The synergistic effects of saxagliptin and metformin on CD34+ endothelial progenitor cells in early type 2 diabetes patients: a randomized clinical trial

42 adults with Type 2 Diabetes diagnosed within 10 years, currently on metformin (1000–2000 mg/day) were enrolled. Subjects were between 40 and 70 years of age, with a BMI of 25–39.9 kg/m², and a HbA1C between 6.0 and 9.0%. Additional inclusion/exclusion criteria can be found in Additional file 1: Appendix S1. This study consisted of a single site at The GW Medical Faculty Associates.

Once subjects signed the informed consent, and were found eligible, there was a 1 month “wash in” period, during which subjects adjusted their exercise level in order to achieve 150 min of moderate-intensity physical activity per week. Diet counseling was also provided. At visit 1, baseline values of the following measures were gathered: blood biochemistries, vitals, biophysical parameters, resting energy expenditure (REE), arterial stiffness measures, and endothelial progenitor cells analysis. Subjects were then randomized to one of two arms: saxagliptin 5 mg/day or placebo, in a blinded manner. Subjects took either saxagliptin (n = 21) or placebo (n = 21) for 12 weeks, while engaging in 150 min of moderate intensity physical activity per week. Visits were conducted every 6 weeks, ending at week 12 (visit 3).

Peripheral blood samples (approximately 60 ml) were drawn from patients and diluted in phosphate buffered saline (1:1). Mononuclear cells (MNCs) were then isolated from whole blood using a Ficoll density centrifugation method. MNCs were counted and an aliquot was used for CFU-Hill colony formation assay following the manufacturer’s instructions (Stem Cell Technologies, Vancouver, BC, Canada). At day 5 colony forming units (CFU) were counted. A fraction of MNCs were stained with (FITC, PE, APC)-conjugated antibodies (Miltenyi Biotec GmbH, Bergisch Gladbach, Germany) in order to analyze specific endothelial cell surface markers (CD34, CD31, CXCR4) by flow cytometry.

To isolate EPCs (CD34+), MNCs were magnetically sorted through a column after cells were stained with CD34 microbeads antibody (Miltenyi Biotec GmbH, Bergisch Gladbach, Germany). According to the manufacturer and based on flow cytometry analysis, the purity of CD34+ cells post sorting is 67% (before gating on white blood cells). An aliquot of CD34+ cells were then stained with trypan blue and counted using an Auto Cellometer Mini (Nexcelom Bioscience, Lawrence, MA).

CD34+ gene expression analysis was performed by quantitative reverse transcriptase polymerase chain reaction (qRT-PCR). CD34+ total mRNA was extracted and purified using the RNeasy mini kit (Qiagen). mRNA was then converted into cDNA by using the high capacity cDNA reverse transcription kit (Applied Biosystems). Possible gene expression changes promoted by Saxagliptin were assessed by a CFX96 real-time qPCR system (Bio-Rad) using TaqMan Universal Master Mix II (Applied Biosystems) and inventoried probes. The gene expression analysis included antioxidants, apoptosis, endothelial function, chemotaxis, inflammation, and endothelial lineage cell surface markers. The expression of individual gene was normalized to either housekeeping 18S or GAPDH and calculated by using the 2^−∆∆Ct method considering the difference in cycle threshold between visit 2 or visit 3 and baseline (visit 1). Gene expression of CD34− cell population was also analyzed along with CD34+ cells.

The migratory capacity of CD34+ was evaluated using the CytoSelect 24-well Cell Migration Assay kit (Cell Biolabs, Inc., San Diego, CA). Cells were suspended in serum-free media and seeded at 100,000 cells per insert. Migration of the cells through a 3 µm polycarbonate membrane to the wells containing serum-free media (control) and chemoattractant SDF-1α (10 or 100 ng/mL) was assessed after cells were kept overnight in a CO₂ incubator at 37 °C. Migratory cells were dissociated from the membrane and subsequently lysed and quantified by fluorescence (480 nm/530 nm) using CyQuant GR dye (Cell Biolabs, Inc., San Diego, CA). The fluorescence ratios between cells exposed to the chemotactic factor and cells exposed to chemoattractant-free media (control) along the visits were used to analyze the migratory capacity of the cells.

