figure 1D, p=0.09). We tested for changes in the regulation of acyl-ghrelin in CF and non-CF ferrets during MMTT and OGTT. Overall, acyl-ghrelin levels were higher in CF compared to non-CF animals across the MMTT time-course (Figure 1E, p=0.005). As expected, circulating acyl-ghrelin levels diminished during MMTT in non-CF ferrets (Figure 1E, p<0.001 and p=0.02 at 30 and 60 minutes, indicated by † symbol). Acyl-ghrelin levels also declined during MMTT in CF ferrets (Figure 1E, p=0.003, 0.002, and 0.04 at 30, 60 and 120 minutes). The area under the curve (AUC) and incremental AUC (iAUC) during MMTT did not differ between non-CF and CF ferrets (Figures 1F-G, p=0.15 and 0.24). The iAUC was negative in both genotypes (Figure 1G, mean differs from zero, p=0.02 for non-CF, p=0.002 for CF), consistent with suppression of acyl-ghrelin levels during MMTT. During OGTT, acyl-ghrelin levels were not statistically different between non-CF and CF ferrets (Figures 1H-J), though only CF ferrets exhibited a robust decrease in acyl-ghrelin (60 minute time point decreased compared to time 0, p=0.004, and negative iAUC p=0.03). Together these data (Figure 1A-J) show that circulating acyl-ghrelin is elevated in CF ferrets, just as in humans with CF, even though it remains responsive to suppression by nutrient intake. We thus examined the stomachs of CF ferrets for changes in ghrelin producing cells. The ghrelin-positive cellular content and ghrelin mRNA in CF stomach did not differ from non-CF stomachs (Figure 2A-F, p=0.50 for difference in ghrelin area, p=0.065 for difference in ghrelin mRNA). Likewise, CF stomachs exhibited normal content and pattern of ghrelin-O-acyltransferase (GOAT) expression (Figure 2G-H) that did not qualitatively differ from that in non-CF stomachs.

Figure 2: Normal ghrelin and ghrelin-O-acyltransferase (GOAT) expression in CF stomach.

Figure 2:

Open in a new tab

Representative immunohistochemistry images from ghrelin (A-D) and GOAT (G-H) stained (dark brown) stomach sections from non-CF (A,C,G) and CF (B, D, H) ferrets. Scale bars: 50 μm. (E) Percentage ghrelin positive area in non-CF and CF stomach (n=5–7). (F) Ghrelin mRNA expression in non-CF and CF stomach (n=6 ). Bars show the mean ± standard deviation.

3.2. Conservation of the Ghrelin Axis in Ferrets

To better understand the conservation of the ghrelin axis between ferrets and other organisms, we examined the sequence of the three proteins forming the core ghrelin axis: ghrelin, GHSR, and GOAT. These proteins showed strong conservation across mammals, including the ferret in which the ghrelin sequence is nearly identical to that of humans (Figure 3A). We also examined whether ferrets experience human-like responses to administration of acyl-ghrelin. Subcutaneous injection of 0.1 mg/kg acyl-ghrelin produced a rapid supraphysiological increase in serum acyl-ghrelin (Figure 3B). As expected [11], acyl-ghrelin administration rapidly and potently raised circulating growth hormone levels in non-CF ferrets (Figure 3C, p<0.0001 for changes at 15 and 30 minutes, and p=0.0015 at 45 minutes). Consistent with effects of acyl-ghrelin in humans [15], insulin was rapidly and briefly suppressed (Figure 3D, p=0.002 for altered level at 15 minutes). Glucose levels climbed modestly after acyl-ghrelin (Figure 3E, p=0.03 at 15 and 90 minutes, p<0.001 at 30, 45, and 60 minutes). Although GLP-1 (Figure 3F) and glucagon (not shown) were not significantly altered by acyl-ghrelin administration, GIP levels were slowly diminished (Figure 3G).

Figure 3: Conservation of the ghrelin axis and acyl-ghrelin actions in wild type ferrets.

