Comparing methods to normalize insulin secretion shows the process may not be needed

in Journal of Endocrinology
View More View Less
  • 1 Department of Biomedical Sciences, Heritage College of Osteopathic Medicine, Ohio University, Athens, Ohio, USA
  • | 2 Diabetes Institute, Heritage College of Osteopathic Medicine, Ohio University, Athens, Ohio, USA

Correspondence should be addressed to C S Nunemaker: nunemake@ohio.edu
Free access

Glucose-stimulated insulin secretion (GSIS) is a well-accepted method to investigate the physiological and pathophysiological function of islets. However, there is little consensus about which method is best for normalizing and presenting GSIS data. In this study, we evaluated the sufficiency of islet area, total protein, total DNA and total insulin content as parameters to normalize GSIS data. First, we tested if there is a linear correlation between each parameter and the number of islets (10, 20, 30 and 40 islets). Islet area, total protein and insulin content produced excellent linear correlations with islet number (R 2 > 0.9 for each) from the same islet material. Insulin secretion in 11 mM glucose also correlated reasonably well for islet area (R 2 = 0.69), protein (R 2 = 0.49) and insulin content (R 2 = 0.58). DNA content was difficult to reliably measure and was excluded from additional comparisons. We next measured GSIS for 18 replicates of 20 islets each, measuring 3 mM and 11 mM glucose to calculate the stimulation index and to compare each normalization parameter. Using these similar islet masses for each replicate, none of the parameters produced linear correlations with GSIS (R 2 < 0.05), suggesting that inherent differences in GSIS dominate small differences in islet mass. We conclude that when comparing GSIS for islets of reasonably similar size (<50% variance), normalization does not improve the representation of GSIS data. Normalization may be beneficial when substantial differences in islet mass are involved. In such situations, we suggest that using islet cross-sectional area is superior to other commonly used techniques for normalizing GSIS data.

Abstract

Glucose-stimulated insulin secretion (GSIS) is a well-accepted method to investigate the physiological and pathophysiological function of islets. However, there is little consensus about which method is best for normalizing and presenting GSIS data. In this study, we evaluated the sufficiency of islet area, total protein, total DNA and total insulin content as parameters to normalize GSIS data. First, we tested if there is a linear correlation between each parameter and the number of islets (10, 20, 30 and 40 islets). Islet area, total protein and insulin content produced excellent linear correlations with islet number (R 2 > 0.9 for each) from the same islet material. Insulin secretion in 11 mM glucose also correlated reasonably well for islet area (R 2 = 0.69), protein (R 2 = 0.49) and insulin content (R 2 = 0.58). DNA content was difficult to reliably measure and was excluded from additional comparisons. We next measured GSIS for 18 replicates of 20 islets each, measuring 3 mM and 11 mM glucose to calculate the stimulation index and to compare each normalization parameter. Using these similar islet masses for each replicate, none of the parameters produced linear correlations with GSIS (R 2 < 0.05), suggesting that inherent differences in GSIS dominate small differences in islet mass. We conclude that when comparing GSIS for islets of reasonably similar size (<50% variance), normalization does not improve the representation of GSIS data. Normalization may be beneficial when substantial differences in islet mass are involved. In such situations, we suggest that using islet cross-sectional area is superior to other commonly used techniques for normalizing GSIS data.

Introduction

Islets of Langerhans contain beta, alpha and delta cells, with beta cells responsible for the secretion of the hormone insulin (Steiner et al. 2010, Da Silva Xavier 2018), a hormone that is critical for lowering blood glucose levels. Islets are dispersed widely throughout the pancreas, accounting for ~1–2% of total pancreatic volume. The structure and function of these organs is conserved between murine and human islets (Bonner-Weir et al. 2015). A common test of islet function is known as glucose-stimulated insulin secretion (GSIS). For this test, islets are stimulated with two glucose concentrations, one that is relatively low (typically 1–3 mM) and one that is high (typically 10–28 mM). These stimulations are either run in parallel or sequentially from low to high glucose. Supernatants are collected at the end of each stimulation period to measure insulin secretion. GSIS has been instrumental in furthering our understanding of physiological and pathophysiological function of islets and beta cells.

Many research groups present GSIS data that has been normalized, in an attempt to correct for any inherent variation from trial to trial. Differences in islet size, the number of beta cells per islet (Farhat et al. 2013), changes to experimental protocols, different experimenters and even the intrinsic differences in the ability of each islet to secrete insulin (Colella et al. 1985) could all contribute to variability in measured insulin secretion. One method commonly used for data normalization is to visually match the cross-sectional area and number of islets for experiments; there is a variation of this method, where islets are selected as a mixture of equal numbers of small, medium and large islets (Carter et al. 2009, Huang et al. 2011). This method does not require any additional processing of islets after the GSIS data are collected, so islets can be used for immunostaining or other experiments in which islet material is required. The other three methods to normalize GSIS data employ total protein, insulin content or DNA content (Lernmark 1974, Hopcroft et al. 1985) as normalization parameters. These methods require additional steps of extracting islet content and measuring the said parameters, followed by subsequent calculations to normalize GSIS data. An additional way to present GSIS data is stimulation index (SI). SI is a measurement of the ratio of insulin secreted in high glucose to insulin secreted in low glucose.

The most important function of normalization is to have a reliable correction factor to be able to compare data sets. Trial-to-trial changes in personnel conducting experiments, changes in mouse colonies over time, small changes in experimental protocol, and many other factors can contribute to variability. Having a convenient parameter to correct for these differences is thus valuable, but which method is best?

In this study, we systematically tested the effectiveness of each parameter to normalize GSIS data by comparing their correlation to islet number and relationship to GSIS and by comparing coefficients of variation between the normalized and raw GSIS data. Our intent was to identify the most accurate method of normalizing GSIS data, and we have found in the initial trial that the islet number scales vary tightly with all normalization parameters. However, when examined using a smaller range of islet areas as would normally be carried out in standard GSIS experiments, we found no relationship between any normalization parameter and GSIS. In addition, raw GSIS data (i.e. non-normalized data) had the lowest variation when compared to GSIS data that were normalized to islet area, protein or insulin content or presented as SI. In light of these observations, we suggest it is preferable to present raw GSIS data from islets of similar size and number. Under circumstances in which islet size and/or number are substantially different, our data suggest the easiest and most effective method is to photograph the islets and use cross-sectional area to normalize the data. Employing these suggestions eliminates additional processing of the islets and additional quantification steps, which in turn will save time and preserve raw islet material for other applications.

