The role of peripheral β-amyloid in insulin resistance, insulin secretion, and prediabetes: in vitro and population-based studies

Background Previous experimental studies have shown that mice overexpressing amyloid precursor protein, in which β-amyloid (Aβ) is overproduced, exhibit peripheral insulin resistance, pancreatic impairment, and hyperglycemia. We aimed to explore the effects of Aβ on insulin action and insulin secretion in vitro and the association of plasma Aβ with prediabetes in human. Methods We examined the effects of Aβ40 and Aβ42 on insulin-inhibited glucose production in HepG2 cells, insulin-promoted glucose uptake in C2C12 myotubes, and insulin secretion in INS-1 cells. Furthermore, we conducted a case-control study (N = 1142) and a nested case-control study (N = 300) within the prospective Tongji-Ezhou cohort. Odds ratios (ORs) and 95% confidence intervals (CIs) for prediabetes were estimated by using conditional logistic regression analyses. Results In the in vitro studies, Aβ40 and Aβ42 dose-dependently attenuated insulin-inhibited glucose production in HepG2 cells, insulin-promoted glucose uptake in C2C12 myotubes, and basal and glucose-stimulated insulin secretion in INS-1 cells. In the case-control study, plasma Aβ40 (adjusted OR: 2.00; 95% CI: 1.34, 3.01) and Aβ42 (adjusted OR: 1.94; 95% CI: 1.33, 2.83) were positively associated with prediabetes risk when comparing the extreme quartiles. In the nested case-control study, compared to the lowest quartile, the highest quartile of plasma Aβ40 and Aβ42 were associated with 3.51-fold (95% CI: 1.61, 7.62) and 2.75-fold (95% CI: 1.21, 6.22) greater odds of prediabetes, respectively. Conclusion Elevated plasma Aβ40 and Aβ42 levels were associated with increased risk of prediabetes in human subjects, which may be through impairing insulin sensitivity in hepatocytes and myotubes and insulin secretion in pancreatic β-cells.

1 Introduction b-amyloid (Ab), mainly including Ab40 and Ab42, is a natural product from enzymatic proteolysis of amyloid precursor protein (APP) and has been found to be deposited in the brains of Alzheimer's disease patients (1). Notably, Ab is widely generated in the brain as well as various peripheral tissues (2)(3)(4)(5). Accumulating evidence indicates that peripheral Ab involves in regulating metabolism, especially glucose and insulin metabolism. Previous animal studies have demonstrated that mice overexpressing APP, in which Ab is overproduced, exhibit peripheral insulin resistance and pancreatic impairment (6,7). In addition, Ab deposition has been found in human peripheral tissues taking part in glucose regulation, including skeletal muscle, liver, and pancreas (8)(9)(10)(11). However, the effects of Ab40 and Ab42 on these peripheral tissues remain unclear.
Prediabetes, defined as an intermediate condition between normoglycemia and diabetes, is mainly manifested as peripheral insulin resistance and b-cell dysfunction (12). There is increasing number of individuals with prediabetes worldwide, with approximated 470 million people suffering from prediabetes by 2030 (13). Prediabetes individuals have been reported to exert greater risk of type 2 diabetes (T2D), cardiovascular disease (CVD), and mortality later in life (14,15). Previous animal studies found that APP transgenic mice with high plasma Ab concentrations displayed a prediabetes phenotype (16,17). Yet, few epidemiological studies have explored the relationship between plasm Ab and prediabetes. Therefore, we examined the effects of Ab40 and Ab42 on insulin sensitivity in hepatocytes and myotubes as well as insulin secretion in pancreatic b-cells via in vitro studies. Furthermore, we evaluated the associations between plasma Ab40 and Ab42 and prediabetes in two independent populations, including a casecontrol study and a nested case-control study within the prospective Tongji-Ezhou cohort.

