The Dipeptidyl Peptidase 4 Substrate CXCL12 Has Opposing Cardiac Effects in Young Mice and Aged Diabetic Mice Mediated by Ca²⁺ Flux and Phosphoinositide 3-Kinase γ

Historically overlooked, heart failure is now firmly recognized as being a major long-term complication of diabetes. On the one hand, diabetes confers an increased risk of heart failure (1). On the other hand, whereas certain medications used in the treatment of diabetes may reduce the risk of heart failure (2), others are neutral (3) and some actually increase heart failure risk (4). In the case of the dipeptidyl peptidase 4 (DPP-4) inhibitor class of antihyperglycemic medication, some DPP-4 inhibitors have demonstrated neutrality (e.g., sitagliptin) (3), whereas one (saxagliptin) was associated with an unexpected increase in the risk of hospitalization for heart failure (5) and another (alogliptin) was associated with a numerical, albeit nonsignificant, increase in heart failure risk (6). These observations are generally discordant with most preclinical studies that have reported overall favorable cardiac effects of various DPP-4 inhibitors (reviewed in Ussher and Drucker [7]).

DPP-4 inhibitors lower blood glucose levels by preventing the enzymatic degradation of the gut hormones glucagon-like peptide 1 (GLP-1) and gastric inhibitory polypeptide (GIP) by DPP-4, a serine exopeptidase. DPP-4 acts by catalyzing the hydrolysis of N-terminal dipeptides from polypeptides that contain either an alanine or a proline residue at position two. However, GLP-1 and GIP are not the only peptides that possess this amino acid configuration, and as a result, DPP-4 inhibitors have the capacity to increase the biological activity of a number of other peptide substrates. Although non–GLP-1 peptide substrates of DPP-4 may not regulate glucose homeostasis, some, such as substance P (8) and C-X-C motif chemokine ligand 12 (CXCL12, stromal cell-derived factor-1α [SDF-1α]) (9), may play important roles in cardiovascular (patho)physiology. The increased biological activity of these alternative substrates has been proposed as mediating some of the cardiovascular effects of DPP-4 inhibitors (10). However, in most cases, it has not been possible to discern whether the effects of DPP-4 inhibition on cardiovascular function are mediated by conventional GLP-1 receptor (GLP-1R)–dependent processes or by alternative DPP-4 substrates.

Here, we hypothesized that the cardiac effects of DPP-4 inhibition are mediated by mechanisms that do not involve classic GLP-1R–mediated events. During the course of our studies, we became mindful of the historical discordance between the effects of DPP-4 inhibition in experimental rodents and in clinical trials. Accordingly, we went on to compare the effects of DPP-4 substrates in the hearts of young adult mice, which are typically used in preclinical studies, with older obese diabetic mice that may better reflect the comorbid condition of the patient populations typically studied in cardiovascular outcome trials. Through this comparison, we discovered a dichotomous effect of the DPP-4 peptide substrate CXCL12, enhancing calcium (Ca²⁺) handling and improving contractility in young hearts but paradoxically impairing cardiac contractility in aged diabetic hearts in a phosphoinositide 3-kinase γ (PI3Kγ)–dependent manner.

Plasma DPP-4 activity was determined in wild-type (C57BL/6) mice treated with either normal chow or linagliptin (0.083 g/kg) (11) in chow (Harlan Laboratories Inc., Madison, WI), or in the hearts of diabetic high-fat diet–fed (DM-HFD) mice, as previously described (12).

