Loss of β-Cell Identity Occurs in Type 2 Diabetes and Is Associated With Islet Amyloid Deposits

Loss of pancreatic β-cell mass and β-cell dysfunction are central in the development of type 2 diabetes (T2DM) and, in combination with peripheral insulin resistance, lead to hyperglycemia (1). Whereas β-cells, on the one hand, fail to properly secrete insulin at a given glucose level, there is also a progressive decline in the number of β-cells (2,3). Loss of β-cell mass has been ascribed to increased apoptosis in T2DM (4). In patients with T2DM, β-cell mass can be up to 40–60% lower than in healthy control subjects (4–6). In addition, abnormal function of glucagon-producing α-cells leading to hyperglucagonemia is associated with T2DM (7). β-cell dedifferentiation and subsequent transition to other islet cell types were suggested as an alternative explanation for the loss of functional β-cell mass in mice (8,9). In this concept, β-cells lose insulin content and insulin secretory capacity followed by the production of other endocrine hormones such as glucagon (8). We recently showed (10) that loss of β-cell identity with the conversion of β-cells into glucagon-containing α-cells can occur in human pancreatic islets ex vivo.

A number of transcription factors have been identified to be essential for the development and maintenance of functional β-cells (11,12). Recent reports (13,14) indicate that a selective loss of transcription factors MafA, Nkx6.1, and Pdx1 is associated with β-cell dysfunction and T2DM. Chronic hyperglycemia in rats is accompanied by the loss of β-cell transcription factors (15). Moreover, mouse β-cells that genetically lack FOXO1 can dedifferentiate in vivo under conditions of metabolic stress and subsequently can convert (or transdifferentiate) into glucagon-producing α-cells, accompanied by hyperglucagonemia (8). In our ex vivo culture system, lineage tracing showed that mature human β-cells can lose their identity and convert into α-cells following dispersion and reaggregation (10). Conversion was characterized by a transition phase in which β-cell–specific transcription factors (Nkx6.1, Pdx1) were expressed in glucagon⁺ cells, a phenomenon also observed in the mice that genetically lack FOXO1 in β-cells (8,10). It is not clear whether the loss of β-cell identity occurs in humans in vivo and whether this could contribute to a reduction in functional β-cell mass. We hypothesized that the loss of β-cell identity (where β-cell identity is defined by the presence of specific transcription factors and insulin), is involved in the loss of functional β-cell mass that occurs in T2DM. Therefore, we studied pancreata from humans and nonhuman primates to determine the relationships among metabolic status, pathological changes in islet morphology, and loss of β-cell identity.

Human cadaveric donor pancreata were procured through a multiorgan donor program. Pancreatic tissue was used in our study if the pancreas could not be used for clinical pancreas or islet transplantation, according to national laws, and if research consent was present. Tissue from organ donors in whom T2DM had previously been diagnosed (n = 11) were compared with donors who had no history of diabetes (i.e., nondiabetic [ND] donors; n = 9). Control donors were matched for age and BMI. More detailed information on all donors is summarized in Supplementary Table 1. Pancreatic tail samples were fixed overnight in 4% formaldehyde (Klinipath), stored in 70% ethanol, and subsequently embedded in paraffin. Blocks were cut into 4-µm-thick sections.

Fourteen male nonhuman primates (12 Macaca mulatta and 2 Macaca fascicularis) were selected from a larger colony that has been characterized extensively (16). The nonhuman primates were maintained in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals, and all protocols have been reviewed and approved by the university animal care and use committee. The animals were maintained on standard monkey chow (15% protein, 26% fat, and 59% carbohydrate; LabDiet, St. Louis, MO) that was provided ad libitum. Animals were divided into the following three groups, based on the metabolic profile at sacrifice: normoglycemic and normoinsulinemic; normoglycemic and hyperinsulinemic (HI) (fasting immunoreactive insulin >600 pmol/L); or overtly diabetic (DM) (fasting plasma glucose >7.8 mmol/L). The baseline characteristics (including β-cell and amyloid area) (17) for each group are provided in Table 1, and for each animal in Supplementary Table 2. Postmortem samples were obtained at sacrifice and processed for wax embedding. Some tissue blocks (in all three groups) showed poor quality of nuclear staining, including DAPI (likely due to the fixation procedure). For the studies on nonhuman primates, staining quality was considered sufficient when nuclear Nkx6.1 labeling could be identified. This decision was made prior to the analysis of >20 islets. Tail samples were used, except for one animal (animal Q in Supplementary Table 2), from which only a sample of the pancreas body was available.