Glucose production assay
We performed glucose production assay according to the method described previously (21). We seeded HepG2 cells in 12well plate and treated the cells with different doses of Ab for 48 h. After treatment, we further incubated the cells with the medium for glucose production (glucose-free DMEM containing 2 mmol/l sodium pyruvate and 20 mmol/l sodium lactate) without or with 100 nmol/l insulin. After 3 h, the glucose level of supernatant was measure by a fluorescent glucose assay kit (Invitrogen, Carlsbad, CA, USA) and normalized to total protein amount.

Glucose uptake assay
Glucose uptake in C2C12 myotubes was determined as previously described (22). C2C12 myotubes in 12-well plate were treated with different doses of Ab for 48 h. Next, C2C12 myotubes were glucose starved in glucose-free DMEM for 1 h and further incubated in phenol red-free DMEM supplemented without or with 100 nmol/l insulin for 2 h. Subsequently, we detected the glucose level of supernatant with the use of glucose assay kit (Invitrogen). Glucose intake was calculated by subtracting the remaining medium glucose from the fresh medium glucose and further normalized to the total protein amount.

Insulin secretion assay
We conducted insulin secretion assay according to the method described previously (23). We seeded INS-1 cells in 12-well plate and treated the cells with different doses of Ab for 24 h. After treatment, the cells were glucose starved in the glucose-free RPMI 1640 media for 2 h and then incubated in the glucose-free RPMI 1640 media supplemented with 3 mmol/l or 15 mmol/l glucose for 2 h. After 2 h, the insulin level of supernatant was determined with a rat insulin assay kit (Mercodia, Uppsala, Sweden), and total protein amount was quantified to normalize insulin values.

Study design and population
The initial case-control study included 1142 participants (571 newly diagnosed prediabetes cases and 571 healthy controls) in Wuhan, China. We recruited cases from patients for the first time diagnosed as prediabetes in Tongji Hospital from 2012 to 2015. Concurrently, healthy controls were recruited from a general population receiving a regular medical examination in Tongji Hospital and matched 1:1 to cases on basis of age ( ± 3 years) and gender. In this study, we included participants who were aged ≥30 and ≤80 years, body mass index (BMI) <40 kg/m2, no previous diagnosis of prediabetes or diabetes, and no use of antihyperlipidemic medication. Additionally, patients diagnosed as any other clinically systemic illness, infectious disease, chronic inflammatory illness, or acute disease were excluded.
To clarify the prospective association of plasma Ab concentration with risk of prediabetes, we further conducted a nested case-control study within the Tongji-Ezhou cohort, a prospective cohort of 5533 participants from Echeng Stell, including 3101 retired employees and 2432 in-service employees. The cohort enrollment period was 2013-2015, and the first followup survey was completed by mid-2020. We conducted this study within the sub cohort of retired employees because the incidence rate of prediabetes was low in young in-service employees. At the first follow-up, 119 retired employees were diagnosed as new-onset prediabetes cases by detecting fasting plasma glucose (FPG). According to the same inclusion criteria as the case-control study, we excluded 2 new-onset prediabetes cases with age >80 years. We also excluded 17 new-onset prediabetes cases without adequate blood samples. Two control participants were matched to each case on basis of age ( ± 3 years) and gender from retired employees without prediabetes or diabetes. Eventually, 100 cases and 200 ageand gender-matched controls were included for analysis.
These two population-based studies were approved by the Ethics Committee of Tongji Medical College and conducted by complying with the Declaration of Helsinki. Written informed consent was provided by the participants included in the study.

Assessment of prediabetes
In the case-control study, the definition of prediabetes was in accordance with the 1999 WHO criteria (24): FPG ≥6.1 and <7.0 mmol/l or 2-h oral glucose tolerance test (OGTT) ≥7.8 and <11.1 mmol/l. In the Tongji cohort, new-onset prediabetes was diagnosed when FPG was ≥6.1 and <7.0 mmol/l.