GLP-1R^−/− mice were provided by Dr. Daniel J. Drucker at the University of Toronto (13). Male wild-type (C57BL/6) and GLP-1R knockout (GLP-1R^−/−) mice (aged 6–13 weeks) were bred at Charles River (Sulzfeld, Germany), and they were studied at St. Michael’s Hospital. Transverse aortic constriction (TAC) (or sham) surgery was conducted as previously described (14). In brief, a thoracotomy was performed in the second left intercostal space, the aorta was cleared distal to the subclavian artery, and a 7-0 silk suture was used to constrict the aortic arch. Sham surgery consisted of thoracotomy without placement of an aortic suture. After TAC surgery, mice were randomly allocated to receive either linagliptin in chow (0.083 g/kg) (11) or chow alone, and they were followed for 8 weeks. HbA_1c was determined using A1CNow+ (Roxon Medi-Tech Ltd., Etobicoke, Ontario, Canada). After 8 weeks, transthoracic echocardiography was performed using a high-frequency ultrasound system (Vevo 2100, MS-250 transducer; VisualSonics Inc., Toronto, Ontario, Canada), and linear dimensions were analyzed offline (Vevo 2100 software version 1.8) by a single investigator masked to the treatment groups. At sacrifice, mice were anesthetized, intubated, and artificially ventilated. Left ventricular (LV) cardiac catheterization was performed through the right internal carotid artery (Millar Mikro-Tip, 1.4F; AD Instruments, Colorado Springs, CO) (15). Data were acquired and analyzed using LabChart Pro (AD Instruments).

Hearts from wild-type (C57BL/6) mice were perfused using a retrograde isolated perfusion technique as described previously (16). For determination of the effects of DPP-4 peptide substrates ex vivo, Krebs-Henseleit buffer (KHB) was supplemented with GLP-1 (10 nmol/L; Bachem, Bubendorf, Switzerland) (17,18), GLP-2 (10 nmol/L; Prospec Technologies Inc., Mississauga, Ontario, Canada) (19), GIP (10 nmol/L; Phoenix Pharmaceuticals Inc., Burlingame, CA) (20), substance P (1 µmol/L; Tocris Bioscience, Bristol, U.K.) (21), or CXCL12 (25 nmol/L; Shenandoah Biotechnology Inc., Warwick, PA) (22) throughout the procedure. For inhibition of PI3Kγ, hearts from DM-HFD mice were perfused with IPI-549 (30 nmol/L; Chemietek, Indianapolis, IN), and contractility and heart rate parameters were measured after 20 min. The cellular IC₅₀ for IPI-549 is 1.2 nmol/L (vs. 250 nmol/L for PI3Kα, 240 nmol/L for PI3Kβ, and 180 nmol/L for PI3Kδ) (23). LV developed pressure (LVDP), maximum and minimum dP/dt, and heart rate were collected and analyzed using PowerLab 8/35 and LabChart Pro (AD Instruments).

Male C57BL/6 mice aged 8 weeks were obtained from Charles River Laboratories (Montreal, Quebec, Canada) and placed on an HFD (45% kcal fat, 35% kcal carbohydrate, and 0.05% weight for weight cholesterol; Research Diets Inc., New Brunswick, NJ). After 12 weeks, mice received a single intraperitoneal injection of streptozotocin (STZ; 90 mg/kg in 0.1 mol/L sodium citrate, pH 4.5). Diabetes was confirmed 2 weeks after STZ injection by blood glucose testing (One Touch UltraMini; LifeScan Canada Ltd., Burnaby, British Columbia, Canada), and mice that were not diabetic (blood glucose <12 mmol/L) received a second dose of STZ. Animals were followed for a total of 26 weeks before hearts were harvested for perfusion in the Langendorff mode.

All experimental procedures adhered to the guidelines of the Canadian Council on Animal Care and were approved by the St. Michael’s Hospital Animal Care Committee.

Cardiac myocyte hypertrophy was determined on hematoxylin and eosin (H-E)–stained sections as previously reported (15,24) and using a method adapted from Frustaci et al. (25). Nuclear volume was determined in 50 cardiomyocytes on each H-E–stained section using a method adapted from Gerdes et al. (26) in which nuclear volume is considered to be that of a prolate ellipsoid. Picrosirius Red–stained sections of the LV myocardium were scanned (Leica Microsystems Inc., Concord, Ontario, Canada) and quantified as the proportion of positive red staining in 10 randomly chosen fields (original magnification ×20) using ImageScope (Leica Microsystems Inc.).

SYBR green–based quantitative RT-PCR was performed on a ViiA7 Real-Time PCR System (Thermo Fisher Scientific, Rockford, IL). Primer sequences are listed in Supplementary Table 1. Data analysis was performed using Applied Biosystems Comparative C_T method.

