The T Allele of TCF7L2 Rs7903146 Predicts Increased Blood Glucose After Seven Days of Bed Rest in Nondiabetic Older Adults in a Secondary Analysis of a Randomized Controlled Trial

Inpatient populations are at increased risk of hyperglycemia due to factors such as medications, physical inactivity and underlying illness, which increases morbidity and mortality. Unfortunately, clinicians have limited tools available to prospectively identify those at greatest risk. We evaluated the ability of 10 common genetic variants associated with development of type 2 diabetes to predict impaired glucose metabolism. Our research model is a simulated hospital stay (7 day bed rest protocol, standardized diet, and physical inactivity) and included a cohort of healthy older adults (n = 31, 65 ± 8 years) with baseline fasting blood glucose < 100 mg/dL. Participants completed a standard 75 g oral glucose tolerance test (OGTT) at baseline and post-bed rest. In multiple regression modeling, the transcription factor 7-like 2 (TCF7L2) rs7903146 T allele predicted elevated 2-hour OGTT blood glucose (p = 0.03925). We show that the TCF7L2 rs7903146 T allele confers risk for elevated 2-hour OGTT blood glucose in nondiabetic older adults following 7 days of bed rest.


Introduction
Nondiabetic patients who develop hyperglycemia during hospitalization have increased lengths of stay and mortality risk [1][2][3][4][5][6]. In critically ill patients, mortality increases incrementally with rising blood glucose, and patients reaching values above 300mg/dL have high mortality rates independent of diabetes diagnosis [3]. Currently, clinicians are unable to identify nondiabetic inpatients at risk for developing hyperglycemia, limiting their ability to initiate preventive therapy.
Dozens of common genetic variants are associated with increased risk for type 2 diabetes. For example, the odds of developing type 2 diabetes is 1.5 when having the the transcription factor 7-like 2 (TCF7L2) rs7903146 T allele (rs7903146 T ), which is established across many ethnic groups [7,8]. The TCF7L2 rs7903146 T variant is also associated with impaired pancreatic function and elevated glycated hemoglobin in nondiabetic individuals [9,10]. Though some genetic variants, including TCF7L2 rs7903146 T , are associated with elevated glycemic indicators in nondiabetic individuals, more evidence is needed to determine how genetic testing of these risk variants may support clinical decision making in the inpatient setting.
Inpatient hyperglycemia is multifactorial and mediated by factors like physical inactivity, medications, medical nutrition therapies, and underlying acute illnesses/chronic disease [2,4,5,11]. Bed rest in healthy research subjects models the physical inactivity aspect of a hospital stay while avoiding the confounding in uence of variable nutrition therapies and disease-related comorbidities.