Assessment of covariates
Sociodemographic characteristics of all participants, including age, gender, family history of diabetes, lifestyle habits (alcohol drinking, smoking, and physical activity), and history of diseases (CVD and hypertension) were obtained from a standardized questionnaire. The definition of physical activity was regular exercise for more than 60min/week in the past 6 months. The trained staff measured weight (kg) and height (m) and calculated BMI by dividing weight by the square of height. We determined fasting plasma insulin (FPI), FPG, total cholesterol, triacylglycerols, HDL-cholesterol (HDL-C), and LDL-cholesterol (LDL-C) using fasting plasma samples as previously described (25). 2-h OGTT plasma samples were collected from subjects who were enrolled from Tongji Hospital and used to determine 2-h post-glucose load values. We calculated the index of homeostasis model assessment of insulin resistance (HOMA-IR) and b-cell function (HOMA-b) according to the following equations: HOMA-IR = FPI (pmol/l) × FPG (mmol/l) ÷ 156.3; HOMA-b = 2.88 × FPI (pmol/l) ÷ [FPG (mmol/l) -3.5]. The triglyceride-glucose (TyG) index was calculated as: ln [FPG (mg/dl) × fasting TG (mg/dl) ÷ 2].

Measurement of plasma Ab levels
We simultaneously determined Ab40 and Ab42 levels in fasting plasma with the use of validated assay kits (Meso Scale Discovery, Rockville, USA). Both the within-and between-assay coefficients of variation of plasma Ab40 and Ab42 were <10%. Plasma Ab levels of the included participants were above the limit of detection (Ab40: 20.0 ng/l; Ab42: 2.5 ng/l). The assays were performed by investigators who were blind to prediabetes status.

Statistical analysis
For the in vitro studies, all data are expressed as mean ± standard error of mean (SEM) of three independent experiments. We assessed the differences among groups using one-way analysis of variance and further conducted multiple comparisons by least significant difference test.
For the population-based studies, characteristics of prediabetes cases and control subjects were compared using t-test, Mann-Whitney U test, or c 2 test. We further evaluated the correlation between plasma Ab and the variables of interest with Spearman correlation test. ORs and 95% CIs for prediabetes were estimated by conditional logistic regression analyses, in which plasma Ab40 and Ab42 levels were classified into quartiles according the distribution in control group or as continuous variables presented as an increment by 30 ng/l (Ab40) or 5 ng/l (Ab42). Adjustments were made for several potential confounders, including age (years), gender (female or male), BMI (<18.5, 18.5-23.9, 24-27.9, or ≥28 kg/m2), family history of diabetes (yes or no), alcohol drinking habit (current, former, or never), smoking habit (current, former, or never), physical activity (yes or no), CVD (yes or no), and hypertension (yes or no). We conducted the linear trend test by assigning each quartile with the median of plasma Ab40 or Ab42 and using it as a continuous variable. Receiver-operating characteristic (ROC) curves were plotted to estimate the prediction of plasma Ab levels on prediabetes. We compared area under the curves (AUCs) of model 1 including traditional risk factors (age, gender, BMI, family history of diabetes, alcohol drinking habit, smoking habit, physical activity, CVD, hypertension, FPG, and FPI) and model 2 with plasma Ab40 and Ab42 levels further added in.
All data analyses were conducted with the use of SPSS 20.0 (SPSS Inc., Chicago, IL) and R 4.2.2 (The R Foundation, http:// www.r-project.org). All presented p values are 2-tailed and considered as significance at level of 0.05.