Immunoblotting was performed with antibodies in the following concentrations: phosphorylated phospholamban (phospho-PLN) (Ser16) 1:1,000 (A285) (27), PLN (1D11) (28), SERCA2a 1:1,000 (IID8F6) (29), phospho-PLN (Thr17) 1:1,000 (sc-17024-R; Santa Cruz Biotechnology, Dallas, TX), p110γ 1:1,000 (sc-166365; Santa Cruz Biotechnology), and GAPDH 1:5,000 (sc-25778; Santa Cruz Biotechnology). Densitometry was performed using ImageJ version 1.39 (National Institutes of Health, Bethesda, MD).

Adult mouse cardiac myocytes were isolated according to the protocol developed by Ackers-Johnson et al. (30). Cells were loaded with 1 µmol/L Fura-2 AM (Thermo Fisher Scientific) and placed under an Olympus IX81 fluorescence microscope (Olympus, Tokyo, Japan) with an X-cite 120Q light source (EXFO, Quebec City, Quebec, Canada; Fura 2B filter set from Semrock, Rochester, NY). Fura-2 AM was excited using single-band excitation (387 nm), and emission was recorded with a single-band filter (510 nm). Images were acquired using a Rolera MGi plus EMCCD camera (QImaging, Surrey, British Columbia, Canada) and analyzed with MetaMorph version 7.6.6 software (MetaMorph, Nashville, TN). Cardiac myocytes were stimulated at 5 V with a frequency of 1 Hz. After taking control readings, cells were treated with 25 nmol/L CXCL12 (Shenandoah Biotechnology Inc.). Ca²⁺ transients were calculated from 200 frames taken over 1 min.

Phosphodiesterase (PDE) activity was determined in heart homogenates using a PDELight-HTS cAMP Phosphodiesterase Kit (LT07-600; Lonza Inc., Rockland, ME).

Data are expressed as mean ± SD. Statistical significance was determined by one-way ANOVA with a Fisher least significant difference test for comparison of multiple groups and two-tailed Student t test for comparison between two groups. Statistical analyses were performed using GraphPad Prism 6 for Mac OS X (GraphPad Software Inc., San Diego, CA). A P value <0.05 was considered statistically significant.

In our first experiments, we set out to determine whether the cardiac effects of DPP-4 inhibition are mediated through GLP-1R–dependent or GLP-1R–independent actions. LV hypertrophy (LVH) is a major determinant of impaired contractile function in type 2 diabetes (reviewed in Nassif and Kosiborod [31]), occurring in up to one-third of cases (32). Because mouse models of diabetes often do not develop LVH (33), we elected to examine the GLP-1R–independent effects of DPP-4 inhibition in mice subjected to TAC, a model of hypertrophic heart failure (34). We treated wild-type and GLP-1R^−/− mice with the DPP-4 inhibitor linagliptin, which at a dose of 0.083 g/kg resulted in a 77.5% reduction in plasma DPP-4 activity (plasma DPP-4 activity [arbitrary units (AU)]: control [n = 5] 640 ± 399, linagliptin [n = 6] 144 ± 52; P < 0.05). Eight weeks after TAC surgery, heart weight:tibial length and LV mass were increased equivalently in wild-type and GLP-1R^−/− mice (Supplementary Tables 2–5 and Fig. 1A and B). Linagliptin treatment had no effect on LV mass in either wild-type or GLP-1R^−/− mice (Fig. 1A and B). Heart weight:tibial length was also unaltered with DPP-4 inhibition, although it was marginally, but nonsignificantly, lower in linagliptin-treated TAC mice than TAC mice treated with vehicle (Supplementary Tables 1 and 2). Despite little change in cardiac hypertrophy, ejection fraction (Fig. 1C), fractional shortening (Fig. 1D), stroke volume (Fig. 1E), and cardiac output (Fig. 1F) were all higher in linagliptin-fed wild-type and linagliptin-fed GLP-1R^−/− TAC mice in comparison with TAC mice fed normal chow, with an equivalent increase in each of these measures irrespective of the presence or absence of GLP-1R (Fig. 1C–F). With respect to hemodynamic parameters, both end systolic pressure and peak systolic pressure were increased in both wild-type and GLP-1R^−/− TAC mice treated with vehicle or linagliptin and, like heart weight:tibial length, they were numerically but nonsignificantly lower with linagliptin treatment (Supplementary Tables 6 and 7). As we have observed previously (35), the indicator of diastolic function, τ, was unaltered in TAC mice (Supplementary Tables 6 and 7).