Discussion
Older adults with TCF7L2 rs7903146 T risk variants are more likely to have increased 2-hour OGTT blood glucose following seven days of bed rest. TCF7L2 is a transcription factor belonging to the Wnt signaling pathway present in pancreas, liver, and other tissues [12,13]. Whole-genome chromatinimmunoprecipitation (ChIP) combined with massively parallel DNA sequencing (ChIP-Seq) analyses show that TCF7L2 binds directly to several genes involved in glucose metabolism, including PCK1, FBP1, IRS1, IRS2, AKT2, ADIPOR1, PDK4 AND CPT1A [14]. Carriers of the rs7903146 T allele exhibit impaired proinsulin conversion, higher proinsulin:insulin ratios, and greater likelihood of developing insulin-dependent type 2 diabetes [15-18], but not hepatic or extrahepatic insulin resistance [19,20]. Paradoxically, some evidence indicates that liver and other tissues appear to be involved in TCF7L2 rs7903146 T -associated glucose intolerance and insulin secretion [21].
TCF7L2 risk alleles are associated with elevated post-OGTT and nocturnal blood glucose in nondiabetic adults [17,18,22,23]. The rs7903146 T allele also associates with impaired glucose tolerance in adults with metabolic syndrome [22] and obese adolescents [17]. Healthy, middle-aged and older nondiabetic participants with rs7903146 T also exhibit higher nocturnal glucose [23]. However, similar to our ndings, prior studies indicate that rs7903146 T does not affect fasting blood glucose in healthy middle-aged adults [24]. This suggests that fasting blood glucose may not be an optimal biomarker to screen individuals with rs7903146 T for risk of developing prediabetes and type 2 diabetes.
Young, healthy Caucasian men with TCF7L2 rs7903146 T risk alleles exhibit a lower rst-phase insulin response (FPIR) to an intravenous glucose tolerance test (IGTT) compared with those with the homozygous C genotype both before and after 9 days of bed rest (p = 0.01 and p = 0.0001, respectively) [25]. Following bed rest, the participants with the TCF7L2 rs7903146 risk variants also fail to show an incremental rise of FPIR in response to insulin resistance. Though FPIR is not a concept that is directly translatable outside of the context of an IGTT, the ability to rapidly secrete insulin in response to an OGTT in the early phase (up to 30 minutes after consumption of glucose) is a similar concept [26]. The liver responds to a robust early phase insulin response by reducing release of glucose, thereby limitng the overall blood glucose response to an OGTT or a meal, and this physiological trait that is lost in the development of type 2 diabetes [27]. Here we did not observe any relationship between rs7903146 T risk alleles and insulin measures or calculated insulin sensitivity at point after a 75 g glucose load, but we completed 2-hour OGTT, which are not directly comparable to the IGTT or the FPIR.
Periods of physical inactivity promote insulin resistance in healthy adults [28,29]. If the reduced glucose tolerance we observed following a 7 day period of inactivity in healthy, nondiabetic adults persists, patients having rs7903146 T variants could be especially susceptible to long-term impairments in glucose metabolism following inpatient stays. Future research should evaluate how rs7903146 T affects blood glucose throughout hospital stay in both critically ill and non-critically ill hospital patients. Moreover, follow up studies should evaluate if rs7903146 T predicts long-term glucose intolerance following an extended period of disuse in clinical populations in patients after discharge. Finally, utilizing and OGTT, rather than fasting blood glucose, may be more appropriate for patients carrying the rs7903146 T allele.
Our analysis showed a clinically relevant association between the TCF7L2 rs7903146 T allele and risk of glucose intolerance after physical disuse; however, there were limitations. This study was not initially designed to test genotype-phenotype relationships. We recruited volunteers to test the effect of nutrition and physicial activity on a broad range of outcomes following bed rest. We feel con dent that the genotype-phenotype relationship between TCF7L2 rs7903146 T and 2-hour OGTT blood glucose described here was not affected by the group assignment, however (Supplemental Table S4). This small study was conducted in generally healthy older adults and woud not be generalizable to other aged groups or those with acute or chronic illness.

Conclusion
We show for the rst time that the TCF7L2 rs7903146 T allele associates with increased 2-hour OGTT blood glucose in nondiabetic, older adults following seven days of bed rest and physical disuse. If these ndings can be replicated in a clinical setting, the TCF7L2 rs7903146 T allele may help clinicians identify nondiabetic inpatients at greater risk for hyperglycemia.