Arterial stiffness was assessed through pulse wave analysis (PWA), and pulse wave velocity (PWV). PWA was obtained from the radial artery while the subjects were seated at rest. Investigators tried to obtain a minimum of three measures, with an operator index score ≥ 80. PWA measures include: augmentation index (AI), Augmentation Index adjusted for a heart rate of 75 (AI-75), augmentation pressure (AP), and both systolic and diastolic blood pressures (SBP, DBP) measured both centrally and peripherally. PWV measures the velocity of the pulse as it moves from a “proximal” artery to a “distal” artery. The designated proximal artery was the carotid, however, occasionally the radial artery was used if no carotid measurement could be obtained. The designated distal artery was the femoral artery, with no alternative used. PWV was obtained with subjects supine, at rest. Investigators tried to obtain a minimum of two measures, each with a standard deviation of less than 10%. These measured were gathered using the AtCor SphygmoCor CP system.

Basal metabolic rate, otherwise known as resting energy expenditure (REE), was measured using the KORR REEVUE. Subjects were resting, sitting in an exam chair prior to beginning the test. Tests ran between 10 and 15 min. Values gathered include: Measured REE, Predicted REE, estimated total energy expenditure, VO2, and calories per day.

Body composition parameters were gathered both manually and using a Tanita BF-350 body composition scale. Manual measurements include: height, weight, BMI, waist circumference, hip circumference. The Tanita scale works via bio-impedance, and provides measures on: weight, BMI, percent body fat, fat mass (kg), fat free mass (kg), percent body water, water mass (kg), basal metabolic rate (kcal), daily calorie intake (kcal), and impedance.

A venous blood draw was performed for both biochemical analyses and serum ELISA. Standard of care laboratory measures were collected at each visit to monitor trends and changes. The following values were ordered either as plasma, serum, or whole blood through Laboratory Corporation of America: basic metabolic panel, lipid panel, leptin, HbA1C, C-reactive protein, IL-6, Adiponectin, and Insulin. ELISA was performed in order to analyze serum total GLP1 and SDF-1α. GLP-1 was analyzed using a competitive ELISA Immunoassay Kit (Raybiotech, Norcross, GA), and SDF-1α using a sandwich ELISA (EHCXCL12A, Thermo Scientific).

Vitals gathered were congruent with those gathered as a part of standard of care: Resting blood pressure, pulse, and temperature.

Finally, subject’s level of exercise exertion was measured using Actigraph wGT3X-BT activity monitors. Subjects were instructed to wear the Actigraphs during all waking hours for a total of 7 consecutive days. Subjects were provided dietary and exercise advice, as a part of the study. For exercise, all subjects were instructed at screening to achieve 150 min/week of moderate intensity physical activity, as per the ADA guidelines. Actigraphs served as a measure of this exercise compliance, and to verify for exercise as a confounding variable.

Variable distributions were examined using frequency histograms and outliers were excluded. In gene expression variables, outliers with expression values > 50 were dropped, and values were natural-log transformed due to skewness [using log(expression + 1)]. Expression values that still had outliers after log transformation were capped at a value of 2.

For each dependent variable (dv), we used a random effects mixed model with robust standard errors to estimate the group (saxagliptin/placebo) main effect, the visit (1, 2, 3) main effect, and the group x visit interaction. The main effect of group tells whether the saxagliptin patients vs controls differed on the dv independently of visit. The visit main effect tells us whether the dv mean changed across visits, independently of treatment group. The treatment x visit interaction was of primary interest, since this tells us whether the treatment and control groups had different slopes over visits, i.e. whether the pattern of scores over time differed based on treatment. Because there was a slight change in the laboratory procedure for CFU and migration outcome measurements after subject 20, we also considered this variation and included 3-way interaction tests (group x visit x early/late) in the model for these parameters. Since subjects were randomized to treatment, baseline subject variables that differed between treatments could not act as confounds (i.e. could not cause a spurious association between treatment and outcome, because they could not affect treatment assignment). Therefore, randomized controlled trials do not usually adjust for baseline differences. However, as a sensitivity analysis, for dependent variables with significant effects, we included as a covariate in the mixed model, any baseline variable that differed between treatment groups with p < 0.10.