Figure 3:

Open in a new tab

(A) Protein sequence of ghrelin, growth hormone secretagogue receptor (GHSR), and ghrelin-O-acyltransferase (GOAT) across a variety of species, ordered by similarity (“Sim”) to human ghrelin. Human ghrelin sequence is indicated. Conserved ghrelin residues are indicated by a period. Sequence differences are indicated by the substitution, with greater degrees of dissimilarity indicated by darker shading. Shaded spaces indicate deletion. (B-D) Responses to acyl-ghrelin administration at time 0. Circulating (B) acyl-ghrelin, (C) growth hormone (GH), (D) insulin, (E) glucose, (F) GLP-1, and (G) GIP were measured. (N=3 (B) or N=6 (C-G) WT ferrets aged 2–12 months; mean ± standard deviation shown; † p<0.05 for change in value compared to time 0;)

3.3. Ghrelin Signaling Antagonism Improves Glucose Tolerance in CF but not WT Ferrets

Given that acyl-ghrelin increased glucose levels and that acyl-ghrelin levels were elevated in CF animals, we hypothesized that a ghrelin receptor antagonist might improve glucose levels in CF animals. We thus performed oral glucose tolerance tests in CF and wild-type (WT) ferrets after acute treatment with [D-Lys3]-GHRP-6, a ghrelin receptor antagonist (GRA), or saline. WT animals treated with saline exhibited normal blood glucoses which did not exceed 150 mg/dL at any point (Figure 4A, 120 minute glucose 126±9 mg/dL). GRA induced marked hyperglycemia in WT animals, into the diabetic range (Figure 4A,C; 120 minute glucose 340±41 mg/dL; pfdr=0.001, 0.013, 0.005, and 0.002 for difference between GRA and saline at 15, 30, 60, and 120 minutes; p<0.001 for increased AUC and iAUC). In contrast to WT animals, saline treated CF ferrets exhibited relative hyperglycemia during OGTT (Figure 4B, glucose peak of 206±49 mg/dL at 60 minutes). Critically, GRA had the opposite effect on glycemia in CF versus WT ferrets. GRA treated CF ferrets had improved glucoses during the intermediary OGTT time points (Figure 4B, pfdr= 0.029, 0.026, and 0.029 at 15, 30, and 60 minutes for difference between GRA and saline groups). To better understand these differential changes in glycemia with GRA treatment, we examined insulin levels. Plasma insulin levels rose in response to oral glucose in saline treated WT and CF ferrets (Figure 4E,F) reflected by a positive incremental area under the curve (non-zero iAUC saline bars, Figure 4G,H, p=0.0004 and 0.02 for WT and CF). However, insulin levels failed to rise in response to glucose after GRA treatment in WT ferrets (Figure 4E,G; GRA iAUC not significantly different than zero, p=0.23). By contrast, insulin increased in response to glucose challenge in GRA-treated CF ferrets (Figure 4F,H; iAUC positive for GRA, p=0.00003), despite a modest reduction in insulin at the 15 minute time point (Figure 4F, 17% lower in GRA group at 15 minutes, pfdr=0.041) though coupled to a 25% reduction in glucose at this time point (Figure 4B). To visualize the relation of these differential GRA-induced glucose and insulin changes in greater detail, we plotted all data points from the OGTT at 30 minutes (Figure 5). In this figure, tests that were performed on the same ferret are connected with a line. For WT, no ferrets underwent repeat testing and thus there are no connecting lines. The increase in glucose induced by GRA in WT ferrets is apparent (Figure 5A, from 145±28 to 259±38 mg/dL, p=0.013), as is the glucose decrement induced by GRA in CF ferrets (Figure 5B, from 189±30 to 129±34, p=0.007). Insulin at 30 minutes was unchanged by GRA in WT (Figure 5C, 408±65 versus 192±158 pg/mL, p=0.18) and in CF animals (Figure 5D, 993±404 versus 980±626, p=0.93 ). To better envisage the relationship between glucose and insulin levels, we plotted the insulin-to-glucose ratios. The insulin-to-glucose ratio was diminished by GRA in WT ferrets (Figure 5E, 2.75±0.59 to 0.74±0.36, p=0.016 ), but was increased in CF ferrets (Figure 5F, from 5.1±1.4 to 8.2±2.9, p=0.039).

Figure 4: The effect of ghrelin receptor antagonist (GRA) on oral glucose tolerance and insulin secretion.