Materials and methods

Mouse islets isolation

CD-1 male mice between the ages of 2–3 months were used as a source of islets according to protocols approved by the Ohio University Animal Care and Use Committee. The islets were isolated using standard collagenase digestion as described previously (Carter et al. 2009). The average islet yield per animal was ~250–350. The islets were handpicked 2 h after isolation and kept in RPMI 1640 (Gibco) with the addition of 10% FBS and 1% penicillin/streptomycin. Islets were maintained in 95% humidity incubator with 5% CO2 for 24–48 h before they were selected for the experiments. No differences in insulin secretion were observed (P = 0.46) between islets incubated for 24 h (14.3 ± 6.2 ng/mL, N = 5) or 48 h (11.2 ± 4.9 ng/mL, N = 13).

Scaling experiments

For the scaling experiments, 10, 20, 30 and 40 islets were handpicked to match in size, with diameters ranging from 115 µm to 223 µm (about 60% of the isolated islets from CD-1 mice are in this range and they appear as medium sized compared to other islets). For each set of islets, total protein, total insulin and total DNA were measured as described in the 'Extraction of islet content' section below.

Glucose-stimulated insulin secretion

For this set of studies, islets were collected without regard for size. These randomly selected islets had diameters of 85–288 µm. A total of nine mice were used, from which we produced 18 biological replicates; 6 out of 18 replicates came from islets pooled from two mice, the rest was from individual animals, and there was no difference in insulin secretion between pooled vs individual islets, with 9.7 ± 2.9 ng/mL and 13.3 ± 5.9 ng/mL, respectively (P = 0.18). Insulin secretion was measured using modified Krebs Ringer Buffer (KRB), as reported previously (Carter et al. 2009, Dula et al. 2010). Briefly, 20 islets were statically incubated with KRB without glucose for 1 h, followed by basal glucose (3 mM) in KRB for 1 h, followed by stimulatory glucose (11 mM) in KRB and supernatants were collected for basal and stimulatory glucose, spun and stored at −20°C. The insulin concentration was assayed by ultrasensitive mouse insulin ELISA kit (ALPCO, Salem, NH, USA) used as directed by the manufacturer. Three ELISA plates were used to analyze the entire data set with the average coefficient of variation (CV) for plate 1 being 6.8%, plate 2 being 8.3% and plate 3 being 5.6%. The average CV for all three plates was 6.9%. The inter-plate CV was 4.3%.

Extraction of islet content

After the media was collected for GSIS (see the 'Glucose-stimulated insulin secretion' section above), the islets were washed in PBS, PBS was removed and the islets were frozen in liquid nitrogen and stored at −80°C until further processing. To extract the content of the islets, ethanol–acid extraction was performed on ice. The extraction buffer composition was 70% ethanol and 1.5% hydrochloric acid. The buffer was made fresh before extraction and pre-chilled on ice. The islets were resuspended in 50 µL of ice-cold extraction buffer and vortexed for 30 s on the highest vortex setting. Resuspended islets were incubated on ice for 2 h with intermittent vortexing every 30 min. To ensure complete lysis of the islet tissue, sonication was performed with electric sonicator (Fisher Scientific), set to 20% power and sonication was performed for 1 s. The sonicated islet extracts were then spun at 380 g for 15 min at 4°C. The supernatant was recovered and kept on ice. The final islet extract was promptly used to perform the total protein assay and to measure the total insulin and DNA contents as described below.

Total protein content of islets

To measure the total protein of the islet extract (see ‘Extraction of islets content’ section), BCA kit was used (Fisher Scientific) as directed by the manufacturer.

Total insulin content of islets

To quantify the total insulin content of the islets, the islet extract was used to perform insulin ELISA using ALPCO ultrasensitive ELISA kit as directed by the manufacturer and as previously performed for total pancreatic insulin content (Antkowiak et al. 2013). The islet extract had to be diluted 1–1000 times.

Total DNA content of the islets

To quantify the total DNA content, a Qiagen Mini DNA preparation kit was used with the ethanol–acid islet extracts. Twenty microliters of the extract were used to purify the DNA, and the concentration was measured with NanoDrop spectrophotometer (Fisher Scientific).

Total islet area

To measure the cross-sectional area of the islets, digital imaging was performed, using EVOS XL Core imaging system (Invitrogen) with a 4× objective lens. Digital images were analyzed using ImageJ software (NIH) and the threshold function was used to quantify the area of the islets, after setting up the threshold of the program to include all the islets in the image. To quantify a cross-sectional area for individual islets, ImageJ was used. The perimeter of each islet was traced to produce a region of interest (ROI), from which the cross-sectional area was calculated for each individual islet. The diameter of each islet was calculated assuming that an islet is a perfect circle, which is a reasonably accurate first approximation for most islets. Individual islet areas were calculated for the data shown in Fig. 3 in order to show the typical distribution of islets sized in these studies. The threshold function within ImageJ software is more practical for routine measurement of islet area and takes less time and removes operator bias or error. We recommend the threshold method for measuring total islet area in each well.

Normalization of GSIS to the total islet area, total protein and insulin content

To calculate the GSIS normalization for each of the parameters the following formulas were used:

article image
article image
article image

Glucose-stimulated insulin secretion index

SI was calculated by dividing the insulin secreted during stimulatory glucose (11 mM glucose (11G)) by insulin secreted during basal incubation (3 mM glucose (3G)).

Statistical analysis

Prism7 (GraphPad) was used to generate and present the figures. Linear regression analysis was used to quantify the linear relationship between parameters, with P values identifying the slope being scientifically non-zero. Linear regression was used for islet scaling studies to determine R 2 values for each normalization method. Differences in R 2 values were identified by Mann–Whitney U test. A one-sample t-test was used to determine whether measured parameters scaled appropriately for a linear relationship of 2× (from 10 to 20 islets), 3× (from 10 to 30 islets) and 4× (from 10 to 40 islets).

Results

Correlation of islet number to normalization parameters

It is generally thought to be necessary to normalize GSIS in some manner. The most prominent normalization parameters are islet size or ‘size matching’, total insulin content, total protein content and total DNA. As a first step in comparing which of the accepted normalization parameters scale best to the number of islets, each parameter was plotted separately as a function of number of islets (Fig. 1). We chose to use 10, 20, 30 and 40 islets, to attain a linear relationship and to calculate the coefficient of determination (R 2) for each parameter, where values closest to 1.0 were considered as a better linear relationship.

