Volume 19, Issue 4 pp. 1050-1060
ORIGINAL ARTICLE
Free Access

Circulating delta-like Notch ligand 1 is correlated with cardiac allograft vasculopathy and suppressed in heart transplant recipients on everolimus-based immunosuppression

Hilde M. Norum

Corresponding Author

Hilde M. Norum

Research Institute of Internal Medicine, Oslo University Hospital Rikshospitalet, Oslo, Norway

Institute of Clinical Medicine, Medical Faculty, University of Oslo, Oslo, Norway

Division of Emergencies and Critical Care, Department for Research and Development, Oslo University Hospital, Oslo, Norway

Correspondence

Hilde Margrethe Norum, Research Institute of Internal Medicine, Oslo University Hospital Rikshospitalet, Oslo, Norway.

Email: [email protected]

Search for more papers by this author
Annika E. Michelsen

Annika E. Michelsen

Research Institute of Internal Medicine, Oslo University Hospital Rikshospitalet, Oslo, Norway

Institute of Clinical Medicine, Medical Faculty, University of Oslo, Oslo, Norway

Search for more papers by this author
Tove Lekva

Tove Lekva

Research Institute of Internal Medicine, Oslo University Hospital Rikshospitalet, Oslo, Norway

Search for more papers by this author
Satish Arora

Satish Arora

Department of Cardiology, Oslo University Hospital Rikshospitalet, Oslo, Norway

Center for Heart Failure Research, Medical Faculty, University of Oslo, Oslo, Norway

Search for more papers by this author
Kari Otterdal

Kari Otterdal

Research Institute of Internal Medicine, Oslo University Hospital Rikshospitalet, Oslo, Norway

Search for more papers by this author
Maria Belland Olsen

Maria Belland Olsen

Research Institute of Internal Medicine, Oslo University Hospital Rikshospitalet, Oslo, Norway

Search for more papers by this author
Xiang Yi Kong

Xiang Yi Kong

Research Institute of Internal Medicine, Oslo University Hospital Rikshospitalet, Oslo, Norway

Institute of Clinical Medicine, Medical Faculty, University of Oslo, Oslo, Norway

Search for more papers by this author
Einar Gude

Einar Gude

Department of Cardiology, Oslo University Hospital Rikshospitalet, Oslo, Norway

Search for more papers by this author
Arne K. Andreassen

Arne K. Andreassen

Department of Cardiology, Oslo University Hospital Rikshospitalet, Oslo, Norway

Search for more papers by this author
Dag Solbu

Dag Solbu

Novartis Norge AS, Oslo, Norway

Search for more papers by this author
Kristjan Karason

Kristjan Karason

Sahlgrenska University Hospital, Transplant Institute, Gothenburg, Sweden

Institute of Medicine, University of Gothenburg, Gothenburg, Sweden

Search for more papers by this author
Göran Dellgren

Göran Dellgren

Sahlgrenska University Hospital, Transplant Institute, Gothenburg, Sweden

Department of Cardiothoracic Surgery, Sahlgrenska University Hospital, University of Gothenburg, Gothenburg, Sweden

Department of Molecular and Clinical Medicine, Institute of Medicine, Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden

Search for more papers by this author
Lars Gullestad

Lars Gullestad

Institute of Clinical Medicine, Medical Faculty, University of Oslo, Oslo, Norway

Department of Cardiology, Oslo University Hospital Rikshospitalet, Oslo, Norway

Search for more papers by this author
Pål Aukrust

Pål Aukrust

Research Institute of Internal Medicine, Oslo University Hospital Rikshospitalet, Oslo, Norway

Institute of Clinical Medicine, Medical Faculty, University of Oslo, Oslo, Norway

Section of Clinical Immunology and Infectious Diseases, Oslo University Hospital Rikshospitalet, Oslo, Norway

Search for more papers by this author
Thor Ueland

Thor Ueland

Research Institute of Internal Medicine, Oslo University Hospital Rikshospitalet, Oslo, Norway

Institute of Clinical Medicine, Medical Faculty, University of Oslo, Oslo, Norway

K.G. Jebsen TREC, University of Tromsø, Tromsø, Norway

Search for more papers by this author
First published: 12 October 2018
Citations: 10

Abstract

Cardiac allograft vasculopathy (CAV) causes heart failure after heart transplantation (HTx), but its pathogenesis is incompletely understood. Notch signaling, possibly modulated by everolimus (EVR), is essential for processes involved in CAV. We hypothesized that circulating Notch ligands would be dysregulated after HTx. We studied circulating delta-like Notch ligand 1 (DLL1) and periostin (POSTN) and CAV in de novo HTx recipients (n = 70) randomized to standard or EVR-based, calcineurin inhibitor-free immunosuppression and in maintenance HTx recipients (n = 41). Compared to healthy controls, plasma DLL1 and POSTN were elevated in de novo (P < .01; P < .001) and maintenance HTx recipients (P < .001; P < .01). Use of EVR was associated with a treatment effect for DLL1. For de novo HTx recipients, a change in DLL1 correlated with a change in CAV at 1 (P = .021) and 3 years (P = .005). In vitro, activation of T cells increased DLL1 secretion, attenuated by EVR. In vitro data suggest that also endothelial cells and vascular smooth muscle cells (VSMCs) could contribute to circulating DLL1. Immunostaining of myocardial specimens showed colocalization of DLL1 with T cells, endothelial cells, and VSMCs. Our findings suggest a role of DLL1 in CAV progression, and that the beneficial effect of EVR on CAV could reflect a suppressive effect on DLL1.

