Endogenous Secretory RAGE Increases with Improvements in Body Composition and is Associated with Markers of Adipocyte Health

Introduction

The receptor for advanced glycation end products (RAGE) is a multi-ligand cell surface protein belonging to the immunoglobulin superfamily of proteins [1]. Upon ligand binding, RAGE initiates an inflammatory signaling cascade, culminating in the activation of nuclear factor kappa B (NF-κB) and subsequent transcription of inflammatory cytokines, including RAGE itself [2]. This cyclical cascade contributes to the development of metabolic disease such as obesity, diabetes, and atherosclerosis [3–8].

Adipose from individuals with obesity possesses more RAGE protein compared to lean, age-matched individuals [7]. Additionally, high-fat diet-fed (HFD) RAGE-null mice are protected from adipose hypertrophy, inflammation, and insulin resistance [5–8]. Therefore, attenuation of RAGE signaling may be a viable strategy to treat obesity and prevent development of secondary complications.

Soluble RAGE (sRAGE) isoforms act as decoy receptors for RAGE by sequestering RAGE ligands and attenuating RAGE signaling. Isoforms of sRAGE include cleaved RAGE (cRAGE) produced via proteolytic shedding of the RAGE ectodomain and endogenous secretory RAGE (esRAGE) generated by alternative splicing of RAGE pre-RNA [9, 10]. Studies have demonstrated reduced total sRAGE in individuals with obesity compared to lean individuals [11–14]. Total sRAGE is also negatively associated with BMI [14–17]. We recently reported that sRAGE isoforms are reduced in individuals with impaired glucose tolerance and type 2 diabetes mellitus (T2DM) and that esRAGE was strongly correlated to hyperinsulinemic-euglycemic clamp derived glucose disposal rates [14]. Several studies have also demonstrated associations between decreased total sRAGE and increased adiponectin/leptin ratio [18] as well as increased risk of T2DM [14, 19]. In a HFD model of obesity, mice treated with recombinant sRAGE were protected from reduced adiponectin expression, insulin resistance, and weight-gain [6].

While many studies have demonstrated the relationship between obesity and RAGE, few studies have examined the effects of weight-loss on sRAGE isoform concentrations [11, 20–22]. The only study to examine the effects of weight-loss via dietary intervention found no change in total sRAGE given a very low-calorie diet but did find a negative association between the change in BMI and the change in sRAGE [22]. However, this study included a heterogeneous group of individuals with obesity possessing a number of other comorbidities which may affect weight-loss and RAGE biology (i.e. diabetes, psoriasis, hypothyroidism etc.) [22]. Brix et al observed an approximate 20% increase in total sRAGE following weight-loss via bariatric surgery [11], whereas others have demonstrated the ability of baseline total sRAGE to predict weight-loss following bariatric surgery [20, 21].

There is a clear lack of human evidence examining the effect of diet-induced weight-loss on sRAGE in obese individuals given the strong evidence implicating RAGE biology as a driver of obesity. Alternate day fasting (ADF) is a relatively new dieting scheme that has become increasingly popular of late. This is likely in part due to the flexibility this dieting strategy lends by allowing the individual to eat ad libitum every other day. In the context of this study we are interested in ADF due to its ability to confer a negative energy balance and promote weight-loss and other positive health benefits which may be related to modulations in the concentrations of sRAGE isoforms. Here we present post-hoc analyses of our previously published randomized control trial (RCT) (NCT00960505) [23] to test the hypothesis that weight-loss via alternate day fasting would increase sRAGE isoforms and the changes in sRAGE concentrations would relate to improvements in adipokine health.

Methods

Study Population

Participants (n=42) were recruited from the greater Chicago area. Inclusion criteria included a BMI of 25.5–39.9 kg/m², sedentary behavior for the previous 3 months (<60 minutes per week of light activity), and 18–65 years of age. Exclusion criteria included history of cardiovascular disease, type 1 or type 2 diabetes, use of weight-loss medications, glucose control medications, and lipid lowering medications, unstable weight for 3 months prior to screening (weight-loss or gain of >4 kg), perimenopause or irregular menstrual cycle, pregnancy, and current smoking. Participants were randomized into the control (CON) or ADF group. This study was approved by the office for the protection of research subjects at University of Illinois at Chicago. All subjects gave written informed consent prior to enrollment.

Intervention

This report includes data from the first 24 weeks of a 1-year trial, which has been previously described [23, 24]. Briefly, all participants performed a 4-week run-in phase in which subjects were instructed to eat their normal diet and remain weight stable. For the duration of the trial, subjects were instructed not to change their physical activity habits to avoid confounding their daily energy expenditure (determined via doublelabeled water method). Participants were subsequently randomized into the ADF or CON group. Baseline measures were taken after randomization at the beginning of the first week (W1) of the trial.