Figure 1
Figure 1

Linear relationship of normalization parameters to islet number. (A, B and C) Linear relationship between the number of islets (10, 20, 30 and 40 size matched islets) and average total cross-sectional islet area measured in mm2 (A), average total protein measured in µg/mL (B) and average insulin content measured in µg/mL (C). R 2 values displayed in the upper left were calculated from average of the data from all five trials. (D) The scatter plot of R 2 values for each of the data sets for area, protein and insulin content. The data are presented as mean ± standard deviation (SD), each data set consists of five biological replicates. *P < 0.05, as determined by Mann–Whitney U.

Citation: Journal of Endocrinology 241, 2; 10.1530/JOE-18-0542

Total cross-sectional islet area had a strong linear relationship to islet number with R 2 of 0.9923 using data from five trials averaged together and reported as a single combined R 2 value (Fig. 1A). Total protein also had a strong relationship to the number of islets with R 2 of 0.9543 (Fig. 1B). Total islet insulin content had R 2 values of 0.9798 (Fig. 1C). When examined more closely, however, total protein and insulin content did not scale appropriately with the number of islets. The protein content from 40 islets should have been four times as much as for ten islets, but this was not the case (Fig. 1B). Using a one-sample t-test to compare measured and expected scaling factors, protein content fell short of expected values for 30 islets (2.2 ± 0.2 vs 3, P < 0.01) and for 40 islets (2.5 ± 0.5, vs 4, P < 0.01). Insulin content had a similar issue for 40 islets (2.9 ± 0.8 vs 4, P < 0.05). Islet area did not differ significantly from expected scaling factor for 20, 30 or 40 islets. It appears that protein and insulin content values do not scale as expected, which suggests these methods may not be ideally suited for normalizing islet data compared to a simple measurement of islet area.

R 2 values of each biological replicate also were plotted for area, protein and insulin content (Fig. 1D). Both Student t-test and Mann–Whitney U test showed that R 2 values for area (0.99 ± 0.01) were significantly higher than those for total protein (0.93 ± 0.04) or insulin content (0.90 ± 0.07) across the five trials (P < 0.05). Insulin content had the greatest variation between the biological replicates, whereas islet area had the least. These data suggest that islet area may have the greatest precision among normalization techniques, but all three techniques are reasonable.

DNA content was also measured by DNA purification kit; however, the raw materials were not sufficient to reliably quantify the DNA content. DNA yields were generally low or undetectable, resulting in generally poor correlation of determination values for each trial (average R 2 value of 0.50 ± 0.12 (data not shown)). Since we were able to get reliable measurements for every other parameter from the same islet source materials, we chose not to pursue DNA content for subsequent studies. It should be noted that with sufficient material, DNA content can accurately reflect islet mass (see Discussion).

Correlation of GSIS to normalization parameters across islet areas with large difference in mass

When the GSIS of a wide range of islet areas is compared to GSIS normalization parameters (Fig. 2), we have observed that area correlates to GSIS with R 2 of 0.6884 (Fig. 2A), total protein (Fig. 2B) and insulin content (Fig. 2C) correlate to GSIS with R 2 values of 0.4869 and 0.5765, respectively. These data suggest that when GSIS is collected from a wide range of islet masses, all normalization parameters have a linear correlation to GSIS, with area having the strongest correlation and protein content having the weakest correlation to GSIS (Fig. 2D).

Figure 2
Figure 2

Linear relationship of normalization parameters to GSIS from a large range of islets sizes. (A, B and C) Linear relationship between the GSIS and total cross-sectional islets area (mm2) (A), total protein in µg/mL (B) and average insulin content measured in µg/mL (C). R 2 values displayed in the upper left were calculated from average of the data from all five trials. (D) The scatter plot of R 2 values for each of the data sets for area, protein and insulin content.

Citation: Journal of Endocrinology 241, 2; 10.1530/JOE-18-0542

Heterogeneity of islet sizes

For the next step of validation, normalization parameters were measured from equal numbers of islets to measure variability in GSIS. A total of 18 biological replicates were used consisting of 20 islets each. Representative images of islets that were used for GSIS studies are shown in Fig. 3A and B. Some sets of islets had more uniformity in size (Fig. 3A), while others were less uniform in size (Fig. 3B). We quantified the coefficient of variation of islet areas for each group of 20 islets and found that some sets of islets had CV below or equal to 30% (6 out of 18 sets), which we called ‘low CV’. The remaining 12 out of 18 sets had CV equal or greater than 45%, which we called ‘high CV’ (Fig. 3C). The mean individual area of islets for these groups was very similar. The low CV group was 8729 ± 1675 μm2, and the high CV group was 8906 ± 1887 μm2, with a P value >0.9 (Fig. 3D). When we compared GSIS between low CV and high CV groups, we found a significant difference with low CV secreting 6.6 ± 2.0 ng/mL insulin and high CV secreting 14.8 ± 4.1 ng/mL insulin, with P < 0.001 (Fig. 3E). Relationship between GSIS and area CV was plotted and R 2 was found to be 0.4712, with P < 0.01 (Fig. 3F). These findings suggest that the distribution of small, medium and large islets may impact the relative amounts of secreted insulin.

Figure 3
Figure 3

Relationship of GSIS and coefficient of variation in individual islet area measurements. (A and B) Representative images of sets of 20 mouse islets with uniform islet area (A, low variation) and a large range of islet areas (B, high variation), where the scale bar is 100 µm. (C) Coefficient of variation (% CV) in islet area for each group of 20 islets used to measure GSIS based on individual islet area. Data are separated into a low CV group (<30%, n = 6) and a high CV group (>30%, n = 12, P < 0.001). (D) Mean individual islet area for high and low CV groups (P value >0.9). (E) GSIS in low and high CV groups (P value <0.001). (F) Relationship between CV of individual islet area and GSIS. R 2 represents measurement of data fitting to the linear regression line. P value <0.01 rejects the null hypothesis that the coefficient is equal to zero (no effect). A full colour version of this figure is available at https://doi.org/10.1530/JOE-18-0542.

Citation: Journal of Endocrinology 241, 2; 10.1530/JOE-18-0542

Mean GSIS values among replicates

Insulin secretion was measured with 1 h static incubation with 3 mM glucose (3G) for basal secretion and 11 mM glucose (11G) for stimulated secretion (Fig. 4A). The average basal insulin secreted by 20 islets within an hour was 0.99 ± 0.69 ng/mL. Islets secreted 12.09 ± 5.28 ng/mL with stimulating glucose levels. From the basal and stimulated insulin secretion, we calculated the SI, which is a ratio between stimulating and basal secreted insulin, and it was found to be 15.77 ± 9.77 (Fig. 4B).