The β-cell area and the ratio of α-cell area to β-cell area (α/β-ratio) were determined as described previously (18,19). Briefly, sections were immunostained with primary antibodies against insulin (1:200; Millipore or Santa Cruz Biotechnology) and glucagon (1:200; Vector Laboratories) for 1 h followed by horseradish peroxidase–conjugated or alkaline phosphatase–conjugated secondary antibodies for 1 h. Sections were developed with 3,3′-diaminobenzidine tetrahydrochloride or liquid permanent red (Dako), respectively; and counterstained with hematoxylin. Stained sections were completely imaged using a digital slide scanner (Panoramic MIDI; 3DHISTECH). β-cell and α-cell areas were determined using an image analysis program (Stacks version 2.1; Leiden University Medical Center), and were expressed as a percentage relative to the exocrine area after exclusion of large blood vessels and ducts, adipose tissue, and lymph nodes. A threshold of four cells in a cluster was used in order to be included in the analysis. Scattered single cells or duct-associated single cells were not included.

Following rehydration of tissue sections, antigen retrieval was performed by heating slides in citrate buffer (pH 6.0) using a pressure cooker. Primary antibodies against insulin (1:200; Millipore or Santa Cruz Biotechnology), glucagon (1:200; Vector Laboratories or Invitrogen), Pdx1 (1:5; R&D Systems), Nkx6.1 (1:1,000; clone F55A12; Developmental Studies Hybridoma Bank; and 1:250; Sigma-Aldrich), Nkx6.2 (1:50; Santa Cruz Biotechnology), MafA (1:200; LP9872; BetaLogics Venture), Arx (1:1,000; R&D Systems), and FOXO1 (1:200; Cell Signaling Technology) were used. DAPI (Vector Laboratories) was used as nuclear counterstaining. The secondary antibodies used were tetramethyl rhodamine isothiocyanate-anti–guinea pig (1:400; The Jackson Laboratory) and Alexa Fluor 488, 568, and 647 anti-mouse or anti-rabbit, when appropriate (1:1,000; Molecular Probes). Sections were imaged using an LSM 7 MP confocal microscope (Zeiss), and images were collected for subsequent quantification. The investigators were blind to the donor background during quantification of the staining for Nkx6.1, glucagon, and insulin. Labeling was analyzed by counting cells in at least 25 islets per donor, containing >10 hormone⁺ cells per islet. Bihormonal cells in our study were defined as cells that were positive for both insulin and glucagon. The proportion of bihormonal cells is represented relative to the total number of insulin⁺ cells, whereas the proportion of Nkx6.1⁺glucagon⁺insulin⁻ cells is represented relative to the total number of Nkx6.1⁺ cells.

To analyze the presence of glucagon⁺/Nkx6.1⁺ cells in contact with amyloid deposits, amyloid was labeled by incubation in 0.5% Thioflavin S in PBS for 2 min after antigen retrieval, but before immunolabeling for the detection of glucagon and Nkx6.1. Amyloid area was determined from the Thioflavin S⁺ area in >20 islets for all donors. Islet areas (in square micrometers) and amyloid area were measured using ImageJ software (National Institutes of Health). The islet amyloid percentage was calculated from the amyloid area and the islet area.

Pancreas samples from two subjects with T2DM were processed for immunoelectron microscopic analysis. One tissue sample was embedded in LR Gold resin, as described previously (20). The second sample was fixed in 0.2% glutaraldehyde and 2% paraformaldehyde, and was prepared for cryo-immunogold labeling (21). Ultrathin sections from both samples were labeled in two consecutive steps. First, sections were incubated with monoclonal mouse anti–C-peptide (1:2,000 or 1:400 for cryosections or LR Gold, respectively; Millipore) followed by rabbit anti-mouse IgG (1:200). In the second round, polyclonal rabbit anti-glucagon (1:400 or 1:30 for cryosections or LR Gold, respectively; Vector Laboratories) was used. In both steps, antibody binding was identified using protein A gold (particle size 10 or 15 nm), while fixation in 1% glutaraldehyde was performed between both steps to avoid cross-reactivity. Images were made on a Tecnai 12 BioTwin transmission electron microscope (FEI) at 120 kV.