In vitro studies
Cells were treated with different doses of Ab40 and Ab42, and viability was detected at 12, 24, and 48 h. Exposure to 20 mmol/l Ab40 and Ab42 induced a significant decrease in cell viability for all cell lines (Supplemental Figures 1-3). The viability of HepG2 cells and C2C12 myotubes were still above 80% after exposure to 20 mmol/l Ab40 and Ab42 for 48 h. In INS-1 cells treated with 20 mmol/l Ab40 and Ab42 for 48 h, cell viability decreased to 67% and 77%, respectively. To maintain high cell viability (>80%), INS-1 cells were treated with various doses of Ab40 and Ab42 for 24 h in insulin secretion assay. To rule out the effects of cytotoxicity of Ab on subsequent experiments, the amounts of glucose production, glucose uptake, and insulin secretion were normalized to total protein amount of the remaining cells in the culture plate.
As expected, insulin significantly inhibited the glucose production in HepG2 cells and promoted the glucose uptake in C2C12 myotubes (p <0.001) (Figures 1A-D). Ab40 and Ab42 dosedependently attenuated the inhibitory effect of insulin on the glucose production in HepG2 cells (Figures 1A, B). The promotional effect of insulin on the glucose uptake in C2C12 myotubes gradually decline with increasing dose of Ab40 and Ab42 (Figures 1C, D). In addition, Ab40 and Ab42 dosedependently suppressed the basal and glucose-stimulated insulin secretion in INS-1 cells (Figures 1E, F). Conversely, neither Ab40-1 nor Ab42-1 affected the insulin sensitivity in HepG2 cells and C2C12 myotubes and insulin secretion in INS-1 cells (Figures 1A-F).

Case-control study
General characteristics of the 1142 participants (571 newlydiagnosed prediabetes cases and 571 matched controls) are described in Table 1. Prediabetes cases showed higher BMI, higher levels of FPI, FPG, triglyceride, LDL-C, HOMA-IR and TyG index as well as lower levels of HDL-C and HOMA-b compared with control subjects. A higher proportion of prediabetes cases had a family history of diabetes, CVD, and hypertension. Additionally, prediabetes cases exhibited higher plasma Ab40 and Ab42 levels compared to control subjects.