Figure 4
Figure 4

GSIS from 20 islets. (A) Static glucose-stimulated insulin secretion (GSIS) from 20 islets in ng/mL; circles – 1-h incubation with 3 mM glucose (3G), followed by 1-h incubation with 11 mM glucose (11G – squares). The straight line is the mean and the bars are SD. (B) Glucose stimulation index (SI) (11G/3G), calculated from data in A. The straight line is the average and the bars are standard deviation. Each data point represents an average of 20 islets, total of 18 biological replicates. (C) Coefficients of variation (CV) for 11 mM glucose raw GSIS data (i.e. not normalized to any parameter), GSIS data normalized to islet area, total protein and insulin content and SI. The data represent average of all 18 biological replicates.

Citation: Journal of Endocrinology 241, 2; 10.1530/JOE-18-0542

Comparison of CVs for GSIS data normalized to islet area, protein and insulin content

The primary purpose of normalization is to decrease the variance in the samples. To further test the GSIS data normalization parameters, the CV was calculated for each of the 11 mM glucose GSIS data (Fig. 4C). For non-normalized (raw) data, the CV was found to be 43.7%, while GSIS data normalized to area was 46.8% and GSIS data normalized to protein and insulin was 59.2 and 45.5%, respectively. We also determined the CV for SI, which was found to be the largest at 61.9%. These data suggest that normalizing does not decrease the variability in the GSIS data, on the contrary, when data are normalized to protein content or presented as SI, the variability is appreciably increased compared to non-normalized GSIS data.

No correlations found between GSIS and islet area, protein or insulin content

We next examined the relationship between GSIS and the various normalization parameters at a more granular level. When GSIS data were plotted against islet cross-sectional area, there was no relationship found between the total islet area and the amount of insulin secreted by those islets, with R 2 of 0.0240 (Fig. 5A). The same was found for total protein, with R 2 of 0.0177 (Fig. 5B) and insulin content with R 2 of 0.0340 (Fig. 5C). The slopes for all three parameters were found to be non-statistically significant from zero, with P values of 0.54 for area, 0.60 for protein and 0.46 for insulin content. These data suggest that the tested normalization parameters do not correlate closely with GSIS under these conditions.

Figure 5
Figure 5

Correlation between normalization parameters and GSIS (11 mM glucose). (A, B and C) Relationship between GSIS and total cross-sectional islet area (A), total protein content of the islets (B) and total insulin content of the islets (C). Each data point represents an average of 20 islets, total of 18 biological replicates. R 2 represents measurement of data fitting to the linear regression line. P values >0.46 for (A–C) fail to reject the null hypothesis that the coefficient is equal to zero (no effect).

Citation: Journal of Endocrinology 241, 2; 10.1530/JOE-18-0542

Correlation of islet area to total protein content, total insulin content and correlation between protein and total insulin

Lastly, we examined whether there were any correlations among the normalization parameters themselves. It was found that total protein correlated with islet area (R 2 = 0.59), with a slope significantly non-zero and a P value of 0.0002 (Fig. 6A). This relationship makes sense since islet area is an estimate of islet mass, which should be closely linked with total protein. On the other hand, the insulin content had a weaker correlation with islet area, with an R 2 of 0.21 (Fig. 6B) and P value bordering on significance at 0.053. When insulin content and protein content were compared, there was no significant relationship between these two parameters, with R 2 of 0.18 and P value of 0.083 (Fig. 6C). These weaker relationships likely reflect the heterogeneity of insulin content values within islets, as has been reported (Huang et al. 2011, Farhat et al. 2013, Brereton et al. 2016).

Figure 6
Figure 6

Relationship of islet cross-sectional area to GSIS data normalization parameters. (A) A scatter plot of cross-sectional islet area to total protein. (B) A scatter plot of cross-sectional islet area to total islets insulin content. (C) A scatter plot of islet insulin content to islet total protein content. Each data point represents an average of 20 islets from an individual biological replicate, with total of 18 biological replicates. R 2 represents measurement of data fitting to the regression line, with a P value <0.05 rejecting the null hypothesis that the coefficient is equal to zero (no effect).

Citation: Journal of Endocrinology 241, 2; 10.1530/JOE-18-0542

Discussion

With GSIS data commonly used for the assessment of islet function, it is important to consider how these data are analyzed and presented for proper comparison and reproducibility between research groups. In an attempt to correct for the inherent heterogeneity and variability of GSIS data, many groups normalize GSIS data before presenting it. However, there is no consensus in the literature on which method is most appropriate. In this study, we found, somewhat surprisingly, that none of the commonly used normalization parameters improves the reporting of insulin secretion data.

Correlation of islet number and GSIS to insulin content, total protein and islet cross-sectional area

To evaluate the normalization parameters, we performed scaling experiments with linearly increasing number of islets (10, 20, 30 and 40). To quantify how close each parameter correlates to the islet number, we calculated the coefficient of determination (R 2) and compared it between the parameters, where the best parameter would have R 2 values closest to 1. The linear relationship of islet area to the different normalization parameters has been suggested for rat islets (Jahr et al. 1978), and correlation to islet area and secretory capacity was reported as well in perfused rat pancreas (Bonnevie-Nielsen et al. 1983). As we expected, all three normalization parameters had R 2 values close to 0.9 or greater, indicating that each parameter can adequately detect large differences in islet mass. However, when R 2 values for each parameter were compared to one another, islet area was superior to protein content and insulin content. Additionally, when we evaluated the secretory capacity of this wide range of islets masses, as expected, there was a linear correlation between all of the parameters and GSIS, again with area having the strongest correlation.

Interestingly, we observed that scaling factor for total protein was significantly different from expected for 30 and 40 islets, and insulin content for 40 islets also did not scale as expected. This observation suggests that when protein or insulin content are used to normalize GSIS data, larger islet masses would be normalized to reduced insulin amounts relative to smaller islet masses. Even islet area values for 40 islets fell slightly below the trendline, but this effect was small, and the scaling factor was not significantly different from the expected value. Although it is difficult to find an explanation for these observations, the relative reduction in protein and insulin content for 30 and 40 islets was fairly consistent and reproducible. It is possible that because the extraction method used to lyse the islets was optimized for 20 islets, this method was not efficient for larger number of islets and underestimated the amount of total protein and insulin content due to inefficient islet lysis. In any case, this failure of expected linear scaling weakens the use of total protein and insulin content as normalization parameter.