All data are expressed as the mean ± SEM, unless stated otherwise. Mann-Whitney U test was used for assessment of statistical significance of differences between two groups, and Kruskal-Wallis test followed by Dunn multiple comparisons test was used for more than two groups. Spearman rank correlation test was used to test correlations. P < 0.05 was considered to be statistically significant.

Age (61.2 ± 8.7 vs. 59.4 ± 8.7 years, ND vs. T2DM) and BMI (29.2 ± 4.3 vs. 30.0 ± 4.9 kg/m², ND vs. T2DM) were similar in both groups of organ donors (Supplementary Table 1). Immunohistochemical labeling for glucagon and insulin was performed to assess the relative areas of α- and β-cells, respectively (Supplementary Fig. 1A). The mean relative β-cell area was 25% lower in the T2DM subjects compared with the matched ND control subjects (2.04 ± 0.28 vs. 1.54 ± 0.37%, ND vs. T2DM, P < 0.05; Fig. 1A). The α/β-ratio was significantly higher in subjects with T2DM compared with ND control subjects (0.50 ± 0.19 vs. 0.98 ± 0.24, ND vs. T2DM, P < 0.05; Fig. 1B), while overall islet size did not significantly differ between the two groups (Supplementary Fig. 1B). In line with previously reported data (5), β-cell area was increased in the ND donors with the highest BMI (Supplementary Fig. 1C).

Confocal imaging of sequential optical sections (Z-stacks) through the 4-µm tissue section revealed the presence of insulin⁺ cells that also expressed glucagon (i.e., bihormonal cells) (Fig. 2A and Supplementary Fig. 2A). These cells were present at a markedly increased frequency in the T2DM group (0.52 ± 0.18% vs. 4.05 ± 1.37, ND vs. T2DM, P < 0.01; Fig. 2B) and affected 33% of the islets in subjects with T2DM versus 13% in ND control subjects. The presence of these cells was confirmed by flow cytometry of dispersed human isolated islets (Supplementary Fig. 2B). Approximately half of bihormonal cells were negative for the β-cell–specific nuclear transcription factor Nkx6.1 in both the ND and T2DM groups (Fig. 2A and B and Supplementary Table 3). Nkx6.1 expression was exclusively present in the nuclei in ND donors, but localization also occurred in the cytoplasm in subjects with a history of T2DM (Supplementary Fig. 3A).

No significant correlation was found between bihormonal cell proportion and the α/β-ratio or donor factors such as BMI or age (Supplementary Fig. 2C–F). To identify bihormonal cells at the level of single granules, we performed immunoelectron microscopy. Double-immunogold labeling showed the presence of distinct granules labeled for either glucagon or C-peptide in the same cell (Fig. 2C). Granule ultrastructure in bihormonal cells was similar to normal insulin or glucagon granules, showing homogenous and electron-dense glucagon granules compared with a more crystalline structure of insulin granules (Fig. 2C). Lipofuscin bodies could be detected in the bihormonal cells (Fig. 2C).

We observed that of all nuclear Nkx6.1⁺ cells that contained glucagon, 60% did not express insulin. We therefore quantified the presence of glucagon in cells containing nuclear Nkx6.1 that were insulin⁻ as another indication of β-cell identity change. The percentage of Nkx6.1⁺glucagon⁺insulin⁻ cells was 4.5-fold higher in subjects with T2DM compared with the ND control subjects (1.14 ± 0.36% vs. 5.11 ± 1.49%, ND vs. T2DM, P < 0.01; Fig. 3A and B). The percentage of Nkx6.1⁺glucagon⁺insulin⁻ cells significantly correlated with the α/β-ratio (Fig. 3C), the percentage of bihormonal cells (Fig. 3D), and the duration of T2DM, while no correlation was found with BMI or age (Supplementary Fig. 3B–D).

