Serum Amyloid A Truncations in Type 2 Diabetes Mellitus

Serum Amyloid A (SAA) is an acute phase protein complex consisting of several abundant isoforms. The N- terminus of SAA is critical to its function in amyloid formation. SAA is frequently truncated, either missing an arginine or an arginine-serine dipeptide, resulting in isoforms that may influence the capacity to form amyloid. However, the relative abundance of truncated SAA in diabetes and chronic kidney disease is not known. Methods Using mass spectrometric immunoassay, the abundance of SAA truncations relative to the native variants was examined in plasma of 91 participants with type 2 diabetes and chronic kidney disease and 69 participants without diabetes. Results The ratio of SAA 1.1 (missing N-terminal arginine) to native SAA 1.1 was lower in diabetics compared to non-diabetics (p = 0.004), and in males compared to females (p<0.001). This ratio was negatively correlated with glycated hemoglobin (r = −0.32, p<0.001) and triglyceride concentrations (r = −0.37, p<0.001), and positively correlated with HDL cholesterol concentrations (r = 0.32, p<0.001). Conclusion The relative abundance of the N-terminal arginine truncation of SAA1.1 is significantly decreased in diabetes and negatively correlates with measures of glycemic and lipid control.


Introduction
Serum Amyloid A (SAA) proteins are members of the acute phase response protein group that play a significant role in inflammatory processes [1]. Structurally, proteins in the SAA complex immediately frozen at −80°C for all other assays. Demographic information (age, sex, ethnicity), physical exam measurements (blood pressure, waist circumference, weight, height, body mass index), medication use, and medical history (hypertension, hyperlipidemia, smoking, type and duration of diabetes) were also recorded. The glomerular filtration rate (GFR) was estimated using the Modification of Diet in Renal Disease Study (MDRD) equation [23]. Assignment of CKD (stages [1][2][3][4][5] was based on GFR levels as described [24]. Exclusion criteria included the following: type 1 diabetes, participation in an active weight loss program, history of cancer, HIV, active infection, other ongoing serious illness or current steroid use. Diabetes classification was based on clinical and medication history, or glycated hemoglobin greater than 6%. For the SAA MSIA analysis, frozen plasma samples were briefly thawed on ice, centrifuged for 5 min at 3000 rpm, and 50 mL aliquoted in low profile 96-well trays. Prior to the analysis, samples were diluted with 100 mL PBS containing 0.1%Tween.

Mass spectrometric immunoassay affinity pipette derivatization
The first step in MSIA method development is derivatization of the affinity pipettes with corresponding antibody toward the targeted protein. Using a Multimek 96 pipettor (Beckman Coulter, Brea, CA ) affinity pipettes were derivatized with antibody toward SAA by initial rinse with 200 mM HCl (2 × 20 cycles, each cycle consisting of an aspiration and dispense of a 120 mL volume), followed by an acetone rinse (1×20 cycles). The surface of the microcolumns was activated by immersing the pipettes into a tray containing 100 mg/mL 1, 1'carbonyldiimidazole (in N-Methylpyrrolidinone), and 450 cycles of 100 mL aspirations and dispenses through each of the affinity pipettes were performed. Two rinses with NMP (10 cycles each, 150 mL volume) followed. The affinity pipettes were then immersed into a microwell plate containing the antibody solution (5 mg Ab per tip) and 750 cycles of aspirations and dispenses of 50 mL volume were performed to bind the antibody to the activated surfaces. One rinse with 60 mM HCl followed (50 cycles each, 100 mL), ending with two final rinses with assay buffer (PBS w/0.1% Tween, 50 cycles each, 100 mL). Antibody-derivatized pipettes were stored at +4°C until used.