Discussion
In the in vitro studies, Ab40 and Ab42 dose-dependently attenuated insulin-inhibited glucose production in HepG2 cells and insulin-promoted glucose uptake in C2C12 myotubes. Meanwhile, Ab40 and Ab42 dose-dependently inhibited the basal and glucose-stimulated insulin secretion in INS-1 cells. In the crosssectional and prospective population-based studies, we consistently found that elevated plasma Ab40 and Ab42 levels were associated with higher odds of prediabetes. Additionally, plasma Ab40 and Ab42 significantly improved the predictive value for prediabetes. It has been reported that APP transgenic mice with high plasma Ab40 and Ab42 concentrations exhibited peripheral insulin resistance (17,26). Peripheral insulin resistance in the mice was ameliorated by the active immunity and passive immunity for Ab (26,27). Similarly, mice deficient in BACE1, a hydrolase for APP, showed decreased Ab production and improved insulin resistance when fed a high-fat diet (28). These experimental studies suggest Ab as a negative regulator of peripheral insulin sensitivity. The current study found positive relevance of plasma Ab40 and Ab42 to HOMA-IR and TyG index in healthy subjects. Meanwhile, we observed that both Ab40 and Ab42 attenuated insulin action on hepatocytes and myotubes. Our findings indicate that Ab may promote peripheral insulin resistance in human by directly impairing insulin signaling in liver and skeletal muscle. Supporting our findings, Zhang and colleagues discovered that Ab could impair hepatic insulin signaling via activating Janus kinase 2 (17,27). In addition, Ab has also been reported to impair neuronal insulin signaling through competing for insulin binding to the insulin receptor (29), removing insulin receptors from neuronal surface (30,31) and activating c-Jun N-terminal kinase to trigger insulin receptor substrate-1 inhibition (32,33). But whether these mechanisms are involved in Ab-induced insulin resistance in peripheral tissues needs to be further explored.
Previous study, using gene regulation network analyses, identified APP as a top candidate gene for the regulation of insulin secretion from pancreatic islets (34). Meanwhile, the study found that knockout of APP in mice enhanced insulin secretion from pancreatic islets through an unclear mechanism. Our in vitro study revealed the negative effect of Ab on insulin secretion, suggesting APP-derived Ab may mediate the role of APP in pancreatic islets. Supporting this hypothesis, mice overexpressing APP exhibited Ab accumulation in the pancreatic islets and decreased insulin levels in Ab-positive regions (7). Additionally, Ab deposition has been detected in the pancreas of subjects with T2D, leading us to speculate that peripheral Ab might be a biomarker for b-cell dysfunction in human (8,9). However, we did not find significant correlations between plasma Ab40 and Ab42 and HOMA-b among controls. This may be explained by the normal b-cell function in healthy controls, and we found significant and negative association of plasma Ab40 with HOMA-b when combining both cases and controls (data not shown). Further Values are means (standard deviations) for parametrically distributed data, medians (interquartile ranges) for nonparametrically distributed data, or n (%) for categorical data.
epidemiological researches are needed to evaluate the forecast value of plasma Ab for b-cell dysfunction.
Prediabetes is a metabolic intermediate state between normoglycemia and T2D, with 5-10% of prediabetic people progressing to T2D each year (15). Prediabetes is considered as a critical stage in preventing or delaying the onset of T2D due to its reversibility (15), therefore, it is important to understand the underlying mechanisms of prediabetes and to develop corresponding human biomarkers. This study, for the first time, reveals that plasma Ab40 and Ab42 are positively associated with prediabetes in human subjects. Similarly, our previous report demonstrated that higher plasma Ab40 and Ab42 levels were associated with increased risk of T2D (35). Previous case-control study also observed higher serum Ab-autoantibody concentration, a biomarker reflecting Ab level, in subjects with T2D (36). Yet, another case-control study reported that T2D cases exhibited lower plasma Ab40 and Ab42 levels than the controls (37). A plausible explanation for the inconsistency might be that the aforementioned study included cases with using hypoglycemic agents (37), which have been reported to affect plasma Ab40 and Ab42 levels (38)(39)(40). Therefore, the use of hypoglycemic agents should be considered when plasma Ab is used as a biomarker for prediabetes and T2D.
Our study is the first that has combined in vitro studies and population-based studies to systematically explore the role of peripheral Ab in insulin resistance, insulin secretion, and prediabetes. However, our study also has several limitations. Firstly, we were unable to determine the causality between plasma Ab and prediabetes due to the observational nature of populationbased studies. Secondly, the findings from population-based studies might be affected by residual confounding of other unmeasured factors, although we controlled for multiple risk factors of prediabetes. Thirdly, new-onset prediabetes was diagnosed only based on FPG in the Tongji-Ezhou cohort, and subjects who developed T2D might be misdiagnosed as prediabetes. Taken together with our previous study (35), plasm Ab levels were not significantly different between prediabetes and T2D individuals. Hence, the positive association between plasma Ab and new-onset prediabetes could persist if misdiagnosed subjects were excluded. Fourthly, glycated hemoglobin, an important parameter for glucose homeostasis, was not determined in our population-based studies. Fifthly, cognitive function was not evaluated in our populationbased studies. Considering the positive associations of prediabetes and T2D with cognitive impairment and Alzheimer's disease (41,42), further studies are needed to explore the role of plasma Ab in linking these conditions.
In conclusion, elevated plasma Ab40 and Ab42 levels were associated with increased risk of prediabetes in human subjects, which may be through impairing insulin sensitivity in hepatocytes and myotubes and insulin secretion in pancreatic b-cells. Plasma   Ab could be used as a predictor for prediabetes, and reducing plasma Ab level may be a novel therapy for prediabetes. The molecular mechanisms of Ab affecting peripheral insulin sensitivity and insulin secretion need to be further clarified.

Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethics statement
The studies involving human participants were reviewed and approved by Ethics Committee of Tongji Medical College. The patients/participants provided their written informed consent to participate in this study.