The ADF intervention required subjects to consume 25% of baseline energy needs as a lunch (12pm–2pm) on “fast days” and 125% of their energy needs across 3 meals on subsequent “feast days.” Meals were provided for the ADF group for the first 12 weeks and during the last 12 weeks of the trial, participants in the ADF group met regularly with a registered dietitian or nutritionist while continuing the diet on their own. Participants randomized into the CON group were instructed not to change their diet and were not provided meals or dietary counseling.

Anthropometric Measures

Body weight was measured via digital scale with participants wearing a hospital gown at weeks 1, 12, and 24 (W1, W12 and W24). Body composition (lean mass, body fat, and body fat percentage) were measured via dual-x-ray absorptiometry (QDR 4500W; Hologic) and visceral adipose tissue (VAT) was quantified via MRI (3.0 Tesla MRI Scanner, Achieva, Philips Healthcare) at W1 and W24. Subcutaneous adipose tissue (SAT) was derived by subtracting VAT from total body fat.

Blood Measures

Blood samples were taken from an antecubital vein at W1, W12, and W24 following a 12-hour fast. Blood was collected in EDTA-treated vacutainers, immediately centrifuged to isolate plasma, and frozen at −80°C until future analysis. Total sRAGE was quantified via commercially available ELISA (R&D Systems Inc., Minneapolis, MN, USA). This measure of sRAGE uses a monoclonal antibody that detects both the cleaved and alternatively spliced sRAGE isoforms thus giving a measure of total sRAGE. The intra- and inter-assay coefficient of variation for the sRAGE kits were 4.39%, and 2.39–7.54% respectively. The esRAGE isoform was measured via another commercial ELISA (As One International, Mountain View, CA, USA). The esRAGE ELISA uses a monoclonal antibody that exclusively detects esRAGE. The intra- and inter-assay coefficient of variation for the esRAGE kits were 6.37%, and 4.78–8.97% respectively. Plasma cRAGE was calculated by subtracting the concentration of esRAGE from the concentration of total sRAGE in each sample as previously described [14, 25–27]. Subsequently, a ratio of cRAGE to esRAGE (cRAGE:esRAGE) was derived as previously described [14, 27]. Given that cRAGE and esRAGE are potentially produced by independent mechanisms, this metric helps determine if the production of these isoforms are proportionately or disproportionately modulated by our intervention. This data lends insight into potential regulators of cRAGE and esRAGE producing mechanisms. Fasting plasma glucose, fasting insulin, TNF-α, IL-6, adiponectin, and leptin, were all measured as previously described [23]. A HOMA-IR (fasting insulin in mU/L X fasting glucose in mg/dL / 405) value ≥ 3.6 was considered to indicate insulin resistance in accordance with Stern, S.E. et al [28].

Statistical Analysis

All data were tested for normality via Shapiro-Wilk test and found to be nonnormally distributed. Baseline measures and change scores were compared between ADF and CON groups via Mann-Whitney U-test. The entire cohort was also grouped by median of weight-loss achieved at W12 and W24 as well as fat-loss, VAT-loss, and SAT-loss at W24. Mann-Whitney U-test was used to compare sRAGE isoforms between groups stratified in these ways. Within group comparisons were made via Wilcoxon Signed Rank Test. Spearman’s rho correlation coefficient was used to determine relationships between weight-loss, fat-loss, and changes in metabolic health markers with changes in sRAGE isoforms. SPSS v24 (IBM, Armonk, NY, USA) was used to perform statistical analyses and GraphPad Prism 7.04 was used to make the figures (GraphPad Software, La Jolla, California, USA). p<0.05 was considered significant and data are presented as median (Interquartile Range) as follows: median (25^th percentile, 75^th percentile).

Results

Effects of ADF on Anthropometrics, Insulin Resistance, and Markers of Metabolic Health

Baseline characteristics for ADF and CON groups are presented in Table 1. The ADF and CON groups were matched for sex distribution, age (p=0.715), weight (p=0.521), BMI (p=0.840), and fat mass (p=0.291) (Table 1). Both groups were insulin resistant as determined by HOMA-IR. Over 24 weeks, the ADF group lost more weight (p<0.0001), and fat mass (p<0.0001) than CON (Table 1). By week 12, the ADF group had a greater improvement in insulin sensitivity (HOMA-IR) compared to (p=0.031, Table 1). ADF intervention did not have an effect on circulating adipokines or inflammatory markers (Table 1).

Table 1. Baseline group characteristics and changes in anthropometric and metabolic measures and sRAGE isoforms.