Paralleling the changes in LV mass, cardiomyocyte size, nuclear volume, and β-myosin heavy chain mRNA levels were increased equivalently in wild-type and GLP-1R^−/− TAC mice, without significant change with linagliptin treatment (Fig. 2A–D). Likewise, interstitial collagen deposition was also increased in wild-type and GLP-1R^−/− TAC mice and was unchanged with DPP-4 inhibition (Fig. 2E and F). In contrast, the preservation of cardiac contractile function in wild-type and GLP-1R^−/− TAC mice treated with linagliptin was accompanied by an increase in serine 16 phosphorylation of the Ca²⁺ handling protein PLN (phospho-PLN [Ser16]) (Fig. 2G), whereas total PLN and SERCA2a levels were unaltered (Fig. 2H and I).

The equivalent contractility-enhancing effects of linagliptin in both TAC wild-type and TAC GLP-1R^−/− mice suggested to us that at least some of the effects of DPP-4 inhibition are mediated through GLP-1R–independent actions, potentially non–GLP-1 substrates of DPP-4. To investigate this possibility further, we turned to an ex vivo isolated perfused heart system and we subjected mouse hearts to 20 min of no-flow ischemia followed by 40 min of reperfusion. We first studied the hearts of wild-type mice fed normal chow or linagliptin in chow. Baseline LVDP did not differ between the hearts of chow-fed mice and mice treated with linagliptin (Fig. 3A). However, there was an approximate doubling in LVDP with linagliptin after no-flow ischemia (Fig. 3A and B). Likewise, peak maximum and minimum dP/dt were also increased with linagliptin after ischemia reperfusion (Fig. 3C and D). Heart rate did not differ between the groups either at baseline or after ischemia reperfusion (Fig. 3E).

Next, we questioned whether the improvement in postischemic recovery with DPP-4 inhibition could be mimicked by any of the physiological peptide substrates of DPP-4, and we perfused the hearts of normal chow-fed wild-type mice with one of several different peptides, i.e., GLP-1, GLP-2, GIP, substance P, or CXCL12 (36). Forty minutes after 20 min of no-flow ischemia, LVDP was significantly increased in substance P–perfused hearts and CXCL12-perfused hearts, whereas GLP-1, GLP-2, and GIP had no effect (Fig. 4A and B). Peak maximum and minimum dP/dt and heart rate were marginally increased at baseline in GLP-1–perfused hearts, although only peak minimum dP/dt achieved statistical significance (Fig. 4C and D). Maximum dP/dt, minimum dP/dt, and heart rate did not differ significantly between the various experimental conditions after no-flow ischemia (Fig. 4C–E).

Because young mice and aged DM-HFD mice have recently been reported to exhibit discordant responses to DPP-4 knockout or inhibition (37), we next set out to determine whether the contractile effects of either substance P or CXCL12 would be different in the hearts of aged DM-HFD mice (mean body weight 44.4 ± 5.6 g, mean blood glucose 17.5 ± 5.0 mmol/L). Plasma DPP-4 activity was increased in DM-HFD mice (plasma DPP-4 activity [AU] 1102 ± 278 [n = 5], P < 0.01 vs. control). Baseline LVDP and peak maximum and minimum dP/dt did not differ between DM-HFD mice and age-matched nondiabetic normal chow-fed mice (control) (Fig. 5A–C). However, unlike the hearts of the young adult mice earlier studied, perfusion of DM-HFD hearts with either substance P or CXCL12 diminished baseline LVDP and maximum and minimum dP/dt (Fig. 5A–C). Heart rate tended to be lower in DM-HFD mouse hearts and was significantly lower than age-matched control hearts after perfusion with either substance P or CXCL12 (Fig. 5D). After no-flow ischemia, LVDP (Fig. 5A) and the percentage recovery of LVDP (Fig. 5E) were significantly lower in DM-HFD hearts than age-matched controls. Neither CXCL12 nor substance P improved postischemic recovery in DM-HFD hearts (Fig. 5E).

