Serum markers of cardiac complications in a systemic sclerosis cohort

Primary cardiac involvement is one of the leading causes of mortality in systemic sclerosis (SSc), but little is known regarding circulating biomarkers for cardiac SSc. Here, we aimed to investigate potential associations between cardiac SSc and candidate serum markers. Serum samples from patients of the Oslo University SSc cohort and 100 healthy controls were screened against two custom-made candidate marker panels containing molecules deemed relevant for cardiopulmonary and/or fibrotic diseases. Left (LV) and right ventricular (RV) dysfunction was assessed by protocol echocardiography, performed within three years from serum sampling. Patients suspected of pulmonary hypertension underwent right heart catheterization. Vital status at study end was available for all patients. Descriptive analyses, logistic and Cox regressions were conducted to assess associations between cardiac SSc and candidate serum markers. The 371 patients presented an average age of 57.2 (± 13.9) years. Female sex (84%) and limited cutaneous SSc (73%) were predominant. Association between LV diastolic dysfunction and tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) (OR 0.41, 95% CI 0.21–0.78, p = 0.007) was identified. LV systolic dysfunction defined by global longitudinal strain was associated with angiopoietin 2 (ANGPT2) (OR 3.42, 95% CI 1.52–7.71, p = 0.003) and osteopontin (OPN) (OR 1.95, 95% CI 1.08–3.52, p = 0.026). RV systolic dysfunction, measured by tricuspid annular plane systolic excursion, was associated to markers of LV dysfunction (ANGPT2, OPN, and TRAIL) (OR 1.67, 95% CI 1.11–2.50, p = 0.014, OR 1.86, 95% CI 1.25–2.77, p = 0.002, OR 0.32, 95% CI 0.15–0.66, p = 0.002, respectively) and endostatin (OR 1.86, 95% CI 1.22–2.84, p = 0.004). In conclusion, ANGPT2, OPN and TRAIL seem to be circulating biomarkers associated with both LV and RV dysfunction in SSc.