To further characterize bihormonal and Nkx6.1⁺glucagon⁺ cells, we analyzed the presence of other β-cell transcription factors in these cells. First, we observed differences in the subcellular localization of the transcription factors MafA and FOXO1 between the ND and T2DM donors. Using 3,3′-diaminobenzidine tetrahydrochloride immunohistochemistry, we found that MafA was mainly expressed in the nucleus in all ND donors, while it was mainly expressed in the cytoplasm in 6 of 11 donors with T2DM. (Supplementary Fig. 4A and B). Using immunofluorescence, we then examined the expression of MafA in bihormonal cells. Interestingly, MafA was either cytoplasmic or absent, but was rarely found to be nuclear in bihormonal cells even in donors who showed predominant nuclear expression (Fig. 4A). FOXO1 was expressed mainly in the cytoplasm in 7 of 9 ND donors and was localized predominantly in the nucleus in 7 of 11 donors with T2DM (Supplementary Fig. 4B). This pattern was maintained in bihormonal cells (Fig. 4C and D). Furthermore, nuclear Pdx1 was found to be expressed in a subset of glucagon⁺Nkx6.1⁺ cells (Fig. 4C). The α-cell marker Arx was also found to be expressed in the nucleus of bihormonal cells, but not all bihormonal cells expressed Arx (Fig. 4D). Nkx6.2 labeling was negative in islets from both ND donors and donors with T2DM (data not shown). Finally, we checked for the early endocrine marker Ngn3 in bihormonal cells. No Ngn3 was found in islets containing bihormonal cells (data not shown).

To determine whether the changes in islet cell identity occur before the onset of diabetes, we studied tissue from M. mulatta nonhuman primates (Table 1) that develop diabetes having characteristics similar to T2DM in humans, including a significantly decreased β-cell area (17), and that have been well characterized metabolically (16). ND animals had an islet architecture with Nkx6.1 expression similar to that of human islets (Supplementary Fig. 5). In agreement with our human data, the proportion of bihormonal and Nkx6.1⁺glucagon⁺insulin⁻ cells was higher in DM animals (bihormonal cells 0.24 ± 0.10 vs. 17.0 ± 4.4% of Nkx6.1⁺ cells, Nkx6.1⁺glucagon⁺insulin⁻ cells 0.07 ± 0.07 vs. 13.8 ± 3.9% of Nkx6.1⁺ cells, ND vs. DM; Fig. 5A and B). In contrast, the group of HI animals, that are considered to be prediabetic, had a higher proportion of bihormonal and Nkx6.1⁺glucagon⁺insulin⁻ cells compared with the ND group but this difference was not significant (bihormonal cells 0.24 ± 0.10 vs. 1.33 ± 0.29 of insulin⁺ cells, Nkx6.1⁺glucagon⁺insulin⁻ cells 0.07 ± 0.07 vs. 0.74 ± 0.45 of Nkx6.1⁺ cells, ND vs. HI; Fig. 5A and B).

Islet pathology in humans with T2DM and in DM nonhuman primates is characterized by the deposition of amyloid fibrils (22). Therefore, we investigated whether a change in islet cell identity was related to the presence of amyloid deposits both in humans and in nonhuman primates.

In pancreata of human donors, islet amyloid was present in 6 of 11 donors with T2DM and in only 1 of 9 control donors (Supplementary Table 1). In the macaques, five of six DM monkeys, three of four HI monkeys, and no ND monkeys had islet amyloid (Supplementary Table 2). The presence of islet amyloid was positively associated with a higher percentage of Nkx6.1⁺glucagon⁺insulin⁻ cells and bihormonal cells in both monkey and human pancreata (Supplementary Fig. 6). The percentages of bihormonal cells and Nkx6.1⁺glucagon⁺insulin⁻ cells both correlated with the average percentage of islet amyloid area per human donor (Fig. 6A and B). Since the extent of islet amyloidosis was heterogeneous in different islets within a donor, we were able to investigate the relationship between the extent of islet amyloidosis and the presence of Nkx6.1⁺glucagon⁺ cells per islet (Fig. 6C). Islets that had a higher degree of amyloidosis (>20%) showed a higher number of Nkx6.1⁺glucagon⁺ cells compared with islets from the same donor in which <1% of the area was composed of amyloid fibrils (Fig. 6D).

We recently showed that mature human β-cells can lose their identity and convert into α-cells ex vivo (10). Here, we provide strong evidence that the loss of β-cell identity in the islets of humans and nonhuman primates in vivo is associated with diabetes and the presence of islet amyloidosis.