Group baseline and change scores following weeks 12 and 24 of ADF for anthropometric, metabolic and circulating factors. Groups were matched for, age, BMI, fat mass, HOMA-IR and sRAGE measures. Between group comparisons were assessed by Mann-Whitney U-test. p values for between group differences are presented with significant differences bolded (p<0.05). Median and Interquartile Range are presented as follows: median (25^th percentile, 75^th percentile).

Variable, Units	Baseline		p	ΔW1-W12		p	ΔW1-W24		p
	Control	ADF		Control	ADF		Control	ADF
n	22	20		-	-		-	-
Sex, M/F	4/18	3/17	-	-	-	-	-	-	-
Age, y	43 (33, 55)	44 (35, 55)	0.715	-	-	-	-	-	-
Weight, kg	87.7 (78.9, 99.8)	86.7 (84.4, 105.2)	0.521	−0.8 (−2.0, 0.5)	−5.6 (−8.6, −2.7)	<0.001	−0.3 (−1.9, 1.0)	−6.8 (−9.5, −3.5)	<0.001
BMI, kg/m2	34.5 (29.7, 36.1)	33.0 (30.9, 36.8)	0.840	−0.3 (−0.8, 0.2)	−2.4 (−3.1, −0.8)	<0.001	−0.1 (−0.7, 0.4)	−2.2 (−3.1, −0.8)	<0.001
Fat Mass, kg	34.7 (26.7, 39.6)	36.7 (33.2, 45.1)	0.290	−	−	−	−0.7 (−1.3, 0.1)	−3.6 (−7.9, −2.6)	<0.001
Percent Fat, %	41.3 (35.4, 42.5)	43.2 (39.3, 47.4)	0.650	−	−	−	−0.88 (−1.78, −0.01)	−4.8 (−8.4, −2.6)	<0.001
Visceral Fat, kg	2.13 (1.13, 3.42)	1.87 (1.28, 3.18)	0.684	−	−	−	−0.07 (−0.22, 0.10)	−0.35 (−0.74, −0.07)	0.020
Glucose, mg/dL	89.0 (82.0, 92.3)	94.0 (87.8, 101.3)	0.037	0.0 (−3.0, 3.5)	−8.0 (−11.0, 1.0)	0.031	5.5 (−6.3, 13.0)	0.5 (−7.5, 7.5)	0.246
Insulin, mU/L	14.6 (9.4, 19.9)	16.5 (7.0, 22.4)	0.850	−1.7 (−4.2, 1.9)	−4.5 (−11.3, 0.9)	0.115	−0.9 (−9.1,4.9)	−7.4 (−12.0, −2.2)	0.047
HOMA-IR, AU	3.1 (2.0, 4.4)	4.1 (1.5, 4.9)	0.580	−0.3 (−0.7, 0.8)	−1.5 (−2.9, −0.1)	0.031	0.1 (0.0, 0.5)	0.03 (−0.30, 0.18)	0.051
Adiponectin, ng/mL	3289 (1565, 4154)	4655 (2674, 5937)	0.056	74 (−817, 559)	−323 (−1331,405)	0.562	589 (−352, 1192)	−222 (−2020, 2038)	0.199
Leptin, ng/mL	45.7 (27.8, 75.0)	57.3 (34.3, 81.1)	0.279	−2.5 (−15.0, 1.7)	−4.7 (−43.7, 7.8)	0.893	0.1 (−9.7, 39.0)	−2.7 (−39.7, 8.0)	0.078
Adiponectin:Leptin, AU	87.3 (49.9, 117.8)	70.9 (59.1, 116.7)	0.880	3.4 (−15.1,23.6)	8.0 (−22.5, 47.7)	0.520	6.8 (−23.8, 29.0)	5.0 (−22.6, 56.6)	0.706
IL-6, pg/mL	2.0 (1.3, 3.2)	1.6 (1.1,2.4)	0.289	0.1 (−0.3, 0.7)	0.2 (−0.1, 1.0)	0.495	0.2 (−0.1, 1.5)	0.3 (−0.2, 0.7)	0.772
TNF-α, pg/mL	1.6 (1.2, 2.3)	1.4 (1.0, 1.7)	0.146	−0.1 (−0.4, 0..2)	0.0 (−0.1,0.2)	0.141	−0.3 (−0.7, 0.2)	0.0 (−0.3, 0.2)	0.202
Total sRAGE, pg/mL	976 (699, 1189)	1105 (790, 1390)	0.375	−50 (−235, 52)	32 (−168, 122)	0.201	−70 (−258, 6)	18 (−154, 249)	0.056
cRAGE, pg/mL	759 (519, 993)	852 (628, 1129)	0.449	−60 (−173, 17)	−23 (−151,101)	0.258	−68 (−189, 33.6)	−34 (−152, 229)	0.081
esRAGE, pg/mL	203 (157, 272)	237 (170, 318)	0.835	−5 (−87, 55)	16 (−30, 92)	0.108	−21 (−72, 16)	15 (−30, 78)	0.023
cRAGE:esRAGE, AU	3.5 (2.9, 4.4)	3.9 (2.9, 4.8)	0.402	−0.1 (−0.8, 0.5)	−0.2 (−0.7, 0.2)	0.792	0.1 (−0.5, 0.4)	−0.2 (−0.7, 0.3)	0.201