Relationship between GSIS data and each normalization parameter

Normalization methods assume a positive linear relationship between GSIS and the normalization parameter. We confirmed previously reported correlations with the wide ranges of islet mass; however, GSIS data are routinely collected from islets of similar mass (equal number and size of islets). The assumption that there is a linear correlation between GSIS and normalization parameters using a narrow range of islet masses has not been tested rigorously. We thus compared GSIS data from 18 biological replicates of 20 islets per replicate with cross-sectional islet area, total protein and insulin content to contrast it with the previously reported data of 10, 20, 30 and 40 islets. Interestingly, among these 18 sets of islets, we observed no correlation between GSIS data and any of the tested parameters, suggesting that the assumption that there is a relationship between GSIS and islet size, total protein or insulin content is not correct when the range of islet size is relatively small. This indicates that these parameters are not sufficient to normalize for the heterogeneity of GSIS data, when the same number of islets are used as is normally done for GSIS experiments (islets have to be within ~50% of target range).

We compared the CV among the GSIS normalization parameters as another test of their sufficiency to normalize GSIS data, with the premise that the most suitable parameter would produce GSIS data with the least variation. We designated raw GSIS data as a point of comparison, since hypothetically, these data would have the highest variation because they were not normalized to any parameter. We have found that GSIS data normalized to area and insulin content was very similar to raw GSIS data, suggesting that these parameters do not decrease the variation but also do not contribute to the variation. In contrast, GSIS data normalized to protein appear to add unnecessary variation. SI is also employed to present GSIS data, where the ratio of basal and stimulated insulin is presented. We found that GSIS data normalized to SI had the largest variation of all the parameters used. These results suggest that additional normalization of GSIS data does not improve the representation of these data but on the contrary increases variability and may not only be unnecessary but also will be encumbering to the interpretation of GSIS data.

Possible sources of variability in insulin secretion

We made attempts to identify some possible sources of variability in insulin secretion due to differences in protocol. For example, some GSIS trials were conducted within 24 h after islet isolation vs 48 h, but we found no significant differences in insulin secretion due to duration in culture between 24–48 h. One factor that does appear to impact variability is whether islets come from a single mouse or are pooled from more than one mouse. We found that the standard deviation of the mean for GSIS among pooled islets was less than half that of islets from a single mouse source. This is not surprising because our prior work showed that the patterning of calcium oscillations that drives pulsatile insulin secretion differed markedly when comparing islets isolated from one mouse to another mouse (Nunemaker et al. 2009, 2005). This is also consistent with a report suggesting that pooling islets from multiple mice leads to much more consistent gene expression results through RT-PCR (Layden et al. 2010). Pooling islets from multiple mice is one way to reduce variability.

Closely size matching islets may be another means of reducing variability. We observed that the sets of islets with the highest CV for islet area also had the highest rates of insulin secretion. The CV values indicate that there were more small plus large islets in the high CV group and a large number of ‘medium’ in the low CV group. Although the total islet area values for the high CV and low CV groups were quite similar, it appears that smaller and/or larger islets secrete more insulin than a uniform collection of medium-sized islets. However, the low CV group also produced much more tightly distributed GSIS values compared to the high CV group, suggesting that closely area-matching islets can reduce GSIS variability.

It should be noted that there are many other potential causes of variability among islets that could impact insulin secretion but could not be easily addressed experimentally. Recent studies have identified numerous subpopulations of different beta cells with diverse gene expression, sensitivity to glucose, different metabolic activities and secretory outputs (Roscioni et al. 2016, Benninger & Hodson 2018, Nasteska & Hodson 2018). Moreover, there is an additional degree of plasticity among beta-cell subpopulations under conditions of metabolic stress, such as obesity and pregnancy. In pathological conditions such as type 2 diabetes, this plasticity can lead to beta-cell ‘dedifferentiation’, resulting in a loss of insulin secretion (Brereton et al. 2016). Also, the ability of islet cells to communicate with one another impacts islet function and insulin secretion (Benninger et al. 2014, Benninger & Hodson 2018). All these biological changes/differences contribute to insulin secretory capacity and may contribute to the variability and heterogeneity of GSIS data.

Correlations and non-correlations among normalization parameters

Among normalization parameters, the strongest linear correlation was observed between islet area and total protein content. This observation suggests that the cross-sectional area of islets is a reasonable predictor of total protein and vice versa, which makes sense since each technique provides an estimate of islet mass. In contrast, insulin content was found to be loosely related to islet area, but this relationship was marginally significant (P = 0.053). One study reported that smaller human islets had higher insulin content due to a proportionally higher beta-cell number (Farhat et al. 2013). The lack of a significant correlation between islet area and insulin content may be due to each set of islets having different distribution of small vs large islets, which may lead to higher variability in insulin content between these sets. Further, the relative number of beta cells to other islet cell types differs from islet to islet, even when similarly sized. This may also impact the relationship between insulin content and islet area or total protein content. The key conclusion of the linear correlation studies presented in Fig. 5 is that when reasonable attempts are made to keep the islet mass within ±50% of target size, the amount of insulin secretion from any given sample has very little to do with the islet area, protein content or insulin content of the sample.

Normalization of GSIS data to islet area is the best method

We have shown that GSIS data do not require normalization in general; however, in some cases, normalization may still be needed. For example, the islet areas in one experiment may be more than 50% within target range of the next preparation done on a different mouse strain or by different personnel, resulting in a large mismatch in islet mass between groups. In this case, islet area should be used to normalize the GSIS data, by taking pictures of the islet preparations after the GSIS experiments are set up. From these digital representations of the islet mass, islet area can be calculated as described in ‘Methods’ section, and GSIS data can be normalized to this parameter. We recommend this method of normalizing because we report that area has the most accurate correlation with islet number and greatest association with insulin secretion. Additionally, digital images of islet experiments can be stored electronically in perpetuity for use years later, whereas biological samples (proteins and DNA) are not quite so convenient to store or examine at a much later date.