Figure 4:

Open in a new tab

Blood glucose (A-D) and serum insulin (E-H) during oral glucose tolerance testing in ferrets treated with saline (closed symbols) or GRA (open symbols). GRA was administered at −15 minute and glucose administered at 0 minutes. Responses in wild type (WT, N=3) (A,C,E,G) and CF (N=5–6) (B,D,F,H) ferrets are shown. Mean ± standard deviation shown. Symbols summarizing statistical test results are conserved throughout all figures: # pfdr<0.05 for difference between GRA and saline at the indicated time points; # p<0.05 for AUC or iAUC difference between GRA and saline; & p<0.05 for mean value differing from zero.

Figure 5: Glucose and insulin levels and their relation at 30 minutes during oral glucose tolerance test.

Figure 5:

Open in a new tab

Data are the same as presented for the 30 minute time point in Figures 4. Values from wild-type (A, C, E) are shown on the left, and CF (B, D, F) are shown on the right. Measures are glucose (A,B), insulin (C,D), and the insulin-to-glucose ratio (E,F). Each data point is shown. Values collected in the same ferret are connected by a line. # p<0.05 for difference between GRA and saline treated ferrets.

3.4. Both CF and Ghrelin Receptor Antagonist Reduce Intestinal Incretin Levels

To better understand how CF modifies GRA responses, we measured levels of the intestinal incretins glucagon-like peptide-1 (GLP-1) and gastric inhibitory polypeptide (GIP). As expected, oral glucose increased the levels of GLP-1 and GIP in saline-treated WT ferrets (Figure 6A,C,E,G; non-zero saline iAUC p<0.00001 and p=0.02 for GLP1 and GIP). However GRA treatment abolished these responses (Figure 6A,C,E,G; reduced AUC p=0.02 and p=0.002, iAUC not differing from zero p=0.34 and 0.50 for GLP-1 and GIP respectively). CF differed from WT ferrets in that GLP-1 and GIP levels were non-responsive to OGTT even under saline-treated conditions (Figure 6B,D,F,H; iAUC not statistically different than zero, p=0.09 and p=0.30 for GLP-1 and GIP). GRA-treatment produced little additional change in GLP-1 and GIP levels in CF ferrets (Figure 6B,D,F,H; AUC unchanged p=0.41 and p=0.35, iAUC unchanged p=0.76 and p=0.37, iAUC not statistically different than zero p=0.09 and p=0.15, for GLP-1 and GIP respectively).

Figure 6: The effect of ghrelin receptor antagonist (GRA) on incretin responses during oral glucose tolerance testing.

Figure 6:

Open in a new tab

Plasma active glucagon-like peptide-1 (GLP-1) (A-D) and gastric inhibitory polypeptide (GIP) (E-H) during oral glucose tolerance testing in WT (A,C,E,G; N=3) and CF (B,D,F,H; N=5–6) ferrets. Saline-treated animals are represented in solid black symbols; GRA-treated animals, hollow symbols. GRA was administered at −15 minute and glucose administered at 0 minutes. Symbols summarizing statistical test results are conserved throughout all figures: # pfdr<0.05 for time point difference between GRA and saline; # p<0.05 for AUC or iAUC difference between GRA and saline; & p<0.05 for mean value differing from zero.

4. Discussion

CF incurs a strong predilection for hyperglycemia and diabetes, in large part via impaired insulin secretion. However, insulin secretion is diminished in CF to a degree exceeding that expected based solely on loss of beta cell mass [26]. Because gut dysfunction and alteration of gut hormone levels are prominent in CF, it has been postulated that low GLP-1 and GIP levels might contribute to impaired insulin secretion [6,8]. The data presented here extends this gut hormone hypothesis by providing evidence that elevated ghrelin axis activity may contribute to hyperglycemia and poor insulin secretion in CF.

Our results show that acute treatment of CF ferrets with a ghrelin receptor antagonist (GRA) lowers blood glucose levels during OGTT. Although insulin levels were generally unchanged by GRA treatment in CF animals, there was a 60% increase in the ratio of insulin-to-glucose at 30 minutes in CF animals treated with GRA (Figure 5F). These results suggest that GRA treatment improves glucose levels by enhancing the amount of insulin secretion in response to glucose levels. However, definitive demonstration of this would require testing under standardized conditions, such as with hyperglycemic clamp. Interestingly, the effects of GRA in WT ferrets were opposite of those in CF ferrets. GRA induced diabetic-level glucose intolerance in WT ferrets and abolished the responsiveness of insulin to glucose load. This is striking, because normally hyperglycemia should have provoked marked increases in insulin levels. Instead, GRA treatment in WT ferrets markedly reduced the ratio of insulin-to-glucose. As detailed further in discussion, differences in incretin responses between CF and WT ferrets may account for the diametrically opposing impact of GRA on glycemia between these two genotypes.