Trial registration numbers—SCHEDULE trial: ClinicalTrials.gov NCT01266148; NOCTET trial: ClinicalTrials.gov NCT00377962.

Abbreviations

  • ANOVA
  • analysis of variance
  • BPAR
  • biopsy-proven acute rejection
  • CAV
  • cardiac allograft vasculopathy
  • CNI
  • calcineurin inhibitor
  • CsA
  • cyclosporine A
  • DLK1
  • delta-like 1 homologue
  • DLL1
  • delta-like Notch ligand 1
  • EC
  • endothelial cell
  • ECM
  • extracellular matrix
  • EVR
  • everolimus
  • HF
  • heart failure
  • hsCRP
  • high-sensitivity C-reactive protein
  • HTx
  • heart transplantation
  • IVUS
  • intravascular ultrasound
  • LVEDD
  • left ventricular end-diastolic diameter
  • MCP-1
  • monocyte chemoattractant protein 1
  • MIT
  • maximal intimal thickness
  • MMF
  • mycophenolate mofetil
  • mTORC
  • mTOR complex
  • mTOR
  • mammalian target of rapamycin
  • NF
  • nuclear factor
  • NT-proBNP
  • N, terminal pro-B-type natriuretic peptide
  • PAV
  • percent atheroma volume
  • PBMC
  • peripheral blood mononuclear cell
  • POSTN
  • periostin
  • sCD25
  • soluble CD25
  • SIR
  • sirolimus
  • TNF
  • tumor necrosis factor
  • VSMC
  • vascular smooth muscle cell
  • 1 INTRODUCTION

    Heart transplantation (HTx) remains the definitive treatment for advanced heart failure (HF), offering a median survival after HTx of more than >13 years.1 Within 5 years, however, approximately 30% of the patients develop cardiac allograft vasculopathy (CAV), an important contributor to adverse outcome after HTx.2 Clinically, CAV resembles coronary heart disease, but it is pathologically distinct from the usual coronary atherosclerosis and may affect any graft vessel, with concentric, progressive thickening along the length of the vessels that ultimately leads to ischemia, HF, and death.3 Multiple factors contribute to CAV development, including immunologic (eg, chronic rejection process), and non–immune-related responses (eg, hyperlipidemia and viral infection), but the pathogenesis is incompletely understood.

    Everolimus (EVR) is an alternative or supplement to calcineurin inhibitor (CNI)–based immunosuppressive regimen after HTx and reduces the incidence and progression of CAV.4-6 EVR, a mammalian target of rapamycin (mTOR) inhibitor, has antiproliferative effects on lymphocytes, vascular smooth muscle cells (VSMCs), endothelial cells (ECs), and fibroblasts7 and attenuates the activation and inflammatory response of neutrophils in vitro.8 However, the clinical benefit of EVR in HTx correlates poorly with changes in circulating inflammatory and vascular markers, and the mechanism by which EVR slows CAV progression remains unrevealed.5

    The Notch signaling pathway is essential for communication between neighboring cells9 and processes known to contribute to CAV,10 like EC activation,11 accumulation and activation of immune cells,12 proliferation of VSMCs, extracellular matrix (ECM) remodeling13, 14 and T cell activation.15 Activation or inhibition of Notch signaling depends on the configuration and interactions of canonical Notch ligands.16 Proteolysis induces secretion of extracellular domains, which together with a large group of noncanonical ligands may modulate Notch signaling.17 The canonical delta-like Notch ligand 1 (DLL1) regulates the cell fate of monocytes,18 macrophages,19 T cells,20 and ECs,21 and in HF, circulating DLL1 is elevated and associated with impaired cardiac function and adverse outcome.22, 23 To our knowledge, DLL1 is the only detectable canonical Notch ligand in human plasma/serum. For noncanonical Notch ligands, delta-like 1 homologue (DLK1) is the most studied and has inhibitory effects on angiogenesis.24 Periostin (POSTN), a matricellular protein functioning as a noncanonical Notch ligand,25 is regulated in clinical myocardial injury,26 and is associated with cardiac dysfunction in end-stage HF.23 POSTN may modify Notch signaling through inhibition of DLK127 or maintenance of Notch1 receptor expression.28

    Based on the involvement of Notch signaling in processes central to the development and progression of HF and CAV, and potential modulation of Notch signaling by mTOR inhibition,29 we hypothesized that (1) circulating Notch ligands would be dysregulated in HTx recipients, (2) EVR would affect levels of circulating Notch ligands in HTx recipients, and (3) circulating levels of Notch ligands would be associated with CAV progression. We tested our hypothesis by measuring the Notch ligands DLL1, POSTN, and DLK1 in plasma from 2 cohorts of HTx recipients with EVR-based and standard immunosuppression, at different timepoints and accompanied by assessment of CAV.