Next, we set out to determine whether DPP-4 inhibition and the DPP-4 substrates substance P and CXCL12 activated similar signaling pathways in the hearts of nondiabetic mice subjected to no-flow ischemia and whether these pathways are altered in DM-HFD hearts. Because we had observed an increase in PLN (Ser16) phosphorylation in the hearts of TAC mice treated with linagliptin, we focused our studies on Ca²⁺ regulatory protein changes. Like in TAC mice, linagliptin also increased PLN (Ser16) phosphorylation in isolated perfused hearts (Fig. 6A). Interestingly, hearts perfused with CXCL12 also exhibited increased PLN (Ser16) phosphorylation, whereas hearts perfused with substance P did not (Fig. 6A). Total PLN and SERCA2a levels were unaffected by any of the conditions (Fig. 6B and C). In contrast, hearts of DM-HFD mice had markedly diminished levels of phospho-PLN (Ser16), with no improvement with either CXCL12 or substance P (Fig. 6D). We queried whether this diminution in phospho-PLN (Ser16) could be compensated for by an increase in PLN threonine 17 phosphorylation (phospho-PLN [Thr17]) but similarly found a reduction in phospho-PLN (Thr17) in DM-HFD hearts (Supplementary Fig. 1). As in our earlier experiments in the hearts of younger, nondiabetic mice, total PLN and SERCA2a levels were unaltered in any of the experimental groups (Fig. 6E and F). Likewise, transcript levels of the substance P receptor (NK1 receptor) or the principal receptor for CXCL12 (CXCR4) were unaltered in DM-HFD hearts (NK1 receptor:RPL13a mRNA [AU]: control [age 8–13 weeks, n = 4] 1.1 ± 0.4, DM-HFD [age ∼8 months, n = 4] 1.1 ± 0.2; CXCR4:RPL13a mRNA [AU]: control 1.0 ± 0.1, DM-HFD 1.0 ± 0.1).

In our final series of experiments, we aimed to discern the mechanism(s) by which a non–GLP-1 DPP-4 substrate may have different effects in the hearts of young adult mice and aged obese diabetic mice. For these experiments, we focused on the actions of CXCL12 because, unlike substance P, CXCL12 mimicked the increase in PLN (Ser16) phosphorylation in the hearts of young mice caused by linagliptin treatment. In isolated adult mouse ventricular myocytes, recombinant CXCL12 increased PLN (Ser16) phosphorylation (Fig. 7A) and directly enhanced Ca²⁺ flux (Fig. 7B–F). In considering how CXCL12 may paradoxically impair cardiac function in the hearts of DM-HFD mice, we recognized the known role of PDEs in limiting PLN phosphorylation (38), and we found PDE activity to be significantly increased in the hearts of DM-HFD mice (log₁₀ PDE activity [AU]: control [n = 7, age 6–13 weeks] 2.3 ± 0.5, DM-HFD [n = 12] 2.7 ± 0.3; P < 0.05). We observed a general trend toward an increase in PDE isoform transcript abundance in DM-HFD hearts, mRNA levels of the PDE4A isoform being significantly increased (twofold) (Supplementary Table 8). Moreover, the increase in PDE activity in DM-HFD mouse hearts was accompanied by a marked upregulation in protein levels of p110γ (Fig. 8A), the catalytic subunit of the lipid kinase PI3Kγ, which is known to regulate PDE activity in cardiomyocytes (39). Because PI3Kγ is activated by G protein–coupled receptors (GPCRs) (including CXCR4) (40) and because PI3Kγ has also been linked to impaired cardiac contractile function (41), we hypothesized that the deleterious effects of CXCL12 in DM-HFD mice were mediated through the actions of upregulated PI3Kγ. Consistent with this hypothesis, when we perfused the hearts of DM-HFD mice with the PI3Kγ inhibitor IPI-549 concurrently with CXCL12, we found that PI3Kγ inhibition negated the deleterious effects of CXCL12 on basal LVDP (Fig. 8B–E).

