Introduction

Mechanical circulatory support by left ventricular assist device (LVAD) implantation has been established as a valuable option in heart failure (HF) therapy in times of donor heart shortage and concomitant improvements of established device systems.1 Today, the annual number of patients receiving LVAD has surpassed the corresponding numbers for cardiac transplantation in most countries.2, 3 The development of aortic valve regurgitation due to structural alterations of the aortic valve is a frequent phenomenon complicating long-term LVAD therapy.4, 5 Reoperation on the aortic valve after previous LVAD implantation is expected to have a relevant negative impact on the outcome of affected patients. Furthermore, aortic valves of LVAD patients eventually receiving heart transplantation (HTx) are regularly harvested for homografts.6 In the latter circumstance, the question arises whether LVAD therapy may trigger structural remodelling or degeneration of the aortic valve and whether this process is depending on a certain duration of LVAD support. However, knowledge about alterations of aortic valve leaflets during LVAD support, including structural and molecular changes is still limited. Only few studies with mostly small sample sizes have described morphological changes in terms of valve thickening and collagen accumulation in aortic valve tissue of LVAD recipients.7, 8 Furthermore, statements concerning infiltration of inflammatory cells have been inconsistent,9, 10 whereas an increased activation of valvular interstitial cells by increasing amounts of alpha-smooth muscle actin (α-SMA) has been described consistently.9, 11 A very recent study has analysed aortic valves by mass spectrometry revealing up-regulation of proteins associated with transforming growth factor β (TGFβ), the actin/myosin, and the immune system in LVAD patients.11 Thus, the aim of this work is to investigate the impact of LVAD therapy on inflammation, remodelling, and chondro-osteogenic differentiation of aortic valves with emphasis on LVAD duration using a more robust sample number. A detailed research on LVAD-induced alterations of the aortic valve may contribute to the development of an optimized management of patients treated with LVAD as bridge-to-transplant or destination therapy.

Methods Human aortic valve tissue

Aortic valve cusps of the diseased heart of patients undergoing HTx were freshly collected in with patients' informed written consent obtained at time of listing for HTx. The investigation conforms to the principles outlined in the Declaration of Helsinki. The study protocol was approved by the Institutional Ethics Board of the Medical Faculty, Heinrich-Heine-University, Düsseldorf, Germany (reference number: 4567). Immediately after excision, aortic valve cusps were photographed, and individual cusps were snap-frozen and stored in liquid nitrogen as described before.12

Patients

A total of 80 HF patients receiving HTx at a single centre between December 2011 and April 2018 were initially included. During this period, a standardized collection of tissue samples from patients undergoing HTx was applied with a dedicated team of trained postdocs, technicians, and medical students harvesting the tissue intraoperatively. Patients with previous aortic valve replacement or infectious disease (human immunodeficiency virus, a history of hepatitis B/C or endocarditis) and patients of whom the aortic valves could not be collected due to various reasons or of whom the aortic valves have been used for other studies were excluded from the further study (n = 17). Finally, samples of 63 patients were used for analyses involved in this study (refer to Figure 1). Demographic data, primary indication for HTx, cardiovascular risk factors, comorbidities, medication and laboratory values, and duration of LVAD support were collected. According to institutional standard of care, follow-up echocardiography was performed at 3 months post-LVAD implantation or at an earlier time point if clinically indicated. Echocardiographic data on valve function were collected with respect to the presence of aortic valve regurgitation as well as opening movement.

Flow chart of patient enrolment and experimental setting. Patient enrolment depicting initially screened patients, exclusion criteria, and included patients with subsequent experimental setting. AV, aortic valve; CT, computed tomography scan; HTx, heart transplantation; LVAD, left ventricular assist device; PCR, polymerase chain reaction; w, with; w/o, without.