    2 MATERIALS AND METHODS

    2.1 Patient population

    Two adult HTx cohorts were included: (1) de novo HTx recipients from the Scandinavian heart transplant everolimus de novo study with early CNI avoidance (SCHEDULE)4 and (2) maintenance HTx recipients from the NOrdic Certican Trial in HEart and Lung Transplantation (NOCTET).30

    The SCHEDULE trial (ClinicalTrials.gov NCT01266148) was a 1-year, controlled and randomized multicenter trial and has been reported in detail.4 In brief the trial was conducted in 5Scandinavian HTx centers to evaluate if early initiation of EVR and early cessation of the CNI cyclosporine A (CsA) could improve renal function and hamper the progression of CAV. After antithymocyte globulin induction therapy, de novo HTx recipients were randomized to an immunosuppressive regimen consisting of either (1) low-dose EVR, CsA, mycophenolate mofetil (MMF), and corticosteroids with CsA withdrawal, and stepwise increments in EVR dose up to full dose after 7-11 weeks (hereafter: Everolimus group, EVR); or (2) standard regimen with CsA, MMF, and corticosteroids (hereafter: CNI group, CNI). The 2regimens were initiated no later than the fifth postoperative day. After the 1-year trial, immunosuppression was according to the investigator's preference, and the patients were reassessed at 3 years post-HTx.31 A total of 70 patients (CNI, n = 34; EVR, n = 36) with plasma collected at ≥3 timepoints: 7-11 weeks (i.e., baseline), 6 months, 1 year, and 3 years post-HTx, were included in the present follow-up substudy.

    The NOCTET trial (ClinicalTrials.gov: NCT00377962) was a 1-year, controlled and randomized multicenter study conducted in 5Scandinavian transplant centers and reported in detail earlier.30 In brief; maintenance HTx recipients with renal impairment were randomized to either (1) EVR + reduced CNI (hereafter: Everolimus group, EVR) or (2) standard CNI-based regimen for maintenance immunosuppression ≥1 year post-HTx (hereafter: CNI group, CNI). The aim of the study was to evaluate whether the introduction of EVR with a simultaneous protocol-specified reduction in CNI exposure would improve renal function. After the 1-year core trial, the study was extended for 1 year and patients were reassessed at their annual visit 2 years after enrollment.32 A total of 41 patients (CNI, n = 22; EVR, n = 19) (mean 4.7 years, range 0.9-16.9 years post-HTx) with plasma collected at ≥3 timepoints: on inclusion (i.e., baseline), 6 weeks and 1 and 2 years after inclusion, were included in the present follow-up substudy.

    For comparison, 20 age- and sex-matched healthy subjects were included.

    For both studies, written informed consent was obtained after institutional review board approval. The studies were carried out in accordance with the ICH Harmonized Tripartite Guidelines for Good Clinical Practice, applicable local regulations, and the Declaration of Helsinki.

    2.2 T cell isolation and activation

    Peripheral blood mononuclear cells (PBMCs) were obtained from heparin-anticoagulated blood from healthy donors by Isopaque-Ficoll (Lymphoprep; Axis-Shield, Oslo, Norway) gradient centrifugation.33 Isolation of T cells from PBMCs was performed by negative selection using Pan T cell isolation kit and MACS Separator (Miltenyi Biotec, Bergisch Gladbach, Germany). Isolated T cells were resuspended in AimV medium (Gibco, ThermoFisherScientific, Waltham, MA), seeded into a 48-well plate (Costra, Cambridge, MA; 250.000 cells/well), and incubated for 40 minutes in a humidified CO2 incubator at 37°C before adding EVR (Sigma-Aldrich, St. Louis, MO), sirolimus (SIR) (Rapamycin, InvivoGen, San Diego, CA), or CsA (Sigma-Aldrich) in different concentrations. After 90 minutes of incubation, the cells were activated by beads coated with antibodies to CD2, CD3, and CD28 (T cell activation/expansion kit, Miltenyi Biotec). Cell-free supernatants and cell pellets were harvested at 24, 48, and 72 hours and stored at −80°C.

    3 RESULTS

    3.1 Patient population

    Baseline demographics and characteristics were comparable between the 2treatment groups of de novo HTx recipients (SCHEDULE population), except for measured glomerular filtration rate, high-sensitivity C-reactive protein (hsCRP), and total cholesterol that were higher, and uric acid and terminal pro-B-type natriuretic peptide (NT-proBNP) that were lower in the EVR group (Table 1). For maintenance HTx recipients (NOCTET population), no differences in baseline demographics and characteristics were found (Table 1). For comparison, 20 healthy control subjects were recruited. Controls (mean age ± standard deviation [SD] 59.4 ± 8.8 years) were of similar age as the NOCTET population (58.9 ± 9.3), but not significantly older than the SCHEDULE population (50.5 ± 13.7, P = .11). Furthermore, creatinine in the controls was lower (76 ± 11 μmol/L) compared to both SCHEDULE (111 ± 41 μmol/L, P < .001) and NOCTET (121 ± 26 μmol/L, P < .001). Women represented 35% of controls, which was not statistically different from SCHEDULE (24%) or NOCTET (12%) (chi-square P = .11).