Participants
Thirty-one healthy older adults were recruited (65 ± 8 years), provided written informed consent, medically screened, and compensated for their time as part of a larger randomized-controlled trial. A fasting glucose > 100 mg/dl, recent corticosteroid use, or evidence of chronic disease (vascular disease, unmanaged elevated blood pressure, and kidney disease) were considered exclusionary criteria for the study. All enrolled participants were community-dwelling, able to complete activities of daily living, and considered to be generally healthy. The study protocol was conducted conducted within the inpatient unit of the Clinical and Translational Research Center at the University of Texas Medical Branch (UTMB) and in accordance with the Declaration of Helsinki and approved by the UTMB Institutional Review Board. All participants provided informed consent including consent for genetic analyses. Sample size was determined by subject enrollment in the parent study and available blood samples for genetic analyses (supplemental gure S1). Recruitment and collection of blood for this secondary analysis began after the initial enrollment period for the parent clinical trial. Participants included in the present analysis were enrolled between 03/26/2014 and 10/10/2017, the latter of which was the last enrollment for the parent grant. This study was registered though clinicaltrials.gov on 03/05/2013 (NCT01846130).
Participants were assigned to one of ve experimental conditions (protein consumption patterns, small bout of walking, or amino acid supplementation) [30][31][32] that were outside of the scope of this study. Collection and storage of blood for genotyping was initiated in the middle of the parent grant, so not all study participant could be included in the anlaysis. The study statistician con rmed that there were no signi cant relationships between the study interventions and primary study outcomes (2-hour OGTT glucose and Matsuda-ISI) presented here (p > 0.05). A brief description of each study intervention is available in Supplemental Table S4.
The general experimental design is depicted in Fig. 3. Participant baseline characteristics are reported in Table 2. There were no signi cant differences in age, BMI, systolic blood pressure, diastolic blood pressure, fasting blood glucose or baseline 2-hour OGTT blood glucose between TCF7L2 rs7903146 genotype groups (Supplemental Table S1). As previously reported [30][31][32] participants completed three days of diet-stabilization/testing followed by seven days of horizontal bed rest in the UTMB Institute for Translational Sciences-Clinical Research Center (ITS-CRC). Consistent with our previous horizontal bed rest studies, subjects were continuously monitored for safety [33]. All bathing and toiletry activities were performed without bearing weight.
Individualized daily energy requirements were estimated using the Harris-Benedict equation with activity factors of 1.6 and 1.3 used for the ambulatory and bed rest period, respectively [30,31,33]. Water was provided ad libitum. The breakfast meal presented after each OGTT was adjusted to compensate for the 75 g glucose load. Energy and macronutrient intake, taking plate waste into account, were analyzed by Samples were run in triplicate and the core was able to make genotype calls for all participants and single nucleotide polymorphisms (SNP).

Statistical Analyses
Prior to analysis, descriptive statistics were calculated for all available observations from each outcome, baseline characteristics, and genotype frequencies were calculated for each SNP (Supplemental Table  S5). Pre-post bed rest comparisons of quantitative variables were made using paired t-tests. For each primary outcome (fasting glucose, 2-hour OGTT glucose and Matsuda-ISI), a one-way ANOVA was run to screen for any signi cant baseline differences with each parent study intervention and genotype groups (Supplemental Tables S2 and S4 show data for 2-hour OGTT glucose for TCF7L2 rs7903146 genotype and parent study internventions). Hardy Weinberg Equilibrium exact tests were also performed using the R package, HardyWeinberg [35]. Next, candidate multivariate regression models for each outcome were identi ed using the Feasible Solutions Algorithm [36,37]. Each model included adjustments for age and BMI [24] (along with baseline value of the outcome, if appropriate). Potential explanatory variables included the 10 SNPs, age, BMI, and parent study intervention. Based on the candidate models from the Feasible Solutions Algorithm, a model was selected for each outcome variable. All analyses were performed in R version 3.6.1.
Declarations Figure 1 1a. Individual 2-hour OGTT blood glucose changes in C/C genotype group; 1b. Individual 2-hour OGTT blood glucose changes in C/T genotype group 1c. Individual 2-hour OGTT blood glucose changes in T/T genotype group; The dotted line on 1a, 1b and 1c is 140 mg/dl, which is the cutpoint for a normal 2-hour blood glucose value during an OGTT. 1d. Average 2-hour OGTT blood glucose by genotype variant group.
Data are shown as mean ± SEM. (p=0.03925 for overall model) Figure 2 Association of PreBR 2-hour OGTT blood glucose with PostBR 2-hour OGTT blood glucose. The black line represents no change in 2-hour OGTT glucose following 7 days of bed rest. The green line represents a 30mg/dL increase in blood glucose after the 7 day bed rest protocol. Individuals with the C/C risk genotype are shown as black circles, C/T risk variants are red circles and T/T risk variants are blue circles. (p=0.03925 for overall model). Figure 3