A note on DNA content as a GSIS normalization parameter

Islet extracts for these studies were used to measure total protein content, insulin content and DNA content from the same sets of islets. These extract amounts were sufficient to measure insulin and protein; however, this raw material was not sufficient to reliably quantify total DNA content with the spectrophotometer method we have employed (see ‘Materials and Methods’ section) despite using more islet extract for DNA than for insulin or protein content measurements. A key element of this study was to use the same islets as a source material for all experimental measurements to make a direct comparison. In that sense, DNA content failed a key test in being able to produce a reliable relationship with the same amount of islet materials used for the other techniques. This is not to say that laboratories utilizing this technique are not capable of producing viable results. Refinement of our methods, by using fluorescent probe SYBR Green I dye (Vitzthum et al. 1999, Leggate et al. 2006) or increasing the number of islets, may improve the yield of DNA to more accurately quantify content. However, while prior work has shown that DNA correlates with glucose utilization (Colella et al. 1985), islet number (Lernmark 1974) and insulin secretion (Reaven et al. 1981, Hopcroft et al. 1985), the correlation of determination values in these studies were not any better than what we report for islet area, protein content or insulin content. Also, as we showed for every other normalization parameter, correlations fail when the range of islet masses is small. None of the DNA-based studies listed above appear to pass this ‘small range’ test. It is thus unlikely that DNA content is as good, or better, than any other normalization parameter that we tested in this study.

Conclusions

Considering the data reported in this study, we suggest that making reasonable attempts to area-match islets (±50% of target) is a sufficient, economical and time-saving method to collect and present GSIS data, with the added benefit of conserving raw islet material for further testing by western blotting, immunohistochemistry, quantitative PCR and other desired downstream uses where the raw islet material is required. Conversely, the tested parameters are valid measurements of other aspects of the islet’s physiological status, separate from GSIS. For example, the ratio of insulin content to total protein can be used to identify the differences in relative insulin content between experimental treatments or between mouse strains or to estimate the fraction of beta cells (measuring insulin content) to total islet cells (measuring total protein). There appears to be no benefit, however, to using any of these measurements to improve the reporting of glucose-stimulated insulin secretion from pancreatic islets.

Declaration of interest

The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported.

Funding

This work was supported by the National Institute of Diabetes and Digestive and Kidney Diseases Mouse Metabolic Phenotyping Centers (www.mmpc.org) under the MICROMouse Funding Program Grant DK076169, R01 DK089182 to C S N, the Osteopathic Heritage Foundation and the Heritage College of Osteopathic Medicine at Ohio University.

Author contribution statement

K G S performed the experiments, helped with experimental set up, researched data, generated figures, wrote the manuscript and contributed to the references. K L C helped with experiential planning and execution, contributed to discussion and reviewed the manuscript. C S N conceived the study, supervised the experiments and data analysis, secured the funding, reviewed the manuscript and contributed to references.

Acknowledgements

The authors thank Dr Richard Benninger for critically evaluating the manuscript and providing helpful comments. Dr Kelly McCall for generously providing equipment to image the islets (EVOS XL Core).

References

  • Antkowiak PF, Stevens BK, Nunemaker CS, McDuffie M & Epstein FH 2013 Manganese-enhanced magnetic resonance imaging detects declining pancreatic β-cell mass in a cyclophosphamide-accelerated mouse model of type 1 diabetes. Diabetes 62 4448. (https://doi.org/10.2337/db12-0153)

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Benninger RKP & Hodson DJ 2018 New understanding of β-cell heterogeneity and in situ islet function. Diabetes 67 537547. (https://doi.org/10.2337/dbi17-0040)

  • Benninger RP, Hutchens T, Head WS, McCaughey MJ, Zhang M, Le Marchand SJ, Satin LS & Piston DW 2014 Intrinsic islet heterogeneity and gap junction coupling determine spatiotemporal Ca2+ wave dynamics. Biophysical Journal 107 27232733. (https://doi.org/10.1016/j.bpj.2014.10.048)

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bonner-Weir S, Sullivan BA & Weir GC 2015 Human islet morphology revisited. Journal of Histochemistry and Cytochemistry 63 604612. (https://doi.org/10.1369/0022155415570969)

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bonnevie-Nielsen V, Skovgaard LT & Lernmark Å 1983 beta-Cell function relative to islet volume and hormone content in the isolated perfused mouse pancreas. Endocrinology 112 10491056. (https://doi.org/10.1210/endo-112-3-1049)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Brereton MF, Rohm M & Ashcroft FM 2016 β-Cell dysfunction in diabetes: a crisis of identity? Diabetes, Obesity and Metabolism 18 (Supplement 1) 102109. (https://doi.org/10.1111/dom.12732)

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Carter JD, Dula SB, Corbin KL, Wu R & Nunemaker CS 2009 A practical guide to rodent islet isolation and assessment. Biological Procedures Online 11 331. (https://doi.org/10.1007/s12575-009-9021-0)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Colella RM, Bonner-Weir S, Braunstein LP, Schwalke M & Weir GC 1985 Pancreatic islets of variable size – insulin secretion and glucose utilization. Life Sciences 37 10591065. (https://doi.org/10.1016/0024-3205(85)90597-1)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Da Silva Xavier G 2018 The cells of the islets of Langerhans. Journal of Clinical Medicine 7 E54. (https://doi.org/10.3390/jcm7030054)

  • Dula SB, Jecmenica M, Wu R, Jahanshahi P, Verrilli GM, Carter JD, Brayman KL & Nunemaker CS 2010 Evidence that low-grade systemic inflammation can induce islet dysfunction as measured by impaired calcium handling. Cell Calcium 48 133142. (https://doi.org/10.1016/j.ceca.2010.07.007)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Farhat B, Almelkar A, Ramachandran K, Williams SJ, Huang H-H, Zamierowksi D, Novikova L & Stehno-Bittel L 2013 Small human islets comprised of more β-cells with higher insulin content than large islets. Islets 5 8794. (https://doi.org/10.4161/isl.24780)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Hopcroft DW, Mason DR & Scott RS 1985 Standardization of insulin secretion from pancreatic islets: validation of a DNA assay. Hormone and Metabolic Research 17 559561. (https://doi.org/10.1055/s-2007-1013606)

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Huang H-H, Novikova L, Williams SJ, Smirnova IV & Stehno-Bittel L 2011 Low insulin content of large islet population is present in situ and in isolated islets. Islets 3 613. (https://doi.org/10.4161/isl.3.1.14132)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Jahr H, Gottschling D & Zühlke H 1978 Correlation of islet size and biochemical parameters of isolated islets of Langerhans of rats. Acta Biologica and Medica Germanica 37 659662.