We found that acyl-ghrelin was elevated in CF ferrets, just as has been observed in humans with CF [9,10]. One prior report found reduced ghrelin levels in humans with CF [27], however the methods used did not distinguish acyl- from desacyl-ghrelin, implying that desacyl-ghrelin levels may be reduced in CF. The mechanism by which acyl-ghrelin is elevated in CF is unknown, though insufficient weight gain has been postulated to be the upstream cause [10]. We found no significant difference in the content of ghrelin and GOAT expressing cells of the stomach between non-CF and CF ferrets. GOAT exhibited a much wider pattern of expression compared to ghrelin (Figure 2GH versus 2AB), such that many GOAT expressing cells do not express ghrelin. This wide pattern of GOAT expression also occurs in rats but not mice [28]. Elevated ghrelin levels have also been observed in other chronic diseases associated with underweight status [29]. CF ferrets exhibited elevated acyl-ghrelin levels across a range of weights. This result is potentially consistent with underweight status as the cause of increased acyl-ghrelin, since CF ferrets exhibit impaired weight gain throughout life despite aggressive nutritional management including pancreatic enzyme replacement and supplementation with Elecare formula [23]. Postprandial suppression of acyl-ghrelin is mediated in part by hyperosmolar luminal exposure of the small intestine [30]. Interestingly, postprandial suppression of acyl-ghrelin was intact in the CF ferrets, perhaps supported by their treatment with pancreatic enzyme replacement as these enzymes contribute to hydrolysis of foods to nutrient constituents of higher osmolality. The impact of pancreatic enzyme supplementation on acyl-ghrelin levels in CF has not been studied, and all published studies investigating acyl-ghrelin levels in CF humans treated all pancreatic insufficient patients with pancreatic enzyme replacement therapy [9,10].

Prior to this study, the metabolic aspects of the ghrelin axis had not been studied in ferrets. Even among mammals, there are significant interspecies differences in gut peptides. For example, mice and rats lack motilin [31], which is an intestinally produced peptide that is most closely related to ghrelin. Our findings show that not only are the ghrelin axis protein sequences (ghrelin, GHSR, and GOAT) closely related between ferrets and humans, but also that the metabolic physiology of the ghrelin axis in ferrets closely resembles that of humans, including acyl-ghrelin stimulation of growth hormone and suppression of insulin, and nutrient induced suppression of acyl-ghrelin levels.

When interpreting the data in Figures 4–5, it is important to bear in mind that conventional rearing (i.e. “WT” ferrets, fed normal chow and not given medicinal support provided to CF ferrets) results in greater insulin sensitivity than co-rearing (i.e. “non-CF” ferrets, fed a CF-centric diet and given CF medications), as demonstrated by euglycemic hyperinsulinemic clamp of WT (Sui et al. 2014) versus non-CF (Yi et al. 2016) ferrets. As predicted by this, fasting insulin levels in WT ferrets (Figure 4, 179±45 pg/mL, n=6) are substantively lower than for non-CF ferrets (~300–500 pg/mL, as published in Figure 4K,L (Yi et al. 2016) ). Thus, we caution against comparing absolute insulin levels between CF versus WT ferrets shown in Figures 4–5, as the WT ferrets are expected to have lower insulin levels due to their greater insulin sensitivity. The more relevant question is whether adequate insulin secretion occurs to prevent hyperglycemia. In WT ferrets, very modest insulin secretion is sufficient to prevent hyperglycemia (Figure 4A,E, saline treated), indicating their substantial insulin sensitivity. By contrast, in CF ferrets, although insulin levels are higher than in WT ferrets, hyperglycemia occurs by 30 minutes (Figure 4B,F, saline treated) indicating that insulin secretion, especially early in the course, has been insufficient in the context of their rearing environment. This particular inference has been tested previously under controlled and standardized conditions; CF ferrets exhibited markedly less insulin secretion than control co-reared non-CF ferrets during hyperglycemic clamp (Yi et al. 2016). Likewise, insulin sensitivity of CF ferrets was identical to that of co-reared non-CF ferrets when examined by euglycemic hyperinsulinemic clamp, indicating that the hyperglycemia experienced by CF ferrets is due primarily to impaired insulin secretion. In other words, the degree of insulin resistance imparted by the CF ferret diet and medication regimen is easily compensated for in non-CF ferrets via appropriate increases in insulin secretion, whereas insulin secretion in CF ferrets is unable to compensate sufficiently resulting in hyperglycemia. Importantly for the current investigation, ghrelin receptor antagonist acutely improved glucose levels in CF ferrets during OGTT, associated with increased insulin-to-glucose levels early in the time course. Taken together, these results suggest that hyperglycemia driven by insufficient insulin secretion is at least partly ameliorated by blocking acyl-ghrelin receptor activity in CF ferrets, consistent with the major hypothesis of this study.