    Table 1. Patient demographics and baseline characteristics
    SCHEDULE NOCTET
    CNI (n = 34) Everolimus (n = 36) P CNI (n = 22) Everolimus (n = 19) P
    Recipient demographics
    Age(y) 49.7 ± 13.7 51.4 ± 14.0 .612 60.9 ± 7.8 56.6 ± 10.5 .142
    Female gender (%) 8 (24%) 9 (25%) .886 3 (14%) 2 (11%) .762
    BMI (kg/m2) 25.3 ± 3.4 23.7 ± 3.5 .080 25.5 ± 4.6 27.1 ± 6.0 .339
    sBP (mm Hg) 134 ± 15 134 ± 20 .844 144 ± 15 145 ± 17 .940
    dBP (mm Hg) 82 ± 9 80 ± 12 .394 90 ± 10 90 ± 14 .928
    HR (beats/min) 84 ± 12 92 ± 11 .008 82 ± 11 88 ± 10 .099
    Medical history
    Diabetes (%) 6 (18%) 3 (8%) .245 4 (18%) 1 (5%) .207
    Hypertension (%) 5 (15%) 6 (17%) .822 15 (68%) 16 (84%) .233
    Current smoker (%) 18 (53%) 17 (47%) .632 12 (55%) 13 (68%) .364
    Time since HTx (y) 4.4 ± 3.4 5.1 ± 4.0 .517
    Biochemistry
    Hgb (g/dl) 11.4 ± 1.1 11.1 ± 1.4 .409 13.6 ± 1.2 14.2 ± 1.3 .109
    mGFR (mL/min) 55 ± 15 67 ± 18 .003 58 ± 12 62 ± 15 .389
    Creat (μmol/L) 118 ± 32 105 ± 48 .190 121 ± 26 119 ± 20 .787
    hsCRP (mg/L) 4 (2,8) 6 (3,22) .035 2.4 (0.9,3.4) 2.7 (1.0,3.9) .708
    Uric acid (μmol/L) 421 (334,454) 299 (253,382) <.001 452 (387,520) 450 (397,508) .972
    Cholesterol (mmol/L) 4.9 ± 0.9 5.5 ± 1.2 .042 5.2 ± 0.7 5.3 ± 0.7 .654
    NT-proBNP (pg/mL) 155 (102,247) 94 (54,204) .041 50 (27, 118) 32 (22, 58) .202
    Donor age (y) 44 ± 13 40 ± 15 .182 35 ± 12 36 ± 13 .927
    Female donor 9 (30%) 14 (42%) .283 8 (36%) 8 (42%) .707
    • Categoric data are reported as n (%) and continuous data as mean ± SD and median and interquartile range depending on distribution.
    • sBP, systolic blood pressure, dBP, diastolic blood pressure; HR, heart rate; Hgb, hemoglobin; mGFR, measured glomerular filtration rate; Creat, creatinine; hsCRP, high-sensitivity C-reactive protein; NT-proBNP, N-terminal pro-B-type natriuretic peptide.

    3.2 Notch ligands in plasma at baseline

    Compared to healthy controls and adjusting for differences in age and creatinine, baseline plasma DLL1 levels were higher in de novo (SCHEDULE) (P < .01) and in maintenance (NOCTET) (P < .001) HTx recipients. In SCHEDULE, DLL1 levels correlated with creatinine (r = −0.45, P < .001), hsCRP (r = 0.36, P = .004), NT-proBNP (r = 0.38, P = .004), and uric acid (r = 0.53, P < .001). Similar, but more modest associations were observed for baseline DLL1 in NOCTET (creatinine r = −0.27, P = .088; hsCRP r = 0.28, P = .078; uric acid r = 0.33, P = .042, and hemoglobin r = 0.32, P = .044). In addition, DLL1 levels at baseline in NOCTET were correlated with time since HTx (r = 0.35, P = .026) (Figure 1).

    Details are in the caption following the image
    Circulating Notch ligands at baseline in the SCHEDULE (n = 70, 7-11 weeks post heart transplant) and NOCTET (n = 41, mean time since heart transplant: 4.7 years) trials and healthy controls (CTR, n = 20). Data are given as median and interquartile range. **P < .01, *** P < .001. P-values adjusted for age and creatinine using analysis of covariance (ANCOVA) with Bonferroni-adjusted post hoc tests

    Baseline POSTN levels were also higher in both groups of HTx recipients compared to controls, but in contrast to DLL1, the highest levels of POSTN were seen in de novo HTx recipients (SCHEDULE). No significant correlations between POSTN and baseline biochemical variables were observed in SCHEDULE. For maintenance HTx recipients (NOCTET), baseline POSTN levels correlated with uric acid (r = 0.39, P = .014) but not with time since HTx (r = 0.28, P = .085) (Figure 1).

    For DLK1, no significant differences were found (Figure 1).