Effects of ADF on sRAGE Isoforms

There was no effect of time on any sRAGE isoform or cRAGE:esRAGE ratio within the ADF or CON groups when analyzed by a Friedman’s one-way non-parametric repeated measures ANOVA (Table 2). We next performed a number of post hoc analyses to determine the between group effects of ADF and CON on sRAGE isoforms via the Mann-Whitney U-Test. There were no differences in the change in sRAGE isoforms in ADF compared to CON at W12 (Table 1). There was a trend towards significance between ADF and CON for the change in total sRAGE (p=0.056) and cRAGE (p=0.081) at W24 (Table 1). The change in esRAGE at W24 was significantly different in ADF compared to CON (p=0.023, Table 1).

Table 2. Effect of time on sRAGE Isoforms in control and alternate day fasting groups.

The effect of time on concentrations of sRAGE isoforms within each group were compared via Friedman’s non-parametric repeated measures ANOVA. Median and Interquartile Range are presented as follows: median (25^th percentile, 75^th percentile).

Variable, Units	CON			Effect of Time	ADF			Effect of Time
	W1	W12	W24		W1	W12	W24
	Median (IQR)	Median (IQR)	Median (IQR)	p	Median (IQR)	Median (IQR)	Median (IQR)	p
Total sRAGE, pg/mL	976 (699, 1189)	813 (705, 1218)	813 (606, 1121)	0.150	1105 (790, 1390)	1285 (706, 1503)	1093 (663, 1603)	0.819
cRAGE, pg/mL	759 (519, 993)	613 (583, 940)	617 (473, 858)	0.368	852 (628, 1129)	943 (543, 1164)	877 (519, 1225)	0.861
esRAGE, pg/mL	203 (157, 272)	192 (146, 275)	179 (136, 262)	0.368	237 (170, 318)	260 (179, 334)	238 (141, 403)	0.705
cRAGE:esRAGE, AU	3.5 (2.9, 4.4)	3.4 (2.7, 4.8)	3.2 (2.8, 4.4)	0.810	3.9 (2.9, 4.8)	3.6 (3.0, 4.8)	3.4 (2.8, 4.3)	0.387

Effects of Weight- and Fat-Loss on sRAGE Isoforms

A large range of weight change was noted in both groups at W24 (Range: ADF: 18kg to 0kg; CON −7kg to +3kg). Therefore, the effects of weight-loss and fat-loss on changes in sRAGE were evaluated. The change in body weight at W24 negatively correlated with the change in total sRAGE (rho=−0.34, p=0.030), cRAGE (rho=−0.32, p=0.042), and esRAGE (rho=−0.38, p=0.015) (Figure 2).

Figure 2. Correlations between the changes in weight and the changes in sRAGE Isoforms at weeks 12 and 24.

Spearman’s Rho correlations were performed to examine if weight-loss at 12 (A-D) and 24 (E-H) weeks was related to the change in sRAGE isoforms at corresponding time points in the study. Correlations presented are for the entire cohort (correlations were not run in individual groups). Data for each individual are presented with those in the “Above” group represented by open circles (o) and those in the “Below” group are represented by closed squares (∎). Above: Individuals above the median weight-loss by week 12 or 24 (greater weight-loss), Below: Individuals below the median weight-loss by week 12 or 24.

To determine the effect of weight-loss on sRAGE isoforms we stratified our entire cohort by median weight-loss achieved at W12 (1.95kg) and W24 (2.09kg). At W12, the group above the median weight-loss (Above) lost 6 (−9, −4) kg of body weight (p<0.001), and the group below the median weight-loss (Below) did not significantly change their body weight −1 (−2, 0.5) kg (p=0.217). There were no differences in the change in sRAGE isoforms between those above and below median weight-loss at W12 (Figure 1, A-D).

Figure 1. Change in sRAGE isoforms by median weight-loss.

To examine the effect of weight-loss on the change in sRAGE isoforms following 12 (A-D) and 24 weeks (E-H) the entire cohort was stratified by median weight-loss after 12 and 24 weeks respectively. Change scores were compared via Mann-Whitney U-Test. # Significantly different between groups, (p<0.05). Data for each individual is presented with those in the “Above” group represented by open circles (o) and those in the “Below” group are represented by closed squares (∎). The bars identify the median value and the 25^th and 75^th percentiles of the median. Above: Individuals above the median weight-loss by week 12 or 24 (greater weight-loss), Below: Individuals below the median weight-loss by week 12 or 24.