In our ex vivo conversion model, the transition from human β-cell to α-cell is characterized by the presence of β-cell transcription factors (Nkx6.1, Pdx1) in glucagon-expressing cells (10). Now, we report that these cells with a mixed phenotype are present at an increased frequency in the pancreata of subjects with T2DM. It is known that T2DM is associated with an increased α/β-ratio (23–25). Interestingly, the proportion of converted cells in our study correlated with an increased α/β-ratio. Thus, although it cannot be excluded from cross-sectional data that conversion also occurs vice versa, we propose that the presence of these cells reflects ongoing conversion of β-cells to α-cells. The model of the conversion of β-cells to α-cells is also supported by recent data from a mouse model (8) in which the conversion of β-cells to α-cells occurred after β-cell–specific deletion of FOXO1, resulting in hyperglycemia. Furthermore, T2DM-associated oxidative stress was recently associated with the loss of β-cell identity (13). Consequently, we suggest that, in addition to the loss of β-cells by apoptosis in T2DM (4), the loss of β-cell identity and the subsequent conversion of β-cells to α-cells contributes to a decrease in β-cell mass.

An increasing number of reports (26,27) has drawn attention to α-cell dysfunction in T2DM, and it is generally accepted that hyperglucagonemia plays a pivotal role in the dysregulation of glucose homeostasis (7,28). Because only ∼5% of cells had a mixed phenotype, it is unlikely that hyperglucagonemia can be completely attributed to the conversion of β-cells to α-cells, but a contributory role cannot be excluded.

It is not clear how changes in metabolism could affect the conversion of β-cells to α-cells, and whether this conversion is involved in the onset and progression of diabetes. M. mulatta rhesus monkeys fed ad libitum develop diabetes in a similar fashion to the development of T2DM in humans, including the onset of obesity-associated insulin resistance, β-cell dysfunction, and the presence of islet amyloidosis (17). Whereas metabolic data in human organ donors and information on donor organs are limited, in nonhuman primates in-depth characterization of metabolic changes related to islet cell composition is possible. Overt diabetes in rhesus monkeys is preceded by an HI stage that is characterized by increased insulin secretion to maintain normal glucose levels (29). A significantly increased proportion of converted cells (either bihormonal cells or glucagon⁺ cells expressing β-cell transcription factors) was found in animals with overt diabetes, which were previously shown to have a significantly reduced β-cell area (17). Although a ×15 increase in the frequency of cells with this mixed phenotype was found in obese HI animals that are considered to be prediabetic compared with ND control animals, the difference did not reach statistical significance, possibly because of the small numbers of monkeys in each group (29). Nevertheless, these data are in agreement with our previous observations (17) that the population of obese HI monkeys did not exhibit a decrease in β-cell area, while this was evident in DM animals. A recent report (30) on rhesus monkeys that were fed a high-fat/high-sugar diet for 24 months, reported an increased α/β-ratio in the presence of normoglycemia. Both the difference in diet composition and the age of the HI animals in our study (almost twice the age at sacrifice) may have contributed to the different outcome. In addition, no bihormonal cells were reported that could point at β-cell-to-α-cell transdifferentiation in those normoglycemic animals (30). These findings indicate that the conversion of β-cells to α-cells accompanies the progression of diabetes and may worsen hyperglycemia, but its potential contribution to events leading to early β-cell failure is still unclear.

Islet amyloid deposits derived from islet amyloid polypeptide (IAPP) are associated with a reduced β-cell mass in T2DM (22,31,32). The loss of β-cell mass has generally been attributed to increased rates of apoptosis (33,34). We found that the loss of β-cell identity in pancreatic islets from both monkeys and humans with diabetes correlated with the presence of islet amyloid deposits. In monkeys, converted cells were not only absent in ND animals that are characterized by the absence of islet amyloid, but islets in both the HI and DM animals that were amyloid free had a lower (animal A) or even undetectable (animal V) proportion of converted β-cells. Compared with humans, the DM animals had more severe islet amyloidosis (>50% of islet area in 100% of islets from DM animals compared with ∼20% of islet area in 70% of islets from humans with islet amyloid), which could have contributed to the increased presence of converted cells in the DM monkeys compared with humans with diabetes. Of note, whereas we matched human donors for age, the design of the monkey study did not allow matching for age and DM animals were generally older. The two human donors without diabetes who had >1% converted cells either had islet amyloidosis (ND7) or had the highest BMI of the group (ND10, BMI 35 kg/m²). We cannot exclude that undiagnosed T2DM was present in these two organ donors.