    3.3 Effect of EVR on Notch ligands in plasma

    A significant overall treatment effect was observed in the temporal course of circulating DLL1 for each study population (Figure 2). In de novo HTx recipients (SCHEDULE) plasma levels of DLL1 decreased after 6 months of EVR treatment and remained low compared to baseline levels, but above those of healthy controls, throughout the study. Comparing the change in DLL1 after adjusting for baseline DLL1 levels and change in kidney function (ie, creatinine) in the same period revealed a significantly larger decrease in the EVR group at all timepoints. Among maintenance HTx recipients (NOCTET), a similar, but less distinct, pattern was observed with a significant treatment effect at 1-year follow-up.

    Details are in the caption following the image
    Circulating Notch ligands during long-term follow-up in the SCHEDULE and NOCTET trials. Data are given as estimated marginal means and 95% confidence intervals (CIs). The P-value indicates the group effect of treatment from repeated measures ANOVA using baseline levels as a covariate. The gray shaded areas represent the geometric mean (gray line) and 95% CI for healthy controls, n = 20. *P < .05, **P < .01, ***P < .001 vs baseline (BL). P < .05 comparing change at the indicated timepoint between standard treatment (CNI) and everolimus adjusting for baseline levels and change in creatinine in the same period [Color figure can be viewed at wileyonlinelibrary.com]

    No treatment effects were observed for POSTN or DLK1 (Figure 2). POSTN levels declined after commencing on either of the immunosuppressive regimens in de novo HTx (SCHEDULE) recipients and remained low, but above those of healthy controls, for both treatment groups compared to baseline (Figure 2). In contrast, POSTN increased compared to baseline at all subsequent timepoints in both treatment groups in maintenance HTx recipients (NOCTET) (Figure 2). Plasma DLK1 remained stable and comparable to levels in healthy controls in both treatment arms and study populations (Figure 2).

    Whereas DLL1 displayed similar profiles in SCHEDULE and NOCTET, POSTN exhibited a marked early decline in SCHEDULE, independent of study arm. The level of POSTN at 1 and 3 years in SCHEDULE was similar to that in NOCTET levels (ie, median 4.7 years post-HTx), and it seems that at 6 months, “steady state” levels of POSTN are achieved. The reason for this pattern is unclear, but because POSTN is induced during tissue injury and during ECM remodeling,34 we suggest that the initial high POSTN levels after HTx reflect ongoing ECM remodeling.

    3.4 Notch ligands and acute rejections

    Similar to the original report,31 there was a trend association between biopsy-proven acute rejection (BPAR) and the use of EVR (32% vs 13%, P = .056), and the use of CNI, respectively. As depicted in Table S1, we observed, however, no difference in change in DLL1 between patients with and without BPAR, although there was a trend toward lower DLL1 levels in patients who experienced BPAR within the first year (P = .090). No differences in change in or 1-year levels of DLL1 in relation to BPAR were observed within the treatment arms. For the other ligands, the decline in POSTN during 1-year follow-up was larger in those who experienced BPAR.

    3.5 DLL1 and CAV development

    Recently, we reported attenuated CAV development in the EVR group of de novo HTx recipients (SCHEDULE).5 Herein, we observed a treatment effect for DLL1. We therefore evaluated the relationship between change in DLL1 (∆DLL1) and CAV development in the SCHEDULE population. A positive correlation between ∆DLL1 and change in maximal intimal thickness (∆MIT) from baseline to 1 (r = 0.31, P = .016) and 3 years (r = 0.39, P = .006) was noted, primarily reflecting changes in the CNI group (Figures 3A,B). However, we found no correlation between MIT and DLL1 levels at the individual time points. ∆DLL1 could potentially reflect improved allograft function, but correlations with changes in allograft function (ie, left ventricular ejection fraction [LVEF], left ventricular end diastolic diameter [LVEDD], peak early diastolic mitral inflow velocity/peak late diastolic mitral inflow velocity [E/A ratio], and NT-proBNP) were poor for both ∆MIT and ∆DLL1 (Table S2). However, ∆DLL1 correlated with change in CRP (r = 0.30, P = .018), mainly driven by a correlation in the CNI group (r = 0.58, P = .001), and creatinine (r = 0.49, P < .001) where the correlation was present in both treatment groups (Table S2). For the whole study population, the association between ∆DLL1 and ∆MIT was sustained when treatment, ∆NT-proBNP, and ∆creatinine were included in a linear regression model (Figure 3C). These data support that the association between changes in circulating DLL1 and MIT are independent of changes in allograft or kidney function. Moreover, changes in percent atheroma volume (∆PAV), an additional marker of CAV development also downregulated by EVR in the SCHEDULE population,5 were also significantly correlated with ∆DLL1 at 1 and 3 years, primarily reflecting changes in the CNI group, and again, independently of allograft or kidney function (Figure 3C).