The impact of acyl-ghrelin on insulin secretion is multifold. On one hand, acyl-ghrelin dampens insulin secretion from islets, with recent work finding two potential involved mechanisms. In beta cells, GHSR reduces insulin secretion by forming hybrid complexes with somatostatin receptors that signal via Gαi2 to inhibit cAMP accumulation and glucose-stimulated insulin secretion [17]. At the islet level, acyl-ghrelin inhibits insulin secretion via stimulation of somatostatin secretion from delta cells, which robustly express GHSR [18,19]. Acyl-ghrelin also has indirect effects that enhance insulin secretion. Namely, acyl-ghrelin augments meal-induced secretion of GLP-1 [20,32], a hormone secreted by intestinal L-cells that promotes glucose stimulated insulin secretion. Thus in toto, acyl-ghrelin exerts both positive and negative actions on insulin secretion. In CF however, GLP-1 levels are often reduced as they were in our study, in which case the inhibitory actions of acyl-ghrelin on insulin secretion would be unopposed. This predicts, just as we observed, that inhibition of acyl-ghrelin action via GHSR antagonist (GRA) would improve the secretion of insulin in response to glucose raising the ratio of insulin-to-glucose in CF.

Acyl-ghrelin has been shown to augment secretion of meal-induced GLP-1 secretion [20,21]. Interestingly, we found that a supra-pharmacologic dose of acyl-ghrelin had no effect on non-fed GLP-1 levels (Figure 3). GIP levels were diminished by acyl-ghrelin, though this was a later effect and might represent a secondary action.

We were surprised by the impaired glucose tolerance and reduced insulin levels induced by GRA in WT ferrets, especially given that ghrelin antagonists do not worsen insulin secretion or glucose tolerance in a variety of non-diabetic mouse models.. However, this finding is not without precedent, as GHSR antagonist worsens insulin secretion in fed sheep [33], causes hyperglycemia in some rodent models [34] and ablation of GHSR in ob/ob mice worsens hyperglycemia and reduces insulin secretion [35]. The ultimate effect of acyl-ghrelin actions on beta cell insulin secretion will depend on the relative balance of somatostatin, GHSR, and GLP-1 tone. Worsening of insulin secretion with GHSR antagonism, as we observed in normal ferrets, could thus indicate that the GLP-1 stimulatory actions of acyl-ghrelin are the dominant beta-cell influence relative to inhibitory effects in normal ferrets during OGTT, though this remains to be tested. Our results are consistent with a role for acyl-ghrelin in supporting GLP-1 secretion in healthy ferrets, in that GHSR antagonist led to a marked reduction of GLP-1 levels during OGTT. Our results also show that GHSR inhibitor markedly decreases GIP levels. To our knowledge, the effects of ghrelin antagonism on GIP secretion have not been previously studied.

The beneficial effects we observed of inhibiting acyl-ghrelin action via GHSR antagonist in CF ferrets was associated with a partial improvement in insulin-to-glucose ratio. Thus, enhanced insulin secretion in response to glucose may be the mechanism by which glucose levels were improved. However, a possible additional mechanism by which acyl-ghrelin may also affect glucose levels is through alteration of gut motility. In general acyl-ghrelin stimulates gut motility [36], whereas GLP-1 reduces motility [37]. Thus, in CF where acyl-ghrelin is elevated and GLP-1 diminished, one might expect increased motility. In such a case GHSR antagonist might reduce motility and thus perhaps blunt the acute hyperglycemic effects of oral glucose. This potential mechanism is speculative and remains to be tested. Acyl-ghrelin is a potent stimulator of appetite. Because reducing appetite is often not desirous in CF, use of GHSR inhibition to improve glucose in CF would ideally be targeted selectively to islets to avoid inhibiting acyl-ghrelin’s orexigenic actions on appetite centers in the brain.