    • Search Google Scholar
    • Export Citation
  • Layden BT, Durai V, Newman MV, Marinelarena AM, Ahn CW, Feng G, Lin S, Zhang X, Kaufman DB, Jafari N, et al.2010 Regulation of pancreatic islet gene expression in mouse islets by pregnancy. Journal of Endocrinology 207 265279. (https://doi.org/10.1677/JOE-10-0298)

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Leggate J, Allain R, Isaac L & Blais BW 2006 Microplate fluorescence assay for the quantification of double stranded DNA using SYBR Green I dye. Biotechnology Letters 28 15871594. (https://doi.org/10.1007/s10529-006-9128-1)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Lernmark A 1974 The preparation of, and studies on, free cell suspensions from mouse pancreatic islets. Diabetologia 10 431438. (https://doi.org/10.1007/BF01221634)

  • Nasteska D & Hodson DJ 2018 The role of beta cell heterogeneity in islet function and insulin release. Journal of Molecular Endocrinology 61 R43R60. (https://doi.org/10.1530/JME-18-0011)

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Nunemaker CS, Dishinger JF, Dula SB, Wu R, Merrins MJ, Reid KR, Sherman A, Kennedy RT & Satin LS 2009 Glucose metabolism, islet architecture, and genetic homogeneity in imprinting of [Ca2+](i). PLoS ONE 4 e8428. (https://doi.org/10.1371/journal.pone.0008428)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Nunemaker CS, Zhang M, Wasserman DH, McGuinness OP, Powers AC, Bertram R, Sherman A & Satin LS 2005 Individual mice can be distinguished by the period of their islet calcium oscillations: is there an intrinsic islet period that is imprinted in vivo? Diabetes 54 35173522. (https://doi.org/10.2337/diabetes.54.12.3517)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Reaven EP, Gold G, Walker W & Reaven GM 1981 Effect of variations in islet size and shape on glucose-stimulated insulin secretion. Hormone and Metabolic Research 13 673674. (https://doi.org/10.1055/s-2007-1019372)

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Roscioni SS, Migliorini A, Gegg M & Lickert H 2016 Impact of islet architecture on β-cell heterogeneity, plasticity and function. Nature Reviews Endocrinology 12 695709. (https://doi.org/10.1038/nrendo.2016.147)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Steiner DJ, Kim A, Miller K & Hara M 2010 Pancreatic islet plasticity: interspecies comparison of islet architecture and composition. Islets 2 135145. (https://doi.org/10.4161/isl.2.3.11815)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Vitzthum F, Geiger G, Bisswanger H, Brunner H & Bernhagen J 1999 A quantitative fluorescence-based microplate assay for the determination of double-stranded DNA using SYBR Green I and a standard ultraviolet transilluminator gel imaging system. Analytical Biochemistry 276 5964. (https://doi.org/10.1006/abio.1999.4298)

    • Crossref
    • Search Google Scholar
    • Export Citation

 

Society for Endocrinology logo

Sept 2018 onwards Past Year Past 30 Days
Abstract Views 1093 0 0
Full Text Views 899 364 26
PDF Downloads 705 359 26
  • View in gallery

    Linear relationship of normalization parameters to islet number. (A, B and C) Linear relationship between the number of islets (10, 20, 30 and 40 size matched islets) and average total cross-sectional islet area measured in mm2 (A), average total protein measured in µg/mL (B) and average insulin content measured in µg/mL (C). R 2 values displayed in the upper left were calculated from average of the data from all five trials. (D) The scatter plot of R 2 values for each of the data sets for area, protein and insulin content. The data are presented as mean ± standard deviation (SD), each data set consists of five biological replicates. *P < 0.05, as determined by Mann–Whitney U.

  • View in gallery

    Linear relationship of normalization parameters to GSIS from a large range of islets sizes. (A, B and C) Linear relationship between the GSIS and total cross-sectional islets area (mm2) (A), total protein in µg/mL (B) and average insulin content measured in µg/mL (C). R 2 values displayed in the upper left were calculated from average of the data from all five trials. (D) The scatter plot of R 2 values for each of the data sets for area, protein and insulin content.

  • View in gallery

    Relationship of GSIS and coefficient of variation in individual islet area measurements. (A and B) Representative images of sets of 20 mouse islets with uniform islet area (A, low variation) and a large range of islet areas (B, high variation), where the scale bar is 100 µm. (C) Coefficient of variation (% CV) in islet area for each group of 20 islets used to measure GSIS based on individual islet area. Data are separated into a low CV group (<30%, n = 6) and a high CV group (>30%, n = 12, P < 0.001). (D) Mean individual islet area for high and low CV groups (P value >0.9). (E) GSIS in low and high CV groups (P value <0.001). (F) Relationship between CV of individual islet area and GSIS. R 2 represents measurement of data fitting to the linear regression line. P value <0.01 rejects the null hypothesis that the coefficient is equal to zero (no effect). A full colour version of this figure is available at https://doi.org/10.1530/JOE-18-0542.

  • View in gallery

    GSIS from 20 islets. (A) Static glucose-stimulated insulin secretion (GSIS) from 20 islets in ng/mL; circles – 1-h incubation with 3 mM glucose (3G), followed by 1-h incubation with 11 mM glucose (11G – squares). The straight line is the mean and the bars are SD. (B) Glucose stimulation index (SI) (11G/3G), calculated from data in A. The straight line is the average and the bars are standard deviation. Each data point represents an average of 20 islets, total of 18 biological replicates. (C) Coefficients of variation (CV) for 11 mM glucose raw GSIS data (i.e. not normalized to any parameter), GSIS data normalized to islet area, total protein and insulin content and SI. The data represent average of all 18 biological replicates.

  • View in gallery

    Correlation between normalization parameters and GSIS (11 mM glucose). (A, B and C) Relationship between GSIS and total cross-sectional islet area (A), total protein content of the islets (B) and total insulin content of the islets (C). Each data point represents an average of 20 islets, total of 18 biological replicates. R 2 represents measurement of data fitting to the linear regression line. P values >0.46 for (A–C) fail to reject the null hypothesis that the coefficient is equal to zero (no effect).

  • View in gallery

    Relationship of islet cross-sectional area to GSIS data normalization parameters. (A) A scatter plot of cross-sectional islet area to total protein. (B) A scatter plot of cross-sectional islet area to total islets insulin content. (C) A scatter plot of islet insulin content to islet total protein content. Each data point represents an average of 20 islets from an individual biological replicate, with total of 18 biological replicates. R 2 represents measurement of data fitting to the regression line, with a P value <0.05 rejecting the null hypothesis that the coefficient is equal to zero (no effect).