    Details are in the caption following the image
    Association between cardiac allograft vasculopathy progression and change in DLL1. The plots show the correlation between change in maximal intimal thickness (MIT) and percent atheroma volume (PAV) and DLL1 at (A) 1 and (B) 3 years. The numbers show the Spearman correlation coefficient and P-value for the total population (black) and in the CNI (blue) and everolimus (red) treatment arms. C, linear regression analysis of change in MIT and PAV at 1 and 3 years with treatment and change in DLL1, creatinine, and N-terminal pro-B-type natriuretic peptide (NT-proBNP) for the same period as predictors. Beta, standardized regression coefficient; t, t statistic, P, probability [Color figure can be viewed at wileyonlinelibrary.com]

    3.6 Effect of EVR on Notch ligand DLL1 in myocardial tissue

    To assess whether the observed effects of EVR on DLL1 could be observed in the myocardium, we determined expression of mRNA for DLL1 and Notch receptors 1 and 2 in myocardial tissue from both study arms (EVR, n = 14; CNI, n = 18) in the SCHEDULE population. Whereas DLL1 was downregulated in the EVR group, NOTCH1/2 were similar between the groups (Figure 4C). It is notable that DLL1 correlated with MIT 1 year post-HTx for both study groups (n = 28, r = 0.40, P = .003) (Figures 4D,E), and a similar finding was also found between DLL1 and PAV (r = 0.36, P = .057). DLL1 also correlated with LVEDD, and NOTCH1 was correlated with NT-proBNP, suggesting some association with allograft function (Figure 4D).

    Details are in the caption following the image
    Effect of everolimus on the expression of Notch in myocardial tissue. A, integrity of mRNA from endomyocardial biopsies (n = 27) obtained at the 1-year control from the SCHEDULE population. Left lane is the ladder, whereas the2 right lanes are mRNA from biopsies. B, melting curve from real-time polymerase chain reaction showing specific primers (DLL1, NOTCH1/2, GAPDH cardiac tissue; NOTCH3/4). C, mRNA expression of NOTCH, NOTCH2, and DLL1 in endomyocardial biopsies (CNI group n = 15; Everolimus group n = 12) obtained 1 year after heart transplantation. *P < .05 (Mann-Whitney U test). D, correlation between mRNA of DLL1, NOTCH1, NOTCH2 and circulating (C) biomarkers and indices of allograft function. The numbers represent Spearman correlation coefficients (blue P < .05, red P < .01). E, correlation between maximal intimal thickness (MIT) and DLL1 mRNA 1 year after heart transplantation (showed as ln transformed values). Experiments were run in duplicates [Color figure can be viewed at wileyonlinelibrary.com]

    3.7 EVR suppresses DLL1 in activated T cells

    Activated T cells are involved in the pathogenesis of CAV,35 and Notch and mTOR signaling are crucial contributors in T cell–mediated immune responses. We therefore evaluated the effect of different concentrations of EVR (4, 12, and 40 ng/mL) on DLL1 release. Activation of T cells was associated with a time-dependent increase in DLL1 and the T cell activation marker soluble (s)CD25 as compared to unstimulated cells (Figure 5A). EVR attenuated DLL1 and sCD25 release in a time- and dose-dependent manner, but with only modest effects in further decreasing DLL1 and sCD25 release at the highest doses (Figure 5B,C). Using even higher doses (100 and 500 ng/mL) had limited effects (Figure S1A). A similar dose–response pattern has previously been reported of rapamycin on activated T cell production of Th1/Th2 cytokines,36 and it is important to note that the lower doses are relevant to targeted concentrations of EVR in the HTx patients (6-10 ng/mL in the SCHEDULE and 3-8 ng/mL in the NOCTET study). Stimulation of activated T cells with the mTOR inhibitor sirolimus, but not the CNI CsA, dampened the DLL1 and sCD25 release in a pattern similar to EVR, whereas CNI reduced sCD25 release dose dependently at 72 hours (Figure S2). These results suggest that the observed effects on circulating DLL1 in the EVR group may be mediated by T cells, attributable to EVR exposure rather than freedom from CNI. Moreover, in activated T cells, EVR attenuated mRNA expression of the mTOR-responsive gene UTP15 (Figure S1B) in a pattern similar to DLL1 and sCD25, supporting a link between mTOR complex (mTORC) signaling and DLL1 release. Contrary to the effects on DLL1, the effect of EVR on mRNA levels of Notch receptors 1 and 2 was more complex, with a reduction of NOTCH1 in un-activated T cells, and a dose-dependent increase in NOTCH1 in activated T cells at 72 hours (Figure S3).

    Details are in the caption following the image
    Modifying effect of everolimus (EVR) on DLL1 secretion from activated T cells. A, DLL1 and sCD25 levels in conditioned media from T cells activated with beads coated with antibodies to CD2, CD3, and CD28 for 24, 48, and 72 hours. B, Effect of 4, 12, and 40 ng/mL EVR on DLL1 and sCD25 secretion from activated T cells (anti-CD2/3/28) after 24 hours and (C). 72 hours. *P < .05, **P < .01, ***P < .001 vs activated T cells (anti-CD2/3/28). P < .05, ††P < .01, †††P < .001 vs activated T cells treated with 4 ng/mL EVR (Student t-test). T cells from three healthy donors, experiments were run in triplicates. White bars: unstimulated T cells; gray bars: activated T cells; black bars: activated T cells treated with EVR. US, unstimulated T cells

    3.8 ECs and VSMCs may be sources of circulating DLL1

    DLL1 may be released from cells other than T cells, potentially contributing to circulating DLL1 in HTx recipients. To elucidate this, we examined the release of DLL1 from VSMCs, human aortic and umbilical vein ECs, and cardiac fibroblasts. These experiments revealed (Figures S4/S5/S6): (1) VSMCs and ECs, but not cardiac fibroblasts, secrete DLL1 in the conditioned medium and could be potential sources for circulating DLL1 in addition to T cells; (2) activation of VSMCs with the prototypical upstream inflammatory cytokine interleukin 1β enhanced DLL1 secretion, whereas the release of DLL1 upon tumor necrosis factor (TNF)–mediated EC activation, another prototypical upstream inflammatory cytokine, was more modest; and (3) EVR attenuated DLL1 release.