At W24, those above the median lost 7 (−10, −4) kg (p<0.001), and those below the median did not significantly change their body weight 0 (−1.5, 1) kg (p=0.715). The changes in all sRAGE isoforms and cRAGE:esRAGE were significantly different between those above and below the W24 median weight-loss (Figure 1E-H). Withingroup comparisons revealed those below the W24 median weight-loss significantly decreased their cRAGE (p=0.008) and esRAGE (p=0.025) whereas those above the W24 median maintained their sRAGE concentrations. Interestingly, there was a significant increase in cRAGE:esRAGE in the group above the W24 median weight-loss indicating that changes in cRAGE and esRAGE may not be similarly responsive to weight-loss.

Given the role of RAGE in adipose biology, we examined the relationship between fat-loss and sRAGE isoforms. The change in fat mass (DXA) was negatively correlated with the change in total sRAGE (rho=−0.33, p=0.039) and esRAGE (rho=0.47, p=0.002). We also stratified our cohort by median fat-loss at W24 (1.61kg) (Figure 3). Those above the W24 median lost 4 (−7, −3) kg of fat (p=0.001), and those below lost 1 (−1, 0) kg of fat (p=0.033). The changes in total sRAGE, esRAGE, and cRAGE:esRAGE were significantly different between those above and below the W24 median fat-loss (Figure 3, A,B, and D). Within group comparisons demonstrate that those above the W24 median fat-loss significantly increased their esRAGE (p=0.032) and decreased their cRAGE:esRAGE ratio (p=0.021). Conversely, those below the W24 median fat-loss significantly decreased their total sRAGE (p=0.019), esRAGE (p=0.014), and cRAGE (p=0.030).

Figure 3. Changes in sRAGE isoforms by median fat-loss.

To examine the effect of fat-loss on changes in sRAGE isoforms, the entire cohort was stratified by median fatloss achieved at week 24. Fat mass was quantified via dual x-ray absorptiometry (DXA) at weeks 1 and 24 only. Changes in Total sRAGE (A), cRAGE (B), esRAGE (C), and cRAGE:esRAGE (D) at week 24 were compared between groups via Mann Whitney UTest. # Significantly different between groups, (p<0.05). Data for each individual are presented with those in the “Above” group represented by open circles (o) and those in the “Below” group represented by closed squares (∎). The bars identify the median value and the 25^th and 75^th percentiles of the median. Spearman’s Rho correlations were performed to examine if fat-loss at 24 weeks was related to the change in Total sRAGE (E), cRAGE (F), esRAGE (G), and cRAGE:esRAGE (H) at week 24. Correlations presented are for the entire cohort (correlations were not run in individual groups). Data for each individual is presented with those in the “Above” group represented by open circles (o) and those in the “Below” group are represented by closed squares (∎). Data are presented as Mean±SEM. Above: Individuals above the median fat-loss by week 24, Below: Individuals below the median fat-loss by week 24.

Effects of Subcutaneous and Visceral Fat-Loss on sRAGE Isoforms

Given the relationship between fat-loss and sRAGE we examined the effect of changes in specific fat depots on sRAGE isoforms. We stratified our cohort by the median VAT-loss (0.12 kg) (MRI). Those above median VAT-loss lost 0.51 (−0.78, −0.29) kg of VAT (p<0.001), and those below the median did not alter their VAT −0.02 (−0.08, 0.11) kg (p=0.687). The changes in sRAGE isoforms were not significantly different between those above and below median VAT-loss nor was the change in VAT correlated to the change in any sRAGE isoforms at week 24 (Figure 4). However, there was a significant in-group difference observed in cRAGE:esRAGE ratio whereby those above the median VAT-loss significantly decreased their cRAGE:esRAGE ratio (p=0.035).

Figure 4. Changes in sRAGE isoforms by median visceral fat-loss.

To examine the effect of visceral fat-loss on changes in sRAGE isoforms, the entire cohort was stratified by median visceral fat-loss achieved at week 24. Visceral fat mass was quantified via dual x-ray absorptiometry (DXA) at weeks 1 and 24 only. Changes in Total sRAGE (A), cRAGE (B), esRAGE (C), and cRAGE:esRAGE (D) at week 24 were compared between groups via Mann Whitney U-Test. # Significantly different between groups, (p<0.05). Data for each individual are presented with those in the “Above” group represented by open circles (o) and those in the “Below” group represented by closed squares (∎). The bars identify the median value and the 25^th and 75^th percentiles of the median.. Spearman’s Rho correlations were performed to examine if visceral fat-loss at 24 weeks was related to the change in Total sRAGE (E), cRAGE (F), esRAGE (G), and cRAGE:esRAGE (H) at week 24. Correlations presented are for the entire cohort (correlations were not run in individual groups). Data for each individual is presented with those in the “Above” group represented by open circles (o) and those in the “Below” group are represented by closed squares (∎). Above: Individuals above the median visceral fat-loss by week 24, Below: Individuals below the median visceral fatloss by week 24.