  • Antkowiak PF, Stevens BK, Nunemaker CS, McDuffie M & Epstein FH 2013 Manganese-enhanced magnetic resonance imaging detects declining pancreatic β-cell mass in a cyclophosphamide-accelerated mouse model of type 1 diabetes. Diabetes 62 4448. (https://doi.org/10.2337/db12-0153)

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Benninger RKP & Hodson DJ 2018 New understanding of β-cell heterogeneity and in situ islet function. Diabetes 67 537547. (https://doi.org/10.2337/dbi17-0040)

  • Benninger RP, Hutchens T, Head WS, McCaughey MJ, Zhang M, Le Marchand SJ, Satin LS & Piston DW 2014 Intrinsic islet heterogeneity and gap junction coupling determine spatiotemporal Ca2+ wave dynamics. Biophysical Journal 107 27232733. (https://doi.org/10.1016/j.bpj.2014.10.048)

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bonner-Weir S, Sullivan BA & Weir GC 2015 Human islet morphology revisited. Journal of Histochemistry and Cytochemistry 63 604612. (https://doi.org/10.1369/0022155415570969)

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bonnevie-Nielsen V, Skovgaard LT & Lernmark Å 1983 beta-Cell function relative to islet volume and hormone content in the isolated perfused mouse pancreas. Endocrinology 112 10491056. (https://doi.org/10.1210/endo-112-3-1049)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Brereton MF, Rohm M & Ashcroft FM 2016 β-Cell dysfunction in diabetes: a crisis of identity? Diabetes, Obesity and Metabolism 18 (Supplement 1) 102109. (https://doi.org/10.1111/dom.12732)

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Carter JD, Dula SB, Corbin KL, Wu R & Nunemaker CS 2009 A practical guide to rodent islet isolation and assessment. Biological Procedures Online 11 331. (https://doi.org/10.1007/s12575-009-9021-0)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Colella RM, Bonner-Weir S, Braunstein LP, Schwalke M & Weir GC 1985 Pancreatic islets of variable size – insulin secretion and glucose utilization. Life Sciences 37 10591065. (https://doi.org/10.1016/0024-3205(85)90597-1)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Da Silva Xavier G 2018 The cells of the islets of Langerhans. Journal of Clinical Medicine 7 E54. (https://doi.org/10.3390/jcm7030054)

  • Dula SB, Jecmenica M, Wu R, Jahanshahi P, Verrilli GM, Carter JD, Brayman KL & Nunemaker CS 2010 Evidence that low-grade systemic inflammation can induce islet dysfunction as measured by impaired calcium handling. Cell Calcium 48 133142. (https://doi.org/10.1016/j.ceca.2010.07.007)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Farhat B, Almelkar A, Ramachandran K, Williams SJ, Huang H-H, Zamierowksi D, Novikova L & Stehno-Bittel L 2013 Small human islets comprised of more β-cells with higher insulin content than large islets. Islets 5 8794. (https://doi.org/10.4161/isl.24780)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Hopcroft DW, Mason DR & Scott RS 1985 Standardization of insulin secretion from pancreatic islets: validation of a DNA assay. Hormone and Metabolic Research 17 559561. (https://doi.org/10.1055/s-2007-1013606)

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Huang H-H, Novikova L, Williams SJ, Smirnova IV & Stehno-Bittel L 2011 Low insulin content of large islet population is present in situ and in isolated islets. Islets 3 613. (https://doi.org/10.4161/isl.3.1.14132)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Jahr H, Gottschling D & Zühlke H 1978 Correlation of islet size and biochemical parameters of isolated islets of Langerhans of rats. Acta Biologica and Medica Germanica 37 659662.

    • Search Google Scholar
    • Export Citation
  • Layden BT, Durai V, Newman MV, Marinelarena AM, Ahn CW, Feng G, Lin S, Zhang X, Kaufman DB, Jafari N, et al.2010 Regulation of pancreatic islet gene expression in mouse islets by pregnancy. Journal of Endocrinology 207 265279. (https://doi.org/10.1677/JOE-10-0298)

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Leggate J, Allain R, Isaac L & Blais BW 2006 Microplate fluorescence assay for the quantification of double stranded DNA using SYBR Green I dye. Biotechnology Letters 28 15871594. (https://doi.org/10.1007/s10529-006-9128-1)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Lernmark A 1974 The preparation of, and studies on, free cell suspensions from mouse pancreatic islets. Diabetologia 10 431438. (https://doi.org/10.1007/BF01221634)

  • Nasteska D & Hodson DJ 2018 The role of beta cell heterogeneity in islet function and insulin release. Journal of Molecular Endocrinology 61 R43R60. (https://doi.org/10.1530/JME-18-0011)

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Nunemaker CS, Dishinger JF, Dula SB, Wu R, Merrins MJ, Reid KR, Sherman A, Kennedy RT & Satin LS 2009 Glucose metabolism, islet architecture, and genetic homogeneity in imprinting of [Ca2+](i). PLoS ONE 4 e8428. (https://doi.org/10.1371/journal.pone.0008428)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Nunemaker CS, Zhang M, Wasserman DH, McGuinness OP, Powers AC, Bertram R, Sherman A & Satin LS 2005 Individual mice can be distinguished by the period of their islet calcium oscillations: is there an intrinsic islet period that is imprinted in vivo? Diabetes 54 35173522. (https://doi.org/10.2337/diabetes.54.12.3517)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Reaven EP, Gold G, Walker W & Reaven GM 1981 Effect of variations in islet size and shape on glucose-stimulated insulin secretion. Hormone and Metabolic Research 13 673674. (https://doi.org/10.1055/s-2007-1019372)

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Roscioni SS, Migliorini A, Gegg M & Lickert H 2016 Impact of islet architecture on β-cell heterogeneity, plasticity and function. Nature Reviews Endocrinology 12 695709. (https://doi.org/10.1038/nrendo.2016.147)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Steiner DJ, Kim A, Miller K & Hara M 2010 Pancreatic islet plasticity: interspecies comparison of islet architecture and composition. Islets 2 135145. (https://doi.org/10.4161/isl.2.3.11815)

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Vitzthum F, Geiger G, Bisswanger H, Brunner H & Bernhagen J 1999 A quantitative fluorescence-based microplate assay for the determination of double-stranded DNA using SYBR Green I and a standard ultraviolet transilluminator gel imaging system. Analytical Biochemistry 276 5964. (https://doi.org/10.1006/abio.1999.4298)

    • Crossref
    • Search Google Scholar
    • Export Citation