Mass spectrometric immunoassay
Prior to sample protein extraction, the derivatized affinity pipettes were pre-rinsed with assay buffer (PBS, 0.1%Tween, 50 aspiration/dispense cycles, 100 mL each), followed by two 60 mM HCL rinses (50 aspiration/dispense cycles, 100 mL each) and another assay buffer rinse. Based on prior experiments, pre-rinsing reduces non-specific binding to the tips during protein extraction. Pipettes were then immersed into a microplate containing the samples and 250 aspirations and dispense cycles were performed (100 mL each) allowing for affinity capture of SAA from the samples. The pipettes were then rinsed with mixture of 1 M ammonium acetate and acetonitrile (AA: ACN = 3:1 (v/v)) (100 cycles), and twice with water (10 cycles and 20 cycles respectively, 100 mL aspiration/dispense each). SAA-loaded tips were then exposed to six-microliter aliquots of MALDI matrix solution (25 g/L sinapic acid in aqueous solution containing 33% (v/v) acetonitrile, and 0.4% (v/v) trifluoroacetic acid). After a 10 second delay (to allow for the dissociation of the protein from the capturing antibody), the eluates were dispensed directly onto a 96-well formatted MALDI target. Following drying, linear-mode mass spectra were acquired from each sample spot, each consisting of ten thousand laser shots using an Autoflex III MALDI-TOF mass spectrometer (Bruker, Billerica, MA). The mass spectra were internally calibrated using low molecular mass protein standards, and further processed (baseline subtracted and smoothed) with the Flex Analysis software (Bruker Daltonics). Antibody specificity was verified by confirming the absence of SAA-related mass spectral peaks from negative control tips on which 1) no antibody was present, and 2) on which monoclonal antibodies specific to other proteins were immobilized. The peak areas for all SAA signals were measured in Zebra software (Beavis Informatics, Ltd.). Relative percent abundance of every SAA isoform was calculated in correspondence to the total SAA in the corresponding mass spectra.

SAA quantification
Total SAA concentrations were measured using a commercial ELISA from Invitrogen Life technologies (Novex, Cat. KHA0012) that has been extensively used [25][26][27][28][29] and tested on serum, plasma and tissue culture samples. The kit is based on antibody affinity to SAA 1.1, and thus does not capture the other forms of SAA.

Statistical Analysis
Mean (SD) or median (25th and 75th percentiles) were calculated for continuous variables. The diabetes and non-diabetes groups were compared by independent t-tests (normally distributed variables) or Wilcoxon rank sum tests (non-normally distributed variables). Categorical variables were compared using the chi-square test. The relationship between SAA variants and diabetes status was measured by logistic regression, with diabetes status as the binary dependent variable. The relation between SAA truncations and clinical variables was modeled using linear regression; all regression models used log-transformed SAA 1.1R as the dependent variable, and included age, BMI, and GFR as adjusting covariates. Regression models were performed in the total sample and by gender; a product interaction term tested whether the associations of clinical variables with SAA variants differed in males vs. females. Statistical analyses used SAS version 9.3 software package; a p-value of < 0.007 (0.05 divided by 7) was used to define significance levels in the SAA variants between the diabetes and non-diabetes groups accounting for the seven SAA variants (SAA 1.1R, SAA 1.1RS, SAA 2.1R, SAA 2.1 RS, SAA 2.2R, SAA 1.3R and SAA 1.3 RS truncations) of interest in this study. All other statistical tests used a p-value of 0.05.