We also examined effect of SAT-loss by stratifying our cohort by median SATloss SAT-loss (1.26 kg). Those above the median SAT-loss lost 3.8 (−7.5, −1.8) kg of SAT (p<0.001), and those below the median did not significantly alter their SAT −0.6 (1.0, 0.1) kg (p=0.091). The changes in esRAGE and cRAGE:esRAGE were significantly different between those above and below median SAT-loss (Figure 5). Those above median SAT-loss decreased their cRAGE:esRAGE (p=0.022) and those below the median decreased their esRAGE (p=0.043) (Figure 5). There was also a trend toward a significant reduction in total sRAGE in those below the median SAT-loss (p=0.053). The change in SAT was also negatively correlated to the change in esRAGE (rho=−0.38, p=0.026) and positively correlated to the change in cRAGE:esRAGE ratio (rho=0.35, p=0.042) (Figure 5 G and H).

Figure 5. Changes in sRAGE isoforms by median subcutaneous fat-loss.

To examine the effect of subcutaneous fat-loss on changes in sRAGE isoforms, the entire cohort was stratified by median subcutaneous fat-loss achieved at week 24. Subcutaneous fat mass was quantified via dual x-ray absorptiometry (DXA) at weeks 1 and 24 only. Changes in Total sRAGE (A), cRAGE (B), esRAGE (C), and cRAGE:esRAGE (D) at week 24 were compared between groups via Mann Whitney UTest. # Significantly different between groups, (p<0.05). Data for each individual are presented with those in the “Above” group represented by open circles (o) and those in the “Below” group represented by closed squares (∎). The bars identify the median value and the 25^th and 75^th percentiles of the median. Spearman’s Rho correlations were performed to examine if subcutaneous fat-loss at 24 weeks was related to the change in Total sRAGE (E), cRAGE (F), esRAGE (G), and cRAGE:esRAGE (H) at week 24. Correlations presented are for the entire cohort (correlations were not run in individual groups). Data for each individual is presented with those in the “Above” group represented by open circles (o) and those in the “Below” group are represented by closed squares (∎). Above: Individuals above the median subcutaneous fat-loss by week 24, Below: Individuals below the median subcutaneous fat-loss by week 24.

Relationships Between sRAGE Isoforms and Markers of Metabolic Health

We also examined if markers of metabolic health were related to circulating sRAGE isoforms. Baseline cRAGE:esRAGE ratio was negatively related to baseline IL6 (rho=−0.46, p=0.003). Moreover, TNF-α at W1 was negatively correlated with the change in esRAGE by W12 (rho=−0.40, p=0.025). Interestingly, we found that the change in total sRAGE was negatively correlated to the change in IL-6 (rho=−0.34, p=0.030) and leptin (rho=−0.34, p=0.038) and positively correlated with the change in adiponectin (rho=0.33, p=0.040) at W12. However, these relationships were lost at the W24 time point. The change in cRAGE:esRAGE ratio at W24 negatively correlated with the change in adiponectin at (rho=−0.35, p=0.027). Given our interesting findings demonstrating the relationship between weight and fat loss with esRAGE, we next developed a post-hoc step-wise regression model to determine if the change in esRAGE could by predicted the change (W1 – W24) in body weight, BMI, fat mass, subcutaneous fat and TNF alpha . These independent variables were chosen because we had previously demonstrated their linear correlation with the change in esRAGE. The only variable that was entered into the model and found to predict the change in esRAGE was the change in fat mass (β = −0.447, −16.4 to −2.8, p = 0.007). All other variables were excluded based on the probability of their F-statistic to result in statistical significance (p<0.05).

Discussion

To our knowledge, this is the first study to examine the effects of diet-induced weight-loss on sRAGE isoforms and the potential relationships between sRAGE and changes in specific fat depots and markers of metabolic health in individuals with obesity who are otherwise healthy. Obesity is a critical risk factor for the development of diabetes. RAGE expression is elevated in adipose tissue of individuals with obesity and has been demonstrated to play a role in adipocyte hypertrophy, insulin resistance and atherosclerosis in mice [5–8]. Conversely, sRAGE acts a competitive inhibitor of RAGE and is reduced in individuals with obesity and T2DM [14, 16, 17, 22, 27, 29, 30]. Individuals with obesity can reduce their risk for development of T2DM by losing weight however, the mechanisms by which weight-loss confers this benefit is not fully elucidated.