In summary, we find that acyl-ghrelin levels are elevated in CF ferrets, just as they are in humans with CF. The ghrelin axis and acute physiological responses to acyl-ghrelin are conserved in ferrets. Importantly, a ghrelin receptor antagonist improved glucose levels in CF animals associated with an enhanced insulin-to-glucose ratio and no further loss of GLP-1 of GIP. By contrast, glucose tolerance and the insulin-to-glucose ratio were markedly impaired in WT animals by ghrelin receptor antagonist, associated with loss of GLP-1 and GIP responses to oral glucose. Taken together, these results suggest that inhibition of acyl-ghrelin signaling, especially in pancreatic islets, is a novel approach that may improve glucose levels in CF.

Supplementary Material

1

Highlights.

  • Like in humans, serum acyl-ghrelin levels were elevated in CF ferrets.

  • Treatment of CF ferrets with ghrelin receptor antagonist (GRA) lowered blood sugar.

  • Treatment of CF ferrets with GRA raised the insulin-to-glucose ratio.

  • GRA had the opposite effects in normal ferrets.

  • Altered beta-cell supporting gut hormones may account for these differences.

5. Acknowledgements

The authors thank the Comparative Pathology Laboratory in the Department of Pathology for their technical assistance.

Financial Support: This work was supported by NIH grant R24 DK096518 (to J.F.E., A.W.N.), R01 DK097820 (to A.U., A.W.N.), R01 DK115791 (to A.W.N., J.F.E.), a Fraternal Order of Eagles Diabetes Research Center scholar award (to A.W.N.), the University of Iowa Center for Gene Therapy (DK54759), National Ferret Resource and Research Center (R24 HL123482), and the Carver Chair in Molecular Medicine (to J.F.E).

Abbreviations

(AG)

Acyl-ghrelin

(CFRD)

cystic fibrosis related diabetes

(GIP)

gastric inhibitory polypeptide

(GLP-1)

glucagon like peptide-1

(GHSR)

growth hormone secretagogue receptor

(MMTT)

mixed meal tolerance test

(OGTT)

oral glucose tolerance tests

Footnotes

Conflict of interest statement: The authors have nothing to disclose.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