SAS (Version 9.3 or 9.4, Cary, NC) was used for data analysis with p < 0.05 considered significant.

The study population was representative of subjects with uncontrolled type 2 diabetes, but with no preexisting macro-vascular complications. All adverse effects that occurred throughout the duration of the study were either not related to the study medication and design, or fell within the expected side effects profile for saxagliptin.

Table 2 shows blood biochemistries and arterial stiffness measures across the three visits.

Actigraph energy monitor use was analyzed to account for any exercise or activity level difference between the placebo and the saxagliptin treatment group. Post Actigraph analysis, we noted no difference between the two groups for the amount of hours per day spent in any level of physical activity intensity (sedentary, light, moderate, moderate-to-vigorous, or vigorous).

Body composition measures showed no statistically significant changes. A high correlative effect was seen in the Tanita body composition scale measures for fat free mass (FFM, p = 0.07) and percent body fat, (p = 0.08). Mean FFM, by visit, drops more from visit 1 to visit 2 in the saxagliptin group, than the control group at a trend level significance (Fig. 6a). Percent body fat, while being higher on average in the saxagliptin group than the placebo group, showed a decline after visit 2. Conversely, the placebo group had an increase after visit 2 (Fig. 6b). A trend was also noticed in total body water (TBW, p = 0.1), which was seen to decline from visit 1 to visit 2, and then a rise again at visit 3 in the saxagliptin group (Fig. 6c). In the control group, however, TBW remains stable until visit 2, when it begins to fall for visit 3.

Venous blood biochemistries were gathered both through Labcorp of America and through serum ELISA. Both standard of care, and research values were collected. In the Labcorp values, only adiponectin (p = 0.01) was statistically significant across the saxagliptin and placebo groups. The placebo group had a steady increase across visit 1 through 3, but the saxagliptin saw an overall stable value of adiponectin (Fig. 7a). Serum creatinine remained relatively stable throughout the study in the saxagliptin group, but showed a drastic decrease after visit 1 in the placebo group, before the values leveled out (p = 0.12, Fig. 7b). There was no correlating significant or trend level significant noticed in eGFR (p = 0.36), or BUN. There were no statistically significant changes between groups in either GLP1 or SDF-1α, performed by ELISA.

For basal metabolic rate, there were no values that showed any statistically significance, when compared to the control group for any of the parameters.

When analyzing the arterial stiffness, augmentation index that was adjusted for a heart rate of 75 (AI-75), was found to be statistically significant (p = 0.04), with the values for the control group increasing across visit 1 through 3. Saxagliptin subjects have a slight decline across the three visits (Fig. 8a). Systolic blood pressure, measured radially, was found to be statistically significant (Fig. 8b; p = 0.009). There was a trend level significant reduction for systolic blood pressure, measured arterially (Table 2; p = 0.061), however not all subjects had PWV measured, therefore lack of statistical difference (though there was a trend) between the groups may not be a true reflection.

In general, it was noted that adding HbA1c as a time-varying covariate had little impact on the results.

We have defined EPCs as CD34+ cells. Werner et al. defined EPCs as CD34/KDR positive cells [27]. However, if an investigator uses two markers to isolate EPCs from mononuclear cells (MNCs) rather than one, the percentage of positive cells drops by almost 10-fold. In this study, only 60–80 mls of whole blood such a drop makes a cell based clinical trial difficult to interpret. Therefore, we have used only one marker, CD34+ve, so that we get consistently close to 1% population of MNC, as EPCs.