In the current study, the change in total sRAGE (60±85 pg/mL) was significantly different in individuals experiencing weight-loss at W24 compared to those who did not change their weight. Moreover, those who did not significantly reduce their weight reduced their total sRAGE (−159±54 pg/mL, p=0.004). In a cohort of 274 individuals spanning the glucose tolerance continuum, we previously published a proportional odds model using sRAGE isoforms along with age, gender, race and BMI to predict the odds of progression towards T2DM [14]. The model revealed a 9% increase in the odds towards T2DM for every 100 pg/mL decrease of total sRAGE [14]. Given our previously published model [14], the reduction in total sRAGE experienced by those in this study who did not lose weight translates to ~14% increased odds for T2DM development. These data suggest that weight-loss preserves circulating sRAGE which may explain some of the protective effects of weight-loss.

We also previously demonstrated an inverse relationship between obesity and sRAGE isoforms with esRAGE being the strongest correlate to BMI and body fat percentage [14, 27]. Here we corroborate those results with the change in esRAGE negatively correlating with the change in fat mass at W24 (rho=−0.47, p=0.002). In line with our current findings, Brix et al found a significant increase in total sRAGE as a result of weight-loss following bariatric surgery [11]. Our findings expand the existing literature by examining the change of all sRAGE isoforms following a highly popular dietary method (ADF) [23] and by characterizing their change in relation to the change in not only body weight but also VAT and SAT.

VAT is an inflammatory fat depot linked to metabolic disease risk [31, 32]. Additionally, RAGE has been shown to be more highly expressed in VAT of individuals with obesity compared to SAT [7]. In the context of obesity, SAT also becomes inflamed and insulin resistant following macrophage infiltration [33]. Adipose of RAGE-null mice fed HFD are protected from macrophage infiltration, impaired adipokine expression, and insulin resistance demonstrating a role for RAGE in this process [6]. Although we did not collect adipose tissue samples, we determined fat mass via DXA and used the gold standard MRI to determine VAT and subsequently derived SAT. Curiously, we found no statistically significant differences between the change in sRAGE isoforms between those Above and Below median VAT loss. This disconnect between changes in VAT and sRAGE are surprising given the role of VAT in inflammation and the AGE/RAGE axis [7]. However, post hoc analysis demonstrated a significant within group reduction of cRAGE:esRAGE ration in individuals above the median VAT-loss. These relatively limited findings may in part be due to the small magnitude of VAT-loss (0.6 kg) across the entire cohort. Comparatively, the loss of SAT in the individuals above the median SAT-loss was much more substantial (4.8 kg). The change in esRAGE was significantly different between those above and below the median fat-loss and SAT-loss. Individuals who lost a significant amount of fat mass increased their esRAGE whereas those who did not decreased their esRAGE and there was a similar trend for SAT-loss. However, the change in cRAGE between those above and below median fat- and SAT-loss were not different. These data suggest that esRAGE is particularly important in changes occurring in these fat depots with weight-loss and may give insight into mechanisms governing esRAGE production. With this in mind we developed a stepwise regression model to determine the variables that are able to predict the change in esRAGE. Our model corroborated the relationship between the change in fat mass and the change in esRAGE providing further evidence of the importance of fat loss in augmenting esRAGE appearance. By deriving a ratio between cRAGE and esRAGE, we can determine if the two isoforms change in the same proportion between groups. Because one factor in this ratio (cRAGE) is not differentially changing, we can assume that differences in cRAGE:esRAGE ratios between groups are largely due to changes in esRAGE. Together these data suggest that fat- and SAT-loss are related to esRAGE suggesting that these tissues are important in regulating esRAGE production. Given cRAGE and esRAGE are produced by different mechanisms, it is intuitive that they may be regulated by different stimuli.

Mechanisms of esRAGE production are largely unexplored. However, the splicing enhancer transformer-2 protein homolog beta (Tra2β) has been implicated to promote splicing of RAGE pre-mRNA into esRAGE, whereas the splicing repressor heterogeneous nuclear ribonucleoprotein A1 (hnRNPA1) directly antagonizes this event [34]. It is possible that these splicing mechanisms are dysregulated in obesity due to inflamed adipose tissue, which may lead to attenuated esRAGE production. However, more research is needed to test this hypothesis.