6. References

  • [1].Lewis C, Blackman SM, Nelson A, Oberdorfer E, Wells D, Dunitz J, et al. Diabetes-related mortality in adults with cystic fibrosis. Role of genotype and sex. Am J Respir Crit Care Med 2015;191:194–200. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [2].Yi Y, Norris AW, Wang K, Sun X, Uc A, Moran A, et al. Abnormal Glucose Tolerance in Infants and Young Children with Cystic Fibrosis. Am J Respir Crit Care Med 2016;194:974–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [3].Ode KL, Frohnert B, Laguna T, Phillips J, Holme B, Regelmann W, et al. Oral glucose tolerance testing in children with cystic fibrosis. Pediatr Diabetes 2010;11:487–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [4].Street ME, Spaggiari C, Ziveri MA, Rossi M, Volta C, Viani I, et al. Insulin production and resistance in cystic fibrosis: effect of age, disease activity, and genotype. J Endocrinol Invest 2012;35:246–53. [DOI] [PubMed] [Google Scholar]
  • [5].Keiser NW, Engelhardt JF. New animal models of cystic fibrosis: what are they teaching us? Curr Opin Pulm Med 2011;17:478–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [6].Hillman M, Eriksson L, Mared L, Helgesson K, Landin-Olsson M. Reduced levels of active GLP-1 in patients with cystic fibrosis with and without diabetes mellitus. J Cyst Fibros 2012;11:144–9. [DOI] [PubMed] [Google Scholar]
  • [7].Kuo P, Stevens JE, Russo A, Maddox A, Wishart JM, Jones KL, et al. Gastric emptying, incretin hormone secretion, and postprandial glycemia in cystic fibrosis--effects of pancreatic enzyme supplementation. J Clin Endocrinol Metab 2011;96:E851–5. [DOI] [PubMed] [Google Scholar]
  • [8].Sheikh S, Gudipaty L, De Leon DD, Hadjiliadis D, Kubrak C, Rosenfeld NK, et al. Reduced β-Cell Secretory Capacity in Pancreatic Insufficient, But Not Pancreatic Sufficient, Cystic Fibrosis Despite Normal Glucose Tolerance. Diabetes 2017;66:134–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [9].Monajemzadeh M, Mokhtari S, Motamed F, Shams S, Ashtiani MTH, Abbasi A, et al. Plasma ghrelin levels in children with cystic fibrosis and healthy children. Arch Med Sci 2013;9:93–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [10].Cohen RI, Tsang D, Koenig S, Wilson D, McCloskey T, Chandra S. Plasma ghrelin and leptin in adult cystic fibrosis patients. J Cyst Fibros 2008;7:398–402. [DOI] [PubMed] [Google Scholar]
  • [11].Shintani M, Ogawa Y, Ebihara K, Aizawa-Abe M, Miyanaga F, Takaya K, et al. Ghrelin, an endogenous growth hormone secretagogue, is a novel orexigenic peptide that antagonizes leptin action through the activation of hypothalamic neuropeptide Y/Y1 receptor pathway. Diabetes 2001;50:227–32. [DOI] [PubMed] [Google Scholar]
  • [12].Huang L, Tong Y, Zhang F, Yang Q, Li D, Xie S, et al. Increased acyl ghrelin but decreased total ghrelin and unacyl ghrelin in Chinese Han people with impaired fasting glucose combined with impaired glucose tolerance. Peptides 2014;60:86–94. [DOI] [PubMed] [Google Scholar]
  • [13].Parvaresh Rizi E, Loh TP, Baig S, Chhay V, Huang S, Caleb Quek J, et al. A high carbohydrate, but not fat or protein meal attenuates postprandial ghrelin, PYY and GLP-1 responses in Chinese men. PLoS One 2018;13:e0191609. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [14].Kojima M, Hosoda H, Date Y, Nakazato M, Matsuo H, Kangawa K. Ghrelin is a growth-hormone-releasing acylated peptide from stomach. Nature 1999;402:656–60. [DOI] [PubMed] [Google Scholar]
  • [15].Tong J, Prigeon RL, Davis HW, Bidlingmaier M, Kahn SE, Cummings DE, et al. Ghrelin suppresses glucose-stimulated insulin secretion and deteriorates glucose tolerance in healthy humans. Diabetes 2010;59:2145–51. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [16].Damdindorj B, Dezaki K, Kurashina T, Sone H, Rita R, Kakei M, et al. Exogenous and endogenous ghrelin counteracts GLP-1 action to stimulate cAMP signaling and insulin secretion in islet β-cells. FEBS Lett 2012;586:2555–62. [DOI] [PubMed] [Google Scholar]
  • [17].Park S, Jiang H, Zhang H, Smith RG. Modification of ghrelin receptor signaling by somatostatin receptor-5 regulates insulin release. Proc Natl Acad Sci U S A 2012;109:19003–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [18].Adriaenssens AE, Svendsen B, Lam BYH, Yeo GSH, Holst JJ, Reimann F, et al. Transcriptomic profiling of pancreatic alpha, beta and delta cell populations identifies delta cells as a principal target for ghrelin in mouse islets. Diabetologia 2016;59:2156–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [19].DiGruccio MR, Mawla AM, Donaldson CJ, Noguchi GM, Vaughan J, Cowing-Zitron C, et al. Comprehensive alpha, beta and delta cell transcriptomes reveal that ghrelin selectively activates delta cells and promotes somatostatin release from pancreatic islets. Mol Metab 2016;5:449–58. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [20].Gagnon J, Baggio LL, Drucker DJ, Brubaker PL. Ghrelin Is a Novel Regulator of GLP-1 Secretion. Diabetes 2015;64:1513–21. [DOI] [PubMed] [Google Scholar]