    3.9 Expression of DLL1 in myocardial tissue

    Our findings so far show that DLL1 mRNA is expressed in myocardial tissue following HTx. Our in vitro studies suggest that T cells, ECs, and VSMCs could be important cellular sources of DLL1. To further elucidate these findings, we performed immunohistochemistry analysis of myocardial specimens from 2HTx recipients with high MIT (0.90 and 0.81 mm), indicative of the presence of CAV. The biopsies were obtained at the first annual post-HTx visit for participants in the SCHEDULE trial (Figure 6). DLL1 protein was expressed in myocardial tissue and exhibited a distribution pattern corresponding to the presence of T cells, ECs, and VSMCs. Although we were unable to firmly establish the proximity to CAV in our specimens, the strong staining of the EC marker CD31, shown to the left in Figure 6C, could reflect endothelial activation, a common feature of CAV.37

    Details are in the caption following the image
    Expression of DLL1 in endomyocardial biopsies obtained at the first annual post-HTx visit. Localization of (A) CD3+ cells as a marker of T cells, and (B) expression in consecutive tissue slide of DLL1 as identified by immunostaining with 3,3′-diaminobenzidine (DAB); (C) CD31+ cells as a marker of endothelial cells, (D) expression in consecutive tissue slide of DLL1; (E) Alpha smooth muscle actin (αSMA)+ cells as a marker of vascular smooth muscle cells, and (F) expression in consecutive tissue slide of DLL1. Immunofluorescence visualization was performed by the Tyramide signal amplification (TSA) method. Arrowheads indicate T cells (panel A), endothelial cells (panel C), and vascular smooth muscle cells (panel E) and corresponding DLL1 expression (panels B, D, and F). Scale bar 100 μm in 10× and 10 μm in 40x [Color figure can be viewed at wileyonlinelibrary.com]

    4 DISCUSSION

    We studied circulating Notch ligands DLL1, POSTN, and DLK1 and CAV in HTx recipients receiving standard or EVR-based, CNI-free immunosuppression and in maintenance HTx recipients. Our main findings were the following: (1) In HTx recipients, plasma DLL1 and POSTN were elevated compared to healthy controls; (2) use of EVR-based, CNI-free immunosuppression was associated with decreased plasma DLL1 levels; (3) during long-term follow-up, changes in plasma DLL1 in de novo HTx recipients correlated with changes in markers of CAV (MIT and PAV); (4) in myocardial tissue from de novo HTx recipients, the DLL1 protein was expressed in a distribution pattern similar to that of T cells, ECs, and VSMCs, and DLL1 mRNA levels were attenuated in the EVR group and correlated with CAV (MIT and PAV); (5) in vitro, activation of T cells increased DLL1 secretion, and EVR and sirolimus, but not CsA, attenuated this release, and (6) ECs and VSMCs released DLL1, with attenuating effects of EVR in EC. These findings suggest a link between CAV development and DLL1 involving a complex interaction between T cells, VSMCs, and ECs, also within the heart allograft.

    As in chronic HF, circulating DLL1 levels at baseline in de novo HTx recipients are correlated with kidney function.38 More importantly, enhanced DLL1 levels compared to healthy controls persisted after adjustment for creatinine. DLL1 was correlated positively with systemic inflammation, as reflected by hsCRP and in particular uric acid, which is markedly elevated early after HTx, potentially reflecting increased metabolic and oxidative stress.39 The persistently elevated DLL1 levels, suggesting enhanced Notch pathway activity, could mean that DLL1 is a marker of these processes, but could also indicate that Notch signaling and DLL1 contribute to inflammation, endothelial activation, and metabolic and oxidative stress in HTx recipients. Furthermore, DLL1 is central in T cell–mediated responses,40 and differentiation and activation of T cells require DLL1-mediated Notch signaling.41, 40 Although T cell activation may be important for maintaining adaptive immune responses, dysregulated T cell activation may cause maladaptive immune response characterized by extensive production of inflammatory cytokines. The release of DLL1 from activated T cells in our in vitro study may propose DLL1 to be a part in a pathogenic loop leading to enhanced T cell activation in HTx recipients and potentially maladaptive T cell responses.

    A major finding was that EVR downregulated DLL1 in vivo. Our in vitro findings show a similar effect of sirolimus but not the CNI CsA, suggestive of mTORC involvement but the underlying mechanisms are not clear. mTORC regulates several pathways relevant for T cell activation such as nuclear factor (NF)-κB, and is activated by the Akt pathway, which links T cell activation to mTORC.42 Because the pattern of DLL1 release mimics that of sCD25, it is tempting to hypothesize that DLL1 could be directly regulated through mTORC. TNF, the prototypical activator of NF-κB, has been shown to increase DLL1 expression in fibroblast-like cells.43 Furthermore, another mTORC inhibitor, Torin-1, decreased the expression of DLL1 in vivo in colorectal cancer xenografts.44 Moreover, the release of DSL ligands including DLL1 is achieved by shedding induced by ADAM9/10/12,45-47 and, of interest, rapamycin (ie, sirolimus) has been shown to inhibit ADAM1048 and ADAM12.49, 50 Thus, both direct mTOR-mediated effects as well as protease activity that is also influenced by the mTOR pathway could contribute to DLL1 release.

    Neointimal proliferation and infiltration by macrophages are principal to CAV development and likely driven by several processes including upstream cytokine production from activated T cells. Indeed, T cells dominate among the immune cells accumulating in the thickened intima,51 and the beneficial effect of EVR on CAV development as reported in the SCHEDULE population5 may mirror attenuation of T cell activation. However, the molecular mechanism for the beneficial effect of EVR on CAV is not clear. Our previous attempts to identify associations between the effect of EVR on CAV and effects on a wide range of inflammatory markers, including CRP, have failed. Our present findings that EVR downregulates DLL1 both in plasma and the myocardium, and that DLL1 correlates with MIT and PAV, reflective of CAV development, may suggest that the beneficial effects of EVR on CAV could involve downregulation of DLL1 and Notch signaling. Of interest in experimental HTx, blockade of transcriptional activation downstream of all Notch receptors in T cells eliminated the effects of canonical Notch signaling and improved graft survival.52 EC injury from immunologic and nonimmunologic factors is central to CAV development,3 and Notch signaling is involved in EC dysfunction.11 Herein we show that ECs release DLL1, attenuated by EVR. Thus it is tempting to hypothesize that the beneficial effect of EVR in CAV could involve attenuated Notch signaling with inhibitory effects on DLL1 release from T cells and ECs. Although we lack data on the effects of EVR on DLL1 protein levels in the myocardium, the immunostaining showing co-localization of DLL1 to markers of T cells and ECs within the myocardium in CAV patients, may suggest that such mechanisms also could be operating in vivo within the myocardium in HTx recipients with CAV.

    In vivo, EVR decreased DLL1 in plasma and myocardium from HTx recipients. In contrast to circulating DLL1, where the temporal change in DLL1 and not the level at a specific timepoint correlated with change in MIT and PAV, myocardial DLL1 correlated with MIT and PAV in biopsies taken 1 year post-HTx. This “disassociation” between plasma and myocardial DLL1 may suggest that circulating DLL1 at a single timepoint may reflect the contribution of multiple cell types in the body, some of which may not be involved in CAV progression, whereas the change in circulating DLL1 over time to a larger extent may mirror concerted long-term changes during CAV progression. On the other hand, myocardial mRNA levels of DLL1 presumably mainly reflect DLL1 in ECs, VSMCs, and immune cells in proximity to CAV.

    Limitations to the present study include the relatively small sample sizes and the homogenous Scandinavian population. Thus, the SCHEDULE and NOCTET study populations may differ from the global HTx recipient population, which may constrain the extrapolation of our results. Moreover, lack of blood samples obtained before HTx limits the ability to interpret the pattern of DLL1 levels after HTx. Furthermore, associations do not necessarily imply any causal relationship, and we lack data on Notch signaling within the CAV lesions and relevant cells (ie, T cells and ECs), and analysis of their DLL1 mRNA and protein expression could illuminate our findings. Mechanistic data on the effects of mTOR inhibition on ligand-receptor interactions would have strengthened our findings. For our experiment on the effect of EVR on NOTCH1/2 mRNA expression in activated T cells, it might have been informative to quantify in parallel the corresponding receptor proteins and DLL1, as others have reported an upregulation of NOTCH1 proteins after anti-CD3/28 stimulation of T cells.53 Whereas plasma DLL1 performs technically well and displays good coefficients of variation and no diurnal or postprandial variation,54 less is known for POSTN. To illuminate the robustness of our findings, the stability of DLL1 in samples that were thawed more than once could have been measured. Unfortunately, we have no data on freeze/thaw stability for DLL1 for the present study. Finally, other Notch ligands, such as DLL3/4 and Jagged 1/2 could also be involved in CAV development. Future studies should address the limiting issues of the present trial.

    We conclude that in HTx recipients, DLL1 is increased and downregulated by EVR-based immunosuppression and that change in DLL1 is correlated with CAV progression. Our findings may suggest a role for Notch signaling and DLL1 in CAV progression.

    ACKNOWLEDGMENTS

    HMN thanks the Raagholt and Kardimmun Foundations for grants for the present work.

      DISCLOSURE

      The authors of this manuscript have conflicts of interest to disclose as described by the American Journal of Transplantation. Dag Solbu is an employee of Novartis Scandinavia. The SCHEDULE and NOCTET studies were supported by Novartis Scandinavia. The other authors have no conflicts of interest to disclose.

        The full text of this article hosted at iucr.org is unavailable due to technical difficulties.