Results
Participants' demographic and biochemical characteristics are summarized in Table 1. Participants with diabetes were on average older, had greater BMI, and demonstrated several metabolic characteristics of type 2 diabetes, including low HDL cholesterol, increased fasting insulin and elevated CRP levels when compared to participants without diabetes. In addition, participants with diabetes had a lower GFR and a greater urine albumin excretion (p<0.01 for both) compared with the non-diabetic controls. Males (in the combined group) had significantly lower GFR than females (mean AE SD: male: 85.6 AE 32.5 L/min/1.73m 2 ; female: 98.4 AE 35.6 L/min/1.73m 2 , p = 0.03).
Characteristics of known SAA variants are summarized in Table 2. The SAA variants that can be resolved by MSIA are summarized in Table 3. A total of four protein isoforms, SAA 1.1, SAA 1.3, SAA 2.1 and SAA 2.2 (reflecting protein products from SAA1 and SAA2 genes) were detected among the samples, expressed either as single or as multiple variants. MALDI-MSIA-mass spectra of SAA from five different plasma samples showing all the variant forms and truncations (missing terminal -R and/or terminal -RS) are shown in Fig. 1. All samples that expressed SAA 1.1, presented with R and RS truncations. The MSIA analysis confirmed that the SAA variant 1.1 (MW = 11682.68) was the most abundant and it was detected in all but one (159/160) individual. The distribution of the other identified SAA protein entities within the cohort was as follows: 58.8% for SAA 2.1, 13.1% for SAA 2.2 and 11.9% for SAA 1.3. In addition to these native forms, two N-terminal truncation isoforms were noted for all the variants: those missing arginine (des-R; Δm = −156.19 Da) and/or arginine-serine dipeptide (des-RS; Δm = −243.27 Da).
The relative abundance of each of the truncated isoforms was calculated as the ratio of truncated SAA to native SAA and compared between subjects with and without diabetes (SAA variants R and RS ratios). The log of SAA1.1R ratio was higher in the non-diabetic group compared to the diabetic group, (p = 0.02). The log of SAA1.1R was significantly associated with age (r = 0.16, p = 0.03), but not with BMI (r = −0.12, p = 0.11). As shown in Table 3, after adjustment for age, SAA 1.1R was lower in the diabetes group compared with the non-diabetes group (p = 0.004); the association remained significant with adjustment for both BMI and age (p = 0.02). In contrast, the ratio of other SAA truncation variants to native (SAA 1.1RS, 2.1R, 2.1RS, 2.2R, 1.3R and 1.3 RS truncations) did not differ between the diabetes and non-diabetes groups (all p>0.05, Table 3). In addition, in the whole group (diabetes and non-diabetes combined) the SAA 1.1R ratio to native was lower in males compared with females (median (IQR): male: n = 76, 0.78 (0.29); female: n = 84, 0.97 (0.43), p<0.001), Fig. 2). An inverse correlation was observed between glucose levels and log-transformed SAA 1.1R (r = −0.3, p<0.001). The correlation between log-transformed SAA1.1R and glucose was significant in males (r = −0.44, p<0.001, Fig. 3B) but not in females (r = −0.15, p = 0.17, Fig. 3C). Table 4 summarizes the linear associations of log-transformed SAA 1.1R ratio and several metabolic measures in males and females, adjusting for age, BMI and GFR. Both fasting glucose and glycated hemoglobin were strongly inversely correlated with log-transformed SAA 1.1R ratio. Although none of these associations significantly differed in males and females (all interaction p-values > 0.05), the SAA 1.1R ratio was most strongly and inversely correlated with glucose levels in males. There was no correlation between total SAA concentrations measured by ELISA and the SAA1.1R ratio determined by MSIA.
Since males had a lower GFR than females, the relative abundance of SAA truncations could be explained by differences in kidney clearance rate. To address this possibility, the relationship between log-transformed SAA 1.1R and GFR in both males and females was examined. In males, log-transformed SAA 1.1R was not significantly correlated with GFR (r = 0.21, p = 0.09); the association was reduced with adjustment for glucose (r = 0.12, p = 0.37). In females, log-transformed SAA1.1R and GFR were inversely but not significantly correlated (r = −0.221, p = 0.06); adjustment for glucose did not alter this association (r = −0.21, p = 0.08). These glucose-adjusted associations did not significantly differ among males and females (p-value for interaction = 0.09).
The study group was categorized by CKD stage with the non-CKD group defined based on a GFR > 90 L/min/1.73m 2 . The distribution of CKD stages in the diabetes and nondiabetes groups is presented in Table 5. There were 61 patients in the CKD group and 80 patients in the non-CKD group. The majority of non-diabetic participants did not have evidence of CKD, whereas the diabetic participants recruited in this study had different stages of CKD with 9 participants in stages 4 and 5. The inverse association of log-transformed SAA1.1R with glucose did not differ among persons with CKD vs. non-CKD (interaction p-value = 0.12, adjusted for age and BMI). These findings confirm that glucose levels had an inverse correlation with SAA 1.1R, with the association apparent in persons with and without CKD.
The peak area of SAA 1.1 was highly correlated to the peak area of SAA 1.1R and SAA 1.1RS ( Table 6). The correlation was strongest (r> 0.9, p <0.001, all subjects) among the SAA 1.1 R and RS variants suggesting that the same process favors the formation of both truncations.