Computed tomography scan for calcium detection

Calcium deposition of available aortic valve cusps (n = 18 without and n = 30 with LVAD) was evaluated by computed tomography performed on a dual-source scanner (Somatom Definition Flash, Siemens Healthineers). In order to preserve tissue quality for later gene expression analysis, computed tomography scans were performed on frozen tissue kept on dry ice during the scan process. In prior test runs, it was confirmed that results of computed tomography scans were comparable for the same tissue sample either in a frozen state or after thawing process (data not shown). Equivalent mass of calcium hydroxyapatite, based on Agatston score measurements was determined using an established software programme (syngo.via software version VB30A, Siemens Healthineers).

RNA isolation and semi-quantitative real-time PCR analysis

Analysis of gene expression was performed on aortic valve tissue of all patients enrolled in this study (n = 63). Therefore, total RNA from the cusps was isolated using TRIzol reagent and QIAGEN RNeasy Mini Kit according to the manufacturer's instructions. Complementary DNA was synthesized by using QIAGEN QuantiTect Reverse Transcription Kit. Semi-quantitative real-time PCR and subsequent analysis of gene expression was performed as described before.12, 13 The mRNA of the following targets was analysed using specific primers: interferon gamma (IFNγ), interleukin-1 beta (IL1β), tumour necrosis factor alpha (TNFα), matrix metalloproteinase 2 (MMP2), matrix metalloproteinase 9 (MMP9), osteopontin (OPN), TGFβ1, alkaline phosphatase (ALP), osteocalcin (OCN), and ribosomal protein L13a as housekeeper (primer sequences are listed in Supporting Information, Table S1).

Histology and immunohistochemistry

In a representative subset of aortic valve cusp specimen, histological analyses were performed. Aortic valves of patients without (n = 9) or with LVAD (n = 11) were snap-frozen in cryo compound, and 5 μm sections were prepared. Sections were stained according to standard protocols for haematoxylin/eosin, Movat's pentachrome and von Kossa staining as previously described.14 Immunohistochemical staining for detection of α-SMA (Sigma) and vimentin as well as staining for CD3, CD86, elastin, collagen type 1 (Abcam), and biglycan (Santa Cruz) were performed as previously described.15 Analyses and documentation were performed in a blinded fashion using pseudonyms for sample labelling with a Leica DM2000 and Leica LAS software Version 3.8.0.

Statistical analyses

GraphPad Prism Version 7.0 (GraphPad Software, San Diego, CA, USA) was used for graphics and statistical analyses. Because experimental data were non-parametric, median with interquartile range is depicted in graphical presentation, and statistical analysis was performed using unpaired two-tailed Mann–Whitney U-test. Categorical variables were analysed using Fisher's exact test. Correlations were performed using a two-tailed Spearman test, reporting the Spearman correlation coefficient r and the according P value. In general, P values < 0.05 were considered as statistically significant. Data are presented as mean ± standard error of mean.

Results

Aortic valves and data of HF patients were analysed after HTx to evaluate the impact of LVAD support on structural and molecular alterations.

Patient characteristics

Initially, 80 HF patients who underwent HTx were screened. Because of exclusion criteria like prior aortic valve replacement, infectious disease, and unavailable aortic valve samples, 22 patients without LVAD and 41 with LVAD support at the time of transplantation were included in the analyses. Assignment of aortic valves to computed tomography, histology and semi-quantitative real-time PCR analysis is depicted in Figure 1. Patient enrolment and patient characteristics are shown in Figure 1 and Tables 1 and 2. Descriptive statistics of patient data show that patients did not differ in their sex, age, and body mass index (Table 1). LVAD patients were significantly more often diagnosed with ICM as primary indication for HTx compared with patients without LVAD (P = 0.0068). Nevertheless, other indications, for example, arrhythmogenic right ventricular cardiomyopathy, were diagnosed more often in the group without LVAD (P = 0.0464), whereas for DCM, there was no difference. There was no difference in LVEF between the two groups. New York Heart Association stages did not differ between the two groups, except for New York Heart Association stage IV that was significantly more often present in patients of the LVAD group (P = 0.0360). Moreover, the patient groups showed no difference concerning major cardiovascular risk factors and comorbidities. Patients with LVAD significantly more often received antiplatelet drugs (P = 0.0003), PD5 inhibitors (P = 0.0004), and β-adrenergic blockers (P = 0.0449). Patients without LVAD significantly more frequently needed catecholamines in the immediate period prior to transplantation (P = 0.0388). Laboratory values did not vary except for C-reactive protein that was significantly higher in the group without LVAD, however with a rather small numeric difference of 0.5 mg/dL (P = 0.0331).