  • [21].Page LC, Gastaldelli A, Gray SM, D’Alessio DA, Tong J. Interaction of GLP-1 and Ghrelin on Glucose Tolerance in Healthy Humans. Diabetes 2018;67:1976–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [22].Yi Y, Sun X, Gibson-Corley K, Xie W, Liang B, He N, et al. A Transient Metabolic Recovery from Early Life Glucose Intolerance in Cystic Fibrosis Ferrets Occurs During Pancreatic Remodeling. Endocrinology 2016;157:1852–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [23].Sun X, Olivier AK, Liang B, Yi Y, Sui H, Evans TIA, et al. Lung phenotype of juvenile and adult cystic fibrosis transmembrane conductance regulator-knockout ferrets. Am J Respir Cell Mol Biol 2014;50:502–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [24].Chuang J-C, Sakata I, Kohno D, Perello M, Osborne-Lawrence S, Repa JJ, et al. Ghrelin directly stimulates glucagon secretion from pancreatic alpha-cells. Mol Endocrinol 2011;25:1600–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [25].Xiao N, Cao D-S, Zhu M-F, Xu Q-S. protr/ProtrWeb: R package and web server for generating various numerical representation schemes of protein sequences. Bioinformatics 2015;31:1857–9. [DOI] [PubMed] [Google Scholar]
  • [26].Iannucci A, Mukai K, Johnson D, Burke B. Endocrine pancreas in cystic fibrosis: an immunohistochemical study. Hum Pathol 1984;15:278–84. [DOI] [PubMed] [Google Scholar]
  • [27].Stylianou C, Galli-Tsinopoulou A, Koliakos G, Fotoulaki M, Nousia-Arvanitakis S. Ghrelin and leptin levels in young adults with cystic fibrosis: relationship with body fat. J Cyst Fibros 2007;6:293–6. [DOI] [PubMed] [Google Scholar]
  • [28].Stengel A, Goebel M, Wang L, Taché Y, Sachs G, Lambrecht NWG. Differential distribution of ghrelin-O-acyltransferase (GOAT) immunoreactive cells in the mouse and rat gastric oxyntic mucosa. Biochem Biophys Res Commun 2010;392:67–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [29].Itoh T, Nagaya N, Yoshikawa M, Fukuoka A, Takenaka H, Shimizu Y, et al. Elevated plasma ghrelin level in underweight patients with chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2004;170:879–82. [DOI] [PubMed] [Google Scholar]
  • [30].Overduin J, Tylee TS, Frayo RS, Cummings DE. Hyperosmolarity in the small intestine contributes to postprandial ghrelin suppression. Am J Physiol Gastrointest Liver Physiol 2014;306:G1108–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [31].Sanger GJ, Holbrook JD, Andrews PLR. The translational value of rodent gastrointestinal functions: a cautionary tale. Trends Pharmacol Sci 2011;32:402–9. [DOI] [PubMed] [Google Scholar]
  • [32].Tong J, Davis HW, Gastaldelli A, D’Alessio D. Ghrelin Impairs Prandial Glucose Tolerance and Insulin Secretion in Healthy Humans Despite Increasing GLP-1. J Clin Endocrinol Metab 2016;101:2405–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [33].Takahashi H, Kurose Y, Sakaida M, Suzuki Y, Kobayashi S, Sugino T, et al. Ghrelin differentially modulates glucose-induced insulin secretion according to feeding status in sheep. J Endocrinol 2007;194:621–5. [DOI] [PubMed] [Google Scholar]
  • [34].Mosa R, Huang L, Li H, Grist M, LeRoith D, Chen C. Long-term treatment with the ghrelin receptor antagonist [d-Lys3]-GHRP-6 does not improve glucose homeostasis in nonobese diabetic MKR mice. Am J Physiol Regul Integr Comp Physiol 2018;314:R71–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [35].Ma X, Lin Y, Lin L, Qin G, Pereira FA, Haymond MW, et al. Ablation of ghrelin receptor in leptin-deficient ob/ob mice has paradoxical effects on glucose homeostasis when compared with ablation of ghrelin in ob/ob mice. Am J Physiol Endocrinol Metab 2012;303:E422–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • [36].Bülbül M, Babygirija R, Zheng J, Ludwig K, Xu H, Lazar J, et al. Food intake and interdigestive gastrointestinal motility in ghrelin receptor mutant rats. J Gastroenterol 2011;46:469–78. [DOI] [PubMed] [Google Scholar]
  • [37].Tolessa T, Gutniak M, Holst JJ, Efendic S, Hellström PM. Inhibitory effect of glucagon-like peptide-1 on small bowel motility. Fasting but not fed motility inhibited via nitric oxide independently of insulin and somatostatin. J Clin Invest 1998;102:764–74. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1521005-supplement-1.pdf (70.6KB, pdf)

RESOURCES