Although there is variability in the degree of amyloidosis between islets within a human pancreas (35), it was evident that islets containing amyloid deposits showed a higher proportion of converted cells, which were frequently found directly adjacent to the amyloid deposits. Aggregated IAPP can exert a direct toxic effect on β-cells (36,37), but it is not known whether IAPP fibrils can directly induce the conversion of β-cells into α-cells. T2DM is thought to be associated with a low-grade inflammatory state (38), and IAPP oligomers can elicit activation of the inflammasome (39). It was also recently shown (40) that the overexpression of human IAPP induces β-cell inflammation and dysfunction in mice. Whether a proinflammatory local islet microenvironment may lead to both islet amyloid formation and conversion of β-cells to α-cells or whether amyloid fibrils may in fact have a direct role in the cell conversion process is not known.

The conversion of human β-cells into α-cells ex vivo following dispersion and reaggregation is likely to be affected by β-cell stress induced by breaking down cell-cell and cell-matrix interactions (10). We speculate that a similar, but more gradually developing, β-cell stress occurs in islets when T2DM develops and amyloid fibril formation occurs, leading to the loss of β-cell identity and changes in islet cell composition (41). This hypothesis is supported by the changes in subcellular localization of the β-cell factors MafA and FOXO1, while the α-cell factor Arx is in its functional nuclear compartment in bihormonal cells. A gradual conversion process in vivo could also explain the finding of bihormonal cells in the current study, in contrast to our ex vivo model, which is characterized by severe β-cell degranulation in the days following islet dispersion. Bihormonal cells have recently been described (42–45) in human pancreata from subjects with T2DM or insulin resistance, both in single or clustered β-cells and in islets in combination with mesenchymal markers. Since the ultrastructural morphology remained typical for glucagon and insulin granules, the storage of both hormones appears to remain distinct in bihormonal cells. In addition, the presence of lipofuscin bodies indicates that these cells have had a longer life span (46).

Loss of differentiated β-cells, characterized by a downregulation of key transcription factors, has been associated with β-cell dysfunction and diabetes in several studies (13–15,47,48). Whether the loss of β-cell transcription factors is directly responsible for or is merely associated with β-cell dysfunction remains to be elucidated. We now add that human β-cells (as defined by β-cell–specific transcription factors) can express both glucagon and insulin in T2DM; it is likely that these are dysfunctional β-cells (49). Even though our study was limited by the measurement of relative β-cell area, these data support the idea that conventional quantification of β-cell mass in T2DM, usually measured by insulin immunostaining, can overestimate the functional insulin-secreting β-cell mass (50,51). The evaluation of mechanisms that actively regulate functional β-cell maintenance could create new opportunities in the field of regenerative medicine, such as the induction of α-cell-to-β-cell conversion. In addition, identification of the exact mechanisms that trigger the loss of human β-cell identity may contribute to strategies to prevent and delay the progression of β-cell dysfunction in diabetes.

Acknowledgments. The authors thank BetaLogics Venture for providing the MafA antibody. The authors also thank A. Tӧns, Leiden University Medical Center, for expert technical help.

Funding. H.S.S. was supported by a grant from the Dutch Diabetes Research Foundation. Additional funding for our laboratory was obtained from the Diabetes Cell Therapy Initiative, the DON Foundation, and the Bontius Foundation.

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

Author Contributions. H.S.S. designed the study, performed the key experiment, analyzed the data, and wrote the manuscript. H.S. performed the key experiment and analyzed the data. J.H.E. and M.M.R. assisted in the data analysis. M.A.E. provided human islet preparations. E.B. and A.J.K. assisted and advised on the electron microscopy. T.J.R. gave conceptual advice and helped to write the manuscript. B.C.H. and A.C. provided nonhuman primate tissue, gave conceptual advice, and helped to write the manuscript. F.C. and E.J.P.d.K. designed the study and wrote the manuscript. F.C. and E.J.P.d.K. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.