Table 1. Patient characteristics All patients (n = 63) W/o LVAD (n = 22) LVAD (n = 41) P value Sex (male), n (%) 48 (76) 18 (82) 30 (73) 0.5442 Age (years) 53 ± 1.4 54 ± 2.2 53 ± 1.9 0.9060 Body mass index (kg/m2) 26 ± 0.6 25 ± 0.7 27 ± 0.9 0.1191 Primary indication, n (%) Primary DCM 31 (49) 14 (64) 17 (41) 0.1172 ICM 27 (43) 4 (18) 23 (56) 0.0068 Other 5 (8) 4 (18) 1 (2) 0.0464 LVEF (%) 23 ± 1.1 26 ± 2.1 22 ± 1.2 0.1327 NYHA classification, n (%) I 0 (0) 0 (0) 0 (0) n/a II 5 (8) 3 (14) 2 (5) 0.3327 III 19 (30) 4 (18) 15 (37) 0.1588 IV 31 (49) 15 (68) 16 (39) 0.0360 Cardiovascular risk factors, n (%) History of smoking 29 (46) 8 (36) 21 (51) 0.2987 Arterial hypertension 36 (57) 10 (45) 26 (63) 0.1922 Diabetes mellitus 18 (29) 5 (23) 13 (32) 0.5642 Thereof IDD 4 (6) 2 (9) 2 (5) 0.6063 Dyslipoproteinaemia 34 (54) 8 (36) 26 (63) 0.0629 Comorbidities, n (%) Chronic kidney disease 40 (63) 13 (59) 27 (66) 0.5972 Coronary artery disease 34 (54) 9 (41) 25 (61) 0.1853 Extracardiac vascular disease 10 (16) 2 (9) 8 (20) 0.4718 Inflammatory disease 8 (13) 1 (5) 7 (17) 0.2425 Medication, n (%) Statins 27 (43) 8 (36) 19 (46) 0.5943 Antiplatelet medication 40 (63) 7 (32) 33 (80) 0.0003 ACE inhibitor 37 (59) 15 (68) 22 (54) 0.2961 PD5 inhibitor 24 (38) 2 (9) 22 (54) 0.0004 Oral anticoagulationa 39 (62) 10 (45) 29 (71) 0.0610 β-adrenergic blocker 56 (89) 17 (77) 39 (95) 0.0449 Antiarrhythmic drugs 23 (37) 11 (50) 12 (29) 0.1691 Calcium channel blocker 11 (17) 2 (9) 9 (22) 0.3017 Diuretics 50 (79) 19 (86) 31 (76) 0.5148 Oral antidiabetic drugs 8 (13) 2 (9) 6 (15) 0.7020 Allopurinol 11 (17) 6 (27) 5 (12) 0.1703 Catecholamines 3 (5) 3 (14) 0 (0) 0.0388 Laboratory values Serum creatinine (mg/dL) 1.24 ± 0.052 1.35 ± 0.080 1.19 ± 0.066 0.0895 Platelets (×1000/μL) 237 ± 12.2 235 ± 26.2 237 ± 12.6 0.3761 Leucocytes (×1000/μL) 8.9 ± 0.39 8.6 ± 0.51 9.0 ± 0.54 0.9914 C-reactive protein (mg/dL) 2.8 ± 0.73 3.1 ± 1.77 2.6 ± 0.66 0.0331 DCM, dilated cardiomyopathy; ICM, ischaemic cardiomyopathy; IDD, insulin-dependent diabetes mellitus; LVEF, left ventricular ejection fraction; n/a, not applicable; NYHA, New York Heart Association; w/o, without. Reported data are represented as total number and proportion of whole (%) or as mean ± standard error of mean. Depicted P values were obtained by using Fisher's exact test for categorical and unpaired two-tailed Mann–Whitney U-test for continuous variables; P values < 0.05 were considered as statistically significant. Reported data are represented as mean ± standard error of mean. Table 2. Left ventricular assist device data LVAD (n = 41) LVAD type, n (%) HeartWare 29 (71) Heartmate II 8 (20) Heartmate III 3 (7) ReliantHeart, HeartAssist5 1 (2) LVAD duration (days) 485 ± 59 Driveline infection, n (%) 8 (20) Aortic valve opens, n (%)a 26 (81) Aortic valve insufficiency, n (%)b 12 (32) Additional device support, n (%)c 12 (29) LVAD, Left ventricular assist device. Reported data are represented as total number and proportion of whole (%) or as mean ± standard error of mean.