Discussion
Our findings demonstrate that ratios of SAA 1.1R truncations to their native SAA variants are decreased in diabetes. The occurrence of N-terminally truncated SAA has been observed in several studies confirming that the relative abundance of this variant of SAA is quite high doi:10.1371/journal.pone.0115320.g003 [3,12,15,16,30,31]. The population frequency of these truncations and their physiologic relevance are not known. Since the N terminus of SAA determines its amyloid fibril formation activity [17,18], assessing changes in SAA truncations might provide insight into diseases such as reactive amyloidosis where susceptibility to amyloid formation is dependent on the SAA functionality and clearance. We previously reported a lower relative abundance of SAA truncations in plasma of one subject with acute inflammation (rheumatoid arthritis) compared with subjects without evidence of inflammation [3]. Given that the N-terminal arginine and serine of SAA are highly conserved, these truncations are likely the result of posttranslational modifications [12,32]. Dipeptidyl aminopeptidase cleavage has been reported for several lipoproteins [33], and SAA is an apolipoprotein that forms HDL. In our study, glucose levels had the strongest correlation (inverse relationship) with the ratio of truncations in males, and this persisted after adjusting for changes in kidney function, age or BMI. This inverse relationship with hyperglycemia suggests that glucose might impair the peptidase activity. In one previous study, the clearance of truncated SAA at the N-terminus differed from native SAA suggesting that these truncations may impact SAA turnover [16].
Our findings also suggest that sex may independently modify the activity of this peptidase, leading to a lower truncated SAA to native SAA ratio. Of note, male sex was previously identified as a risk factor for both amyloidosis and diabetes complications. In a study of Familial Mediterranean Fever, the male to female ratio was significantly higher (2:1) in patients with reactive amyloidosis, [34]. In the Pittsburgh Epidemiology of Diabetes Complications Study, males had worse diabetic complications after long term follow up [35]. The role of these truncations in chronic diabetic complications merits further study. We measured total SAA concentrations by a validated commercial ELISA [25][26][27][28][29]. Total SAA concentrations did not differ between the diabetic and non-diabetic groups and SAA concentrations did not correlate with the ratio of SAA1.1R ratio. Although a previous study [36] demonstrated that SAA concentrations were increased in type 2 diabetes with chronic kidney disease, we did not find increased total SAA concentrations in persons with diabetes.
The relative abundance of the SAA RS truncation is lower than that of SAA R, and there has been speculation that it might be a biomarker of certain diseases, such as renal cancer [12]. However, the strong correlation of R and RS truncations with the unmodified SAA 1.1 variant suggest that the same process affects the formation of both of these truncations in patients with and without diabetes ( Table 6). This may argue against RS truncation as a selective marker of diabetes complications.
It is worth noting that MSIA cannot distinguish SAA form 1.5 from 1.1 because their molecular weight differs by only 1 Da, which makes these forms difficult to differentiate from each other on most mass spectrometers. The expression frequencies of SAA 1.4 and 1.5 are not known, but based on previous genetic studies, they are likely uncommon variants [37]. Given the lower relative abundance of SAA 2.1, 2.2 and 1.3 truncations, this study could be underpowered to detect differences in these variants between the diabetes and non-diabetes groups.
In conclusion, our findings indicate that truncated SAA is less common in diabetes. SAA truncations may modify SAA function and clearance. Measuring SAA truncations may help us better understand the susceptibility to reactive amyloidosis in chronic conditions.