In this study we show for the first time that saxagliptin, in addition to metformin, positively modulates CD34+ EPCs, as a marker of vascular endothelial function, in early onset type 2 diabetes subjects who have no overt cardiovascular complications. Vascular complications are common in subjects with diabetes and a reduced number of EPCs in these subjects can predict early onset of vascular complications [28‐31]. Therefore, in order to assess cardiovascular risk using a cell as a biomarker rather than serum biochemistry, it is important to investigate the number and functionality of endothelial progenitors and whether these cells can adequately promote or help with endothelial repair and angiogenesis [32]. Metformin, a very common medication for diabetes, has been shown to contribute to the improvement of EPCs in both type 1 and type 2 diabetes on its own and similarly, saxagliptin has been shown to increase EPC number and function [17, 22, 33]. Curiously, no differences in EPCs number were observed between the metformin and saxagliptin groups [17]. However, here we have demonstrated an additional improvement of EPCs by adding saxagliptin to on-going metformin treatment.

Based on the CFU-Hills’s colony count and migratory response to SDF-1α results (Figs. 1, 2, respectively), we demonstrated that saxagliptin enhances EPCs functionality. Considering that CFU-Hill’s colonies number is inversely related to Framingham risk score, these results also indicate a decrease in CVD risk promoted by the use of saxagliptin and metformin combination, in addition to exercise.

Our data also indicated a trend for an increase of CD34+ cell number for the saxagliptin group. On the other hand, other results recently published showed a significant increase of EPCs number (CD34+ CD133+ KDR+) when patients were subjected to either saxagliptin or metformin alone [17, 22, 33]. These results indicate that the combination of these two drugs may not promote additional increase in number of CD34+ cells in the periphery blood. It is possible that the subjects may have reached the maximum threshold already by using metformin. Metformin has a dual effect by simultaneously increasing VEGFA and reducing both CXCL10 and TIMP1 in CD34+ cells in a model of the diabetic state combined with hypoxia, and also shows a proangiogenic activity [34, 35].

Although we did not observe a statistically significant increase in CD34+ cells, the higher percentage of CD34+ CXCR4+ (double positive cell) for patients undergoing saxagliptin treatment was confirmed by our flow cytometry results (Fig. 3). We also noted increased CD31 positive cells that were CXCR4 positive (Fig. 4). CXCR4 is the receptor to SDF1α ligand. Similarly, other investigators have shown increase in the number of CD31+ cells as a positive outcome marker for cardiovascular risk assessment [36, 37]. It was noted that both CD34+ cells (a progenitor endothelial cell pool) and CD31+ cells (a mature circulating endothelial cell pool) had increased in double positivity with SDF1α receptor CXCR4 indicating positive effect of saxagliptin on endothelial lineage cells both immature and mature. It has been reported that CD34+ cells from patients with type 2 diabetes have defective chemotaxis to SDF-1α resulting in reduced vasculogenic potential [38]. Here, we showed that migratory response to SDF-1α of CD34+ cells is improved by saxagliptin. These results corroborate with the fact that CD34+ cells expressed more CXCR4 receptor and thereby facilitating the responsiveness to the SDF-1α ligand. Altogether, these results indicate that colony formation ability and mobilization of EPCs can be improved in patients with type 2 diabetes subjected to saxagliptin treatment. Similarly CD31+/CXCR4+ cell number also increased post saxagliptin therapy. While saxagliptin and metformin promoted similar beneficial effects on endothelial function when used separately, our results indicate a synergistic positive effect of saxagliptin to the entire endothelial lineage [17].

Effects of sitagliptin, another DPP-4 inhibitor, on EPCs have been previously reported [12, 13]. In type 2 diabetic patients, sitagliptin promoted increase in circulating EPCs followed by upregulation of SDF-1α [12]. In a mouse model, SDF-1α was also increased and in addition to increased EPCs number those results indicated that sitagliptin has a great potential as a promoter of neovascularization [13]. However in our study, saxagliptin therapy did not show an increase in SDF-1α plasma levels (measured by ELISA). There are still no data showing the minimum period of time for administration of saxagliptin in order to demonstrate an increase in SDF-1α levels. Therefore, our results could be secondary to a short-term intervention of 12 weeks and it is likely that plasma levels did not show increase as optimal time for detecting a statistical difference between placebo and saxagliptin. Interestingly, SDF-1α mRNA expression of EPCs was also not upregulated, in spite of increased CXCR4− expressing CD34+ cell number. In fact, there was a trend towards down-regulation of SDF-1α in the mRNA expression from cells that were exposed to saxagliptin, which may indicate that saxagliptin is rescuing the cells from damage and dysfunction, as SDF-1α is a factor that is primarily produced from damaged cells rather than healthy cells. Therefore the gene expression of SDF-1α ligand in healthy EPCs may actually decrease [29, 31]. The lack of effects of saxagliptin on SDF-1α plasma levels needs to be confirmed in larger and longer outcome studies.