Despite the widespread adoption of DPP-4 inhibitors into the glucose-lowering armamentarium for the treatment of type 2 diabetes, several questions remain unanswered regarding their cardiac effects and the peptide substrates they affect. In the current study, we observed that DPP-4 inhibition with linagliptin preserved cardiac contractility in pressure-overloaded mouse hearts even in the absence of GLP-1R, and it also improved contractile recovery of mouse hearts ex vivo. We went on to discover that this latter effect of linagliptin could be mimicked by the non–GLP-1 DPP-4 substrate CXCL12, which enhances cardiomyocyte Ca²⁺ flux. However, in the setting of aging, obesity, and diabetes, CXCL12 paradoxically worsened cardiac function, an effect that appeared to be mediated by the lipid kinase PI3Kγ. Collectively, the findings highlight the extent to which peptide substrates of DPP-4, other than GLP-1, can affect cardiac function, and they demonstrate how these effects are profoundly influenced by diabetic comorbidity.

In our first studies, we treated pressure-overloaded wild-type and GLP-1R^−/− mice with the DPP-4 inhibitor linagliptin. We chose a pressure overload model because of the importance of LVH as a predeterminant of heart failure in diabetes (31). The dosing regimen of linagliptin that we selected (0.083 g/kg) resulted in a 77.5% reduction in plasma DPP-4 activity, just marginally lower than that observed with the 5-mg dose of linagliptin in patients (80–90%) (42). Both wild-type and GLP-1R^−/− mice exhibited an equivalent reduction in ejection fraction, fractional shortening, stroke volume, and cardiac output after pressure overload. However, linagliptin-fed mice demonstrated a notable preservation of systolic function. The preservation of systolic function with linagliptin is similar to that previously reported with the DPP-4 inhibitor vildagliptin (43). However, whereas the cardioprotective effects of vildagliptin were attributed to improved glucose tolerance and increased GLP-1 levels (43), linagliptin was equally efficacious in both wild-type and GLP-1R^−/− mice. This led us to conclude that DPP-4 inhibition has GLP-1–independent cardiac effects, and we speculated that these effects are most likely mediated by the actions of one or more of the other peptide substrates of DPP-4. To help us discern the similarities between the cardiac effects of DPP-4 inhibition and the cardiac effects of individual DPP-4 peptide substrates, we chose an isolated perfused heart system. Such an approach allowed us to study the direct cardiac effects of individual peptide substrates of DPP-4 independent of any actions on the central or autonomic system or on preload or afterload.

We examined the effects of five physiological substrates of DPP-4 on LVDP recovery, GLP-1, GLP-2, GIP, substance P, and CXCL12 (44), and we observed that either substance P or CXCL12 improved LVDP recovery, comparable to linagliptin, whereas the other peptides did not. In a nonreductionist sense, it is not surprising that the effects of linagliptin on the recovery of contractile function could be mimicked by more than one peptide substrate, and indeed substance P (21) and CXCL12 (45) have both been reported to improve contractile recovery after ischemia reperfusion in other studies. GLP-1 has also been previously reported to improve the recovery of cardiac contractile function (46), although we did not observe this in the current study. It is noteworthy, however, that we selected the concentrations of the different peptide substrates with which to perfuse mouse hearts based on previously reported experiments (in the absence of data regarding bioequivalence), and thus a direct comparison between the magnitude of the effect of different peptides should be avoided.

At this point, our cognizance of the absence of cardiovascular benefit in outcome trials of other DPP-4 inhibitors reported to date led us to question whether the cardiac effects of either substance P or CXCL12 would be affected by diabetic comorbidity. To explore this possibility, we studied the effects of the peptides in the hearts of aged obese diabetic mice. These mice had elevated plasma DPP-4 activity and an impairment of LVDP recovery. The increase in plasma DPP-4 activity is similar to that previously reported in patients with type 2 diabetes (47). We observed that substance P or CXCL12 paradoxically impaired basal ex vivo LVDP in DM-HFD mice, and we surmised that this observation was indicative of the persistent biological activity of the peptides even in the setting of increased plasma DPP-4 activity. We speculated that the antithetical effects of substance P and CXCL12 in the hearts of young nondiabetic mice and aged obese diabetic mice could provide insights into the relative importance of the signaling pathways the peptides activate, and we probed for changes in the proteins that regulate cardiomyocyte Ca²⁺ cycling, the signal that regulates cardiac muscle contractility.