RAGE activation increases circulating inflammatory cytokines and attenuates anti-inflammatory adipokines such as adiponectin [6, 7, 33, 35]. In the context of obesity, IL-6 and leptin are regarded as inflammatory mediators [33]. We found that the change in total sRAGE was negatively correlated to changes in IL-6 (rho=−0.34, p<0.05) and leptin (rho=0.34, p<0.05) at week 12 respectively although these relationships were lost by week 24. Conversely, adiponectin possesses anti-inflammatory and insulinsensitizing effects, and is typically increased in response to reductions in fat mass via life-style interventions such as diet and exercise [32, 36]. However, we did not observe a significant effect of ADF on circulating total adiponectin, unlike Kelly et al who saw increases in circulating adiponectin with concomitant decreases in fat mass [32]. This discrepancy in findings may be due to our quantification of total adiponectin versus their quantification of high-molecular-weight adiponectin, which is believed to be the active isoform [32]. The intervention implemented by Kelly et al also consisted of a diet and aerobic exercise training which may help explain the difference in findings. Perhaps the weight-loss achieved by our cohort was not enough of a stimulus to result in significant changes in adiponectin or other circulating factors we measured.

Although ADF did not increase total circulating adiponectin, the change in adiponectin was positively correlated with the change in total sRAGE at W12 (rho=0.33, p<0.05). While this relationship was lost by week 24 of the intervention, at W24 the change in adiponectin was positively related to the change in cRAGE:esRAGE. This relationship between cRAGE:esRAGE ratio is particularly interesting as it suggests potentially independent regulation of the two isoforms following weight-loss. Given that adiponectin is an adipokine, its relationship with cRAGE:esRAGE further insists that changes in cRAGE and esRAGE are differentially regulated by changes in fat-loss. Future mechanistic work is required to determine if changes in adiponectin and the proportion of circulating sRAGE isoforms are causally linked.

This study provides unique insight into the regulation of circulating sRAGE isoforms following diet-induced weight-loss in individuals with obesity. The strengths of this study include our relatively large sample size (n=42), RCT design, gold standard MRI for VAT quantification, and determination of all sRAGE isoforms. However, this study is not without limitations, which include our inability to determine body fat at each time point, inability to provide meals for the entire 24 weeks, our relatively wide range of age (18–62 y), and BMI (23.6–46.3 kg/m²) in our cohort and the disproportionate number of women in this study (83% female). Previous work in 1,201 middle aged, overweight, non-diabetic individuals demonstrated that being male was accompanied by an odds ratio of 1.97 (1.42–2.73) for being in the lowest sRAGE quartile [19]. It is therefore not clear if our results translate to a sample that is more homogeneous in regards to sex distribution. Perhaps the largest limitation in the context of these findings was our inability to quantify circulating or tissue AGEs and AGEs in the diet. Previous reports have demonstrated a role for dietary AGEs in obesity and metabolic syndrome and which may also have had an impact on the outcomes in the current study as well [37]. In addition previous works have also shown that obesity increases circulating AGEs, and tissue AGEs in SAT and VAT which contribute to reduced adiponectin and the development of insulin resistance [7, 37, 38]. Future studies should determine if increases in sRAGE isoforms as a result of dietary weight-loss play a role in reducing both circulating and tissue AGEs. Additionally, because we did not collect adipose tissue we can only speculate on potential mechanisms and tissue sources of sRAGE isoforms. We are also unable to comment on the potential influence of genotype on our results. Specifically, the SNP at codon 82 influences sRAGE concentrations and [39]. Future studies should address these limitations to gain further insight into adipose tissue/RAGE biology. In conclusion, we report for the first time an association between body composition improvement and increased esRAGE following 24 weeks of an alternate day fasting diet.

• The control group significantly reduced their total sRAGE and cRAGE

• The alternate day fasting group was protected from reductions in sRAGE isoforms

• Those who lost more fat significantly increased their esRAGE

• Subcutaneous fat loss was negatively correlated to the change in esRAGE

• The change in cRAGE:esRAGE was negatively related to the change in adiponectin

Abbreviations:

ADF

Alternate Day Fasting

ADAM10

A Disitigrin And Metalloproteinase 10

AGE

Advanced Glycation End Products

CAD

Coronary Artery Disease

CI

Confidence Interval

CON

Control Group

cRAGE

Cleaved Receptor for Advanced Glycation End Products

esRAGE

Endogenous Secretory Receptor for Advanced Glycation End Products

hnRNPA1

Heterogeneous Nuclear Ribonuclear Protein A1

MRI

Magnetic Resonance Imaging

NF-κB

nuclear factor kappa-light-chain-enhancer of activated B cells

RAGE

Receptor for Advanced Glycation End Products

RCT

Randomized Control Trial

SAT

Subcutaneous Adipose Tissue

sRAGE

Soluble Receptor for Advanced Glycation End Products

T2DM

Type 2 Diabetes Mellitus

TRA2β

Transformer 2 Beta

VAT

Visceral Adipose Tissue

W1

Week 1

W12

Week 12

W24

Week 24