Regarding gene expression, we looked at pathways involving inflammation, oxidative stress, apoptosis, EPC chemotaxis pathways and endothelial function markers, in both CD34+ and CD34− cells obtained post magnetic column separation of MNCs. The mRNA expression results that stand out (statistically significant) are the antioxidants such as SOD1 (superoxide dismutase) and GPX1 (glutathione peroxidase) with clear upregulation, post Saxagliptin exposure, compared to placebo. This is concurrent with reduced inflammatory and apoptosis cascade markers such as IL6 and Caspase-3 mRNA expressions in CD34− cells indicating an overall reduction in inflammation and apoptosis pathway, which could be secondary to upregulation in antioxidants. This is particularly interesting as there is sufficient evidence from our previous work, to suggest significant patho-physiological role of ROS (anti-oxidant expression may increase in response to increased intracellular ROS presence) in not only EPCs but also in MSCs in a setting of hyperglycemia, as seen in diabetes [10, 39]. The decrease of IGF-1 mRNA expression in CD34− cells in the saxagliptin intervention group compared to placebo may indicate reduced insulin resistance at the levels of MNCs, which also supports reduction of cellular inflammation.

All subjects were on a stable dose of metformin (1–2 grams/day) for 3 months or greater prior to enrollment. Subjects were instructed to engage in 150 min of moderate intensity physical activity per week, as per the American Diabetes Association guidelines, prior to being randomized into either the saxagliptin or placebo group. The Actigraph energy monitor data analysis showed no difference in activity across all exercise intensity levels between saxagliptin and placebo. This indicates that no changes for any outcome measure in the saxagliptin group can be attributed to exercise alone.

Saxagliptin is a weight neutral medication, and thus it was not expected that we would find a difference between groups in the weight [40]. There also was no change in waist or hip circumference measurements, which is consistent with other studies involving saxagliptin and linagliptin, although these studies did not have concomitant metformin therapy [17, 41]. Previous studies involving mice have shown that treatment with DPP-4 inhibitors in hyperglycemic, obese mice resulted in reductions in adiposity, both in body fat percentage, and abdominal fat mass [42]. This was attributed to an increase in energy expenditure, which was measured via monitoring metabolic rate and food intake. Our research subjects taking saxagliptin had reductions in both fat free mass (kg) and percent fat as compared to placebo, however, there was no difference between groups in resting energy expenditure values or exercise levels. The changes in fat mass, therefore, could likely be attributed to saxagliptin. The reduction in fat mass could be secondary to increased fat oxidation in lean and obese conditions [43]. Total body water showed a rise from visit 2 to 3 in the saxagliptin group, whereas the control group remained steady. One reason might be related to the increase of heart failure rate noticed in the SAVOR-TIMI trial, however this trial, unlike ours, involved a population with pre-existing CVD [44]. Interestingly, reduction in fat mass and increase in TBW would keep the weight of the subjects neutral as seen in most trials with DPP4 inhibitors, however this may not be clinically relevant in subjects with no overt heart failure [16].

Previous studies investigating DPP-IV inhibitor therapy, mainly sitagliptin and vildagliptin, in a type 2 diabetes population found that there was no significant reduction in HbA1C values with treatment [45]. This is consistent with our results, upon comparing the saxagliptin and placebo groups. In the saxagliptin group HbA1c decreased by 0.3% (less than expected), and the baseline HbA1c value was higher than placebo group. The difference in the baseline HbA1c value between groups and the little HbA1c reduction on the saxagliptin group is unlikely to explain the positive effect of saxagliptin on CD34+ EPC primarily based on HbA1C changes. Therefore, the positive changes that we see such as in migration response of EPCs to SDF1a and flow cytometry is unlikely to be secondary to improvement in glycemic indices. One study conducted by Li et al. found a significant improvement in HbA1C with saxagliptin therapy over 12 weeks, but these subjects were naïve to anti-diabetic therapy prior to entering the study [17]. Blood biochemistries, however, did find a significant difference both in adiponectin and serum creatinine between saxagliptin and placebo. While the meta-analysis conducted by Liu et al. found that treatment with DPP-4 inhibitors resulted in an increase in adiponectin, as compared to control, our results showed the opposite [46]. A reason for this may be that in our study we actually noticed a reduction in percent fat, and overall fat mass. We however did not see a decrease in leptin values.