The affinity of SERCA2a for Ca²⁺ is regulated by PLN. In its unphosphorylated state, PLN limits SERCA2a Ca²⁺ affinity, restricting cardiac contractility, whereas when it is phosphorylated, PLN no longer restricts SERCA2a Ca²⁺ affinity, resulting in positive inotropic and lusitropic effects (48). PLN can be phosphorylated at two sites, Ser16 (which is mediated by protein kinase A) and Thr17 (which is mediated by Ca²⁺/calmodulin-dependent protein kinase II [CaMKII]) (49). We found that PLN (Ser16) phosphorylation was increased in the hearts of wild-type and GLP-1R^−/− TAC mice treated with linagliptin and in the hearts of linagliptin-fed mice subjected to no-flow ischemia. CXCL12 mimicked this effect of DPP-4 inhibition in increasing phospho-PLN (Ser16) levels but substance P did not. However, in the hearts of aged obese diabetic mice, PLN (Ser16) phosphorylation levels were diminished, and they were unaltered by either CXCL12 or substance P. This decrease in PLN (Ser16) phosphorylation coincided with an increase in PDE activity, which itself is regulated by the lipid kinase PI3Kγ (39).

PI3Kγ is a class IB PI3K that, unlike class 1A PI3Ks, is activated by GPCRs. PI3Kγ consists of a catalytic p110γ subunit tightly associated with an accessory or adaptor subunit, p101 (50). Upon GPCR activation, the Gβγ complex separates from the Gα subunit, allowing interaction between p110γ and the Gβγ complex and resulting in p110γ translocation to the membrane and activation of its kinase activity. p110γ is expressed by several cardiac cell types, including cardiomyocytes (41,51). PI3Kγ activity is increased in human failing hearts (52), and in the current study, its protein levels were increased in the hearts of aged DM-HFD mice. Loss of p110γ increases cardiomyocyte contractility (41) through cAMP-dependent mechanisms (53) and its activation in response to β-adrenergic receptor stimulation predisposes to myocardial hypertrophy and cardiac fibrosis (51). We hypothesized that the impairment in LVDP in the CXCL12-perfused hearts of DM-HFD mice was due to PI3Kγ upregulation, and we found that inhibition of PI3Kγ negated the deleterious effects of CXCL12 on basal LVDP.

A number of limitations of the experiments herein described are worth emphasizing. First, we studied the effects of linagliptin in vivo in TAC mice and the effects of linagliptin and the peptides ex vivo in hearts subjected to ischemia-reperfusion injury, reflecting our evolving understanding of pathobiological mechanisms during the course of our studies. Accordingly, it should be appreciated that the pathological processes occurring in the two model systems can be quite different. However, it is interesting that linagliptin treatment improved the contractility of the hearts of the young nondiabetic mice in both settings. Likewise, linagliptin (as well as the GLP-1R agonist liraglutide) has been reported to improve recovery after myocardial ischemia in mice in vivo (54). Second, although the hearts of linagliptin-treated mice had improved contractile function that could be mimicked by certain DPP-4 substrates, other than GLP-1, this does not itself demonstrate that these substrates are responsible for the effects of DPP-4 inhibition in vivo or ex vivo. Commercial antibodies do not typically distinguish between active and inactive forms of CXCL12, and an increase in active CXCL12 either systemically or locally within the hearts of linagliptin-treated mice has not been demonstrated. Moreover, even if the cardiac effects of DPP-4 inhibition in mice are mediated through increased CXCL12 activity, the actions of endogenously produced active CXCL12 may be different from those of the recombinant peptide. Third, DPP-4 inhibition does not raise the bioactivity of a single substrate in isolation, and the combined effects of multiple peptides may be quite different from the effects of any one of them alone. Treatment with the DPP-4 inhibitor MK-0626 impaired cardiac function in aged DM-HFD mice subjected to TAC (37). However, a definitive understanding of the cardiovascular effects specifically of linagliptin in patients with type 2 diabetes will be provided by close examination of the results of the Cardiovascular Outcome Study of Linagliptin Versus Glimepiride in Patients With Type 2 Diabetes (CAROLINA) and Cardiovascular and Renal Microvascular Outcome Study With Linagliptin in Patients With Type 2 Diabetes Mellitus (CARMELINA) clinical trials. Notwithstanding these limitations, the present experiments do reveal important biological insights that may be helpful in the development of new treatment approaches in the future. They also complement, without contradicting, a recently proposed hypothesis that in some settings, DPP-4 substrates (including CXCL12) can have adverse cardiac effects (55).