Concerning the type of LVAD, 71% of LVAD patients had a HeartWare system (Medtronic), 27% had a HeartMate II or III system (Abbott), and one patient had a HeartAssist5 (ReliantHeart; Table 2). LVAD duration ranged from 14 to 1452 days (485 ± 59 days) with 20% of the patients suffering from driveline infection. Analysis of the available echocardiography data revealed that in 81% of the LVAD patients, the aortic valve showed regular opening movements. Here, information about aortic valve opening could not be retrieved for nine patients (refer to Table 2). In 32% of the LVAD patients, an aortic valve regurgitation was present. Here, information about aortic regurgitation could not be retrieved for four patients (refer to Table 2). However, unavailable information concerning aortic valve opening and aortic valve regurgitation was due to lacking documentation, HTx within 3 months after LVAD implantation, or LVAD implantation and follow-up care in other centres. Finally, additional device support like right ventricular assist device or extracorporeal membrane oxygenation was implanted in 29% of the LVAD patients.

Up-regulation of inflammatory markers in aortic valves of left ventricular assist device patients

Analysis of mRNA expression showed a significant up-regulation of the inflammatory markers IFNγ (P = 0.0038), IL1β (P < 0.0001), and TNFα (P = 0.0395) in aortic valves of patients with LVAD (Figure 2A–C). Nevertheless, driveline infections, which occurred in 20% of the patients during LVAD support, did not account for higher mRNA expression of those markers (Figure 2D–F; IFNγ: P = 0.195; IL1β: P = 0.961; TNFα: P = 0.176). C-reactive protein levels did also not correlate with mRNA expression of these inflammatory markers (refer to Figure S1). Moreover, immunostaining for CD3 and CD68 showed no infiltration of the valve cusps by leukocytes or macrophages (not shown).

LVAD induces expression of inflammatory markers in aortic valve tissue. Aortic valves of LVAD patients showed a significantly higher IFNγ (A), IL1β (B), and TNFα (C) mRNA expression compared with aortic valves of HF patients without LVAD support. mRNA expression of IFNγ (D), IL1β (E), and TNFα (F) in aortic valves of LVAD patients showed no difference between patients with and without driveline infection. Each data point reflects an individual biological replicate. Drivel. inf., driveline infection; HF, heart failure; LVAD, left ventricular assist device; ns, not significant.*P value < 0.05; **P value < 0.01; ****P value < 0.0001.

Left ventricular assist device induces MMP9 mRNA expression and valvular interstitial cell activation in aortic valves

In order to evaluate whether LVAD-induced changes affect the gross morphology of valvular tissue, we analysed a subset of aortic valves by histology and immunohistochemistry (n = 9 without and n = 11 with LVAD). In this subgroup, LVAD duration was 512 ± 123 days. Haematoxylin/eosin and Movat's pentachrome staining revealed a heterogeneous morphology of the valves (Figure 3). Scoring of relevant factors, such as leaflet thickness, tissue compactness, or distribution of extracellular matrix components, thus showed no differences between the two groups. Immunostaining for elastin, collagen type 1, and biglycan also revealed a heterogeneous morphology with no obvious differences between the groups (not shown).