We noted a stable serum creatinine in the saxagliptin group compared to the placebo group. The eGFR was not statistically significant (p = 0.36), which may be because the eGFRs for the placebo group were initially lower and remained stable throughout the study. On the other hand, despite creatinine starting higher, it improved significantly in the 1st phase of the study. Therefore, improvement in plasma creatinine may be an acute effect rather than persistent or chronic effect. This maybe an interesting finding considering the fact that diabetes related kidney disease is a progressive disease. This could be explained by the fact that saxagliptin most likely has a renal protective effect. A higher percentage of CD34+ CXCR4+ cells may which help by homing-in on the kidney vasculature to repair the endothelial dysfunction present and thereby help prevent progressive kidney damage [47, 48].

Arterial stiffness, is a measure of compliance of one’s arteries, and their ability to constrict and dilate in response to blood pressure [49‐51]. It is measured non-invasively by assessing PWV) and PWA, and has been noted to increase naturally with age [49]. PWV is measured as a velocity in m/s. Higher values of AP, AI, AI-75, and PWV are correlated to higher levels of arterial stiffness. PWV, in addition to PWA measures such as blood pressure and augmentation index, have been found to be a predictor of increased CVD risk in the general population, and especially in those at an increased risk, such as patients with type 2 diabetes [49, 51]. Arterial stiffness, being a direct measure of the radial, carotid and femoral arteries, would be expected to change with significant alterations to the endothelium.

Our study shows a reduction in arterial stiffness in the saxagliptin group, as seen through a reduction in AI-75. This was also seen with other DPP-4 inhibitors, sitagliptin and vildagliptin, which resulted in a reduction in AI-75 [45]. Arterial stiffness, as measured via AI-75, is a strong predictor of CVD in Type 2 Diabetes [52, 53]. The reduction in AI-75 may be attributed to a multi-platform effect. DPP-4 inhibitors cause an increase in systemic incretin levels, which can cause a relaxation of the arteries via nitric oxide (NO) [54, 55]. This could be attributed to a reduction in arterial stiffness. Also, the increase in CFU and higher percentage of CD34+ CXCR4+ that was reported in our cellular analysis may indicate that EPCs are having a regenerative effect on the subjects’ arteries across the 12-week time-period. Finally, DPP-4 inhibitors help patients achieve a better level of glycemic control, and, as observed from our body composition results, a decrease in both percent fat and fat free mass. This decline in adiposity may also contribute to the improvement (reduced PWA) in arterial stiffness.

ELISA was performed on patient serum samples to check for the presence of GLP1, and SDF-1α, in order to confirm that the saxagliptin was effectively blocking DPP-4 in the body. We were anticipating an increase in serum GLP1and SDF-1α in the saxagliptin group, indicating that DPP-4 was successfully inhibited. However, our results found no statistical significance between saxagliptin and placebo groups. It is likely that metformin leads to a significant increase of GLP-1 as seen in obese non diabetic and type 2 diabetes subjects, therefore the addition of saxagliptin would not show any further appreciable changes [56, 57]. Alternatively, it is possible that the dose of 5 mg of saxagliptin or duration of only 12 weeks was not sufficient to show a difference, or increase in values.

Limitations of our study may include the relatively short duration of 12 week saxagliptin therapy, which may have been inadequate to see significant changes in certain clinical and cellular parameters. This may have been because of the small sample size, and due to the difficulty in obtaining all measures, in some patients. Further studies with a larger population and longer duration would be helpful to further explore and consolidate the mechanisms behind our findings.