In summary, DPP-4 inhibition has effects on cardiac function in mice that extend beyond those mediated by the GLP-1R. Some of these effects can be mimicked by peptide substrates of DPP-4, other than GLP-1, in particular CXCL12, which has opposing actions in the hearts of young adult mice and in aged DM-HFD mice. Whereas CXCL12 enhances cardiomyocyte Ca²⁺ handling, it can also impair cardiac contractility in the setting of aging, obesity, and diabetes, the latter effect occurring through PI3Kγ-dependent mechanisms. The cardiac effects of DPP-4 substrates depend upon the metabolic context in which they act.

Acknowledgments. The authors thank Youan Liu and Suzanne L. Advani (Keenan Research Centre for Biomedical Research and Li Ka Shing Knowledge Institute, St. Michael’s Hospital, Toronto, Ontario, Canada) for technical assistance.

Funding. S.N.B. was supported by a Keenan Family Foundation KRESCENT Postdoctoral Fellowship through the Kidney Foundation of Canada, a Heart and Stroke/Richard Lewar Center of Excellence Fellowship Award, and a Banting & Best Diabetes Centre Hugh Sellers Postdoctoral Fellowship. K.T. was supported by a Research Internship Abroad from the São Paulo Research Foundation (Fapesp 2016/04591-1). F.H.Z. was supported by a Queen Elizabeth II Graduate Scholarship in Science and Technology and an Ontario Graduate Scholarship. T.A.A. is supported by a King Abdullah Foreign Scholarship. M.J.H. is a recipient of a scholarship from the Research Training Centre of St. Michael’s Hospital and a Banting & Best Diabetes Centre–Novo Nordisk Studentship. S.M. was supported by a Diabetes Canada Postdoctoral Fellowship. A.O.G. was supported by the Heart and Stroke Foundation of Ontario (T-6281). K.A.C. is a recipient of a Canadian Institutes of Health Research New Investigator Award and an Early Researcher Award from the Ministry of Research, Innovation and Science, Ontario. A.A. is a recipient of a Diabetes Investigator Award from Diabetes Canada.

Duality of Interest. These studies were supported by a grant from Boehringer Ingelheim. K.A.C. and A.A. are named as inventors on a patent application by Boehringer Ingelheim for the use of DPP-4 inhibition in the treatment of heart failure. K.A.C. reports speaker honoraria from Amgen, AstraZeneca, Boehringer Ingelheim, Eli Lilly and Company, Janssen, and Merck and has received research grant support from Amgen, AstraZeneca, Boehringer Ingelheim, Servier, Merck, and Eli Lilly and Company. A.A. has received research support from Boehringer Ingelheim and AstraZeneca. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. S.N.B. designed and performed the experiments, analyzed the data, and wrote the manuscript. K.T., F.H.Z., T.A.A., V.G.Y., M.J.H., S.M., and D.L. performed the experiments and analyzed the data. M.G.K. and B.B.B. performed the in vivo experiments. A.O.G. supervised the in vitro experiments. K.A.C. designed the experiments and oversaw the analysis of in vivo assessment of cardiac function. A.A. designed the experiments, supervised the study, and wrote the manuscript. A.A. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this work were presented at the 76th and 77th Scientific Sessions of the American Diabetes Association, New Orleans, LA, 10–14 June 2016, and San Diego, CA, 9–13 June 2017, respectively.