Materials and methods
SSc study cohort. All SSc patients seen at the department of rheumatology at Oslo University Hospital are included in the prospective Oslo SSc cohort, and are followed on an annual basis 22 . At first visit, patients undergo clinical examination (including assessment of digital ulcers, calcinosis and modified Rodnan skin score), serum sampling (including anti-centromere (ACA) and anti-topoisomerase-I (ATA) antibody status), pulmonary function tests, high-resolution CT (HRCT) scan and echocardiography 10 . Annual follow-up consultations are performed by a rheumatologist, registering data on demographic, clinical, laboratory, functional and imaging parameters in the Norwegian Systemic Connective Tissue Disease and Vasculitis Registry (NOSVAR) 22 . Ischemic heart disease (IHD) was defined presence of an International Classification of Diseases-10 (ICD-10) IHD-code from electronic patient journals (EPJs), while hypertension (HTN) was defined as a composite of an ICD-10 HTN-code and a systolic blood pressure > 140 mmHg at time of echocardiography.
This study included patients presenting at our hospital from 2002-2016. Inclusion criteria for this study were (i) fulfillment of the 1980 American College of Rheumatology criteria for SSc 28 and/or the 2013 European League Against Rheumatism/American College of Rheumatology criteria 29 , (ii) age > 18 years and (iii) serum samples available for serum marker analysis. Demographic, clinical and laboratory data were extracted from NOSVAR and EPJs. Patients were classified as limited (lcSSc) or diffuse cutaneous (dcSSc) SSc according to skin involvement 30 . Data on medication were collected from EPJs and registered if ever used. Vital status was available for all patients at study end, extracted from the Norwegian national registry. Disease duration was defined as time from SSc diagnosis to serum sampling and observation period as time from serum sampling to death or study end. The study was approved by The Regional Committee of Health and Medical Research Ethics in South-East Norway, research protocols no. 2016/119 and 2017/1815. All included participants consented to serum sampling for research purposes. All methods were performed in accordance with the relevant guidelines and regulations.
Serum sampling and biomarker analyses. Serum samples are collected from all SSc patients entering NOSVAR at first visit (baseline) and annual visits, using a standardized protocol for analyzation following the European Scleroderma Trial and Research (EUSTAR) recommendations for biobanking 31 . For this study, baseline sera were used. Serum was centrifuged within 30 min at room temperature and stored at − 70 °C before analyzation. Two custom-made panels were applied including specific circulating markers tailored for cardiac and pulmonary disease.Markers were identified through searches of PubMed with the search terms "biomarker, interstitial lung disease, cardiac disease, pulmonary diseases and systemic sclerosis", in 2012 and 2015. After this search, expert opinion and author discussions led to the final markers. Samples from 2003-2016 were in 2016 analyzed by Luminex (panel A), using Milliplex assays (Merck Millipore), evaluating markers known from immune-related and pro-fibrotic disease states 32,33  Cardiopulmonary assessment. We had comprehensive cardiac assessments available. Systolic and diastolic cardiac function were evaluated by one of the authors (AHT) on available echocardiographies, applying updated international recommendations as previously shown 10,35 . Systolic dysfunction was defined by a global longitudinal strain (GLS) > − 17.0% 11,36 , rather than ejection fraction, due to better reproducibility. LV diastolic dysfunction was evaluated according to the 2016 guidelines as described earlier 10,37 and RV systolic dysfunction was defined by tricuspid annular plane systolic excursion (TAPSE) < 17 mm 11,35 . Suspicion of PH was based on patient history, symptoms/findings of PH, echocardiography findings, or the DETECT algorithm 15,38 , warranting referral to right heart catheterization (RHC) for PH diagnosis. According to recent suggestions from the 6th World Symposium for PH, precapillary PH was defined as mean pulmonary artery pressure (mPAP) ≥ 21 mm Hg, pulmonary arterial wedge pressure ≤ 15 mm Hg and pulmonary vascular resistance (PVR) ≥ 3.0 Wood units 39 . PAH and PH associated with ILD (PH-ILD) were defined as precapillary PH in absence or presence of ILD, respectively 38 . ILD was expressed as extent of ILD in percentage of total lung volumes on HRCT by semi-quantitative scoring method as previously described 39 . Briefly, area measurements were done precisely by drawing the region of interest to score the overall extent of fibrosis and relate this to the total lung volume. Fibrosis was expressed as the percentage of total lung volume. Significant ILD was defined as presence of ILD affecting > 10% of pulmonary tissue 5,40 . Extent of ILD on HRCT of 0.1-10% was defined mild ILD 5 .
Outcome measures. As this was the first proof of concept study assessing circulating biomarkers in cardiac dysfunction in SSc, we aimed to first assess sera levels in SSc compared to healthy controls. In a second step, we tested associations between biomarkers and cardiac dysfunction. Lastly, we assessed biomarker predictive ability for all-cause mortality. Outcome measures were as follows: . Independent t-tests and chi square tests were applied for between-group comparisons as appropriate to assess differences in sera levels between SSc patients and healthy controls. Candidate predictors for logistic and cox regression analyses were selected by expert opinion and based on the published literature 22,[32][33][34] . Logistic regressions with odds ratios (OR) and 95% confidence intervals (CI) were applied for evaluation of association between cardiac outcomes and serological biomarkers. Parameters with significance levels < 0.20 were included in multivariable models adjusting for age and sex. Multivariable logistic regression analyses were preceded by correlation tests in order to avoid multicollinearity. In multivariable models, significance levels < 0.05 were considered significant. Multivariable models were evaluated by area under the receiver-operating curve (AUC), and models presenting values > 0.7 were considered acceptable and reported. Schoenfeld's test was applied for affirmation of proportional hazards. For confirmation of clinical significance within the SSc cohort, serum markers associated with cardiac dysfunction were tested for mortality predicting abilities, adding them to multivariable models including diastolic dysfunction (model A) and precapillary PH (model B), as reported earlier 10 . These reported models included clinical-, imaging-and serological markers, and proved superior predictive ability compared to models with other appreciated risk factors, such as e.g. presence of IHD. Cox regressions with hazard ratios (HR) and 95% CI were calculated for the predictive value of serum biomarkers on mortality. C-statistics were applied to compare multivariable Cox models, with c-indexes > 0.7 representing acceptable models.
Ethics approval and consent to participate. All patients provided consent to participate in the study.
The study was approved by The Regional Committee of Health and Medical Research Ethics in South-East Norway.

Results
Study cohort and serum markers associating with SSc and selected disease features. The study cohort included 371 SSc patients with serum samples that could be analyzed for markers of panel A containing immune-related molecules and other proteins linked to pro-fibrotic and inflammatory pathways. Among these, 255 patients (69%) had serum samples also analyzed for markers of panel B, enriched for molecules previously shown to associate with human cardiac or pulmonary diseases 41 Fig. 1. We identified altered levels of 26 serum markers in SSc with varying levels in dcSSc, and the two major SSc specific autoantibodies, ACA and ATA (Fig. 1).
Patients evaluated for cardiac dysfunction in the study cohort. Echocardiographic data on GLS, LV diastolic dysfunction and/or TAPSE, within three years from serum sampling, was available in 188/371 (51%) study cohort patients (Fig. 2). Patients with available echocardiography within 3 years from serum sampling www.nature.com/scientificreports/ were older than the patients without available echocardiography (60.