LVAD does not influence gross morphology of aortic valves. Representative images of a subset of aortic valves of three different patients with or without LVAD show haematoxylin/eosin (H&E) and Movat's pentachrome staining. Morphology in representative valves is heterogeneous, and compactness of tissue as well as distribution of extracellular matrix components shows no differences between aortic valves of patients without or with LVAD. LVAD, left ventricular assist device; w/o, without; bars = 400 μm.

Analysis of mRNA expression of MMP2 and MMP9 revealed that MMP2 expression did not vary (P = 0.931), whereas MMP9 mRNA expression was significantly higher in aortic valves of LVAD patients (P = 0.0071; Figure 4A and B). Moreover, α-SMA-staining of tissue sections was performed to evaluate whether the present valvular interstitial cells (VIC) are activated. Here, aortic valves of LVAD patients significantly more often showed VIC activation by α-SMA-positive staining (P = 0.0445, Figure 4C and D).

LVAD induces remodelling of aortic valve cusp tissue. Aortic valves of LVAD patients show no difference in MMP2 mRNA expression (A), but a significantly higher MMP9 mRNA expression compared with those of HF patients without LVAD. Representative images of immunohistology staining show vimentin-positive valvular interstitial cells with higher amounts of α-SMA-positive cells in aortic valves of patients with LVAD compared with those of HF patients without LVAD (C and D). Each data point reflects an individual biological replicate. HF, heart failure; LVAD, left ventricular assist device; ns: not significant; w/o, without. **P value < 0.01; bars = 50 μm.

Left ventricular assist device induces early chondro-osteogenic differentiation without full progression to calcification of aortic valves

We further analysed the mRNA expression of OPN, TGFβ1, ALP, and OCN to investigate the impact of LVAD on chondro-osteogenic differentiation in aortic valves. Here, only OPN mRNA expression was significantly up-regulated in the aortic valves of LVAD patients (P = 0.0026), whereas TGFβ1 (P = 0.122), ALP (P = 0.983), and OCN (P = 0.161) showed no regulation (Figure 5A–D).

LVAD leads to early chondro-osteogenic differentiation. LVAD support led to a significantly higher mRNA expression of OPN (A) in aortic valves, whereas TGFβ1 (B), ALP (C), and OCN (D) mRNA expression remains unaltered. Quantification of von Kossa staining revealed a heterogeneous distribution and severity of calcification (E). LVAD support did not alter calcium mass in aortic valves (F) as determined by standardized evaluation of computer tomography scans. Each data point reflects an individual biological replicate. HF, heart failure; LVAD, left ventricular assist device; ns, not significant; w/o, without. **P value < 0.01.

Because von Kossa staining revealed a quite heterogeneous occurrence of calcification ranging from no staining to severe calcification (Figure 5E), an independent analysis was employed to quantify the level of calcification. Using the examiner-independent software-based measurement of equivalent calcium mass in three-dimensional image data of computed tomography scans, we observed no relevant difference in the calcium content of aortic valves of HF patients with or without LVAD support (P = 0.429; Figure 5F).

MMP2 and TGFβ mRNA expression negatively correlate with the duration of left ventricular assist device support

To evaluate whether duration of LVAD support influences markers of inflammation, remodelling, and differentiation, mRNA expression of IFNγ, IL1β, TNFα, MMP2, MMP9, OPN, TGFβ1, ALP, and OCN as well as calcium mass was correlated with the duration of LVAD support for each patient whose aortic valve cusp underwent computed tomography scan (n = 30). Interestingly, there was a significant negative correlation of MMP2 mRNA expression and LVAD support duration (P = 0.038; r = 0.4424; Figure 6A) as well as a strong trend to a negative correlation between TGFβ mRNA expression and LVAD support duration (P = 0.0504; r = −0.3077; Figure 6B). There was no correlation between TNFα (P = 0.061), IFNγ (P = 0.107), IL1β (P = 0.201), MMP9 (P = 0.668), ALP (P = 0.077), OPN (P = 0.669), and OCN (P = 0.139) as well as between calcium mass and LVAD support duration, respectively (P = 0.258; not shown).

Negative correlation of MMP2 and TGFβ mRNA expression with the duration of LVAD support. Spearman correlations were run to determine the relationship between investigated markers and the duration of LVAD support. MMP2 (A) and TGFβ (B) mRNA expression showed a negative correlation with LVAD support duration. Each data point reflects an individual biological replicate. LVAD, left ventricular assist device.

Influence of aortic valve opening and aortic valve insufficiency on molecular alterations

We evaluated mRNA expression patterns and calcium accumulation levels discriminating between aortic valve tissue samples of patients with documented and regular opening as opposed to a mostly closed aortic valve under LVAD support. Here, determined markers showed no difference between these subgroups (Figure S2). Using the aforementioned data set to discriminate between aortic valve samples of LVAD patients with or without documented aortic valve insufficiency, respectively, a significantly higher mRNA expression of IFNγ (P = 0.0074) was observed in patients with aortic valve insufficiency (Figure S3).

Discussion Left ventricular assist device-induced up-regulation of inflammatory markers in the aortic valve

Studies about inflammatory processes in aortic valves due to haemodynamic changes associated with LVAD support are scarce and to some extent inconsistent. The present work shows an up-regulation of IFNγ, IL1β, and TNFα mRNA expression in aortic valves of patients with LVAD support. To our knowledge, this is the first study reporting this effect of LVAD support on valvular cusp tissue.

Similar to our findings, previous reports focusing on myocardial tissue changes have shown elevated TNFα as well as elevated IL1β expression after LVAD implantation,16 whereas also contradictory reports on an adverse effect for IL1β have been published.17 Stephens et al. refer to increased protein expression of immunoglobulin complexes in aortic valves of LVAD patients.11 However, in our cohort, aggravated expression of inflammatory markers of the LVAD group was not related to driveline infections. As also reported in a longitudinal study,18 CRP levels of patients in the present study proved to be significantly lower in the LVAD group and expression of inflammatory markers did not correlate with CRP levels, thus unspecific inflammation might not be the reason for the observed findings. Moreover, we found only rarely CD68+ or CD3+ inflammatory cells with no differences in the frequency between the two patient groups, which is in line with previous literature.9, 10 Mudd et al. even have reported the absence of any inflammatory cells.10

Thus, other inflammatory cells or VIC themselves are likely the source of increased inflammatory markers. It is known that valvular fibroblasts or myofibroblasts contribute to degenerative processes in vitro19 as well as in the heart valve tissue by secretion of cytokines and chemokines (reviewed in Singh and Torzewski20 and Li et al.21). Inflammation plays a crucial role in the initiation and progression of calcific aortic valve disease.22 Here, IL1β and TNFα are key factors in this scenario by activating the canonical NF-?B pathway in VIC inducing remodelling and mineralization.23 Moreover, IFNγ is associated with progression of calcific aortic valve disease24, 25 and atherosclerotic changes (reviewed in Andersson et al.26). Taken together, considering the herein observed up-regulation of IL1β, TNFα, and IFNγ, it seems likely that aortic valves of LVAD patients might suffer from abnormal remodelling and consecutive degeneration.

Remodelling and degeneration of aortic valves during left ventricular assist device support

Left ventricular assist device patients are at risk for developing aortic valve insufficiency or for suffering from a worsening of a pre-operatively present insufficiency, which may significantly impair the long-term outcome.27 In the present study, we could not observe gross differences concerning leaflet thickness and overall extracellular matrix morphology of the aortic valve tissue of LVAD carriers compared with other HF patients undergoing HTx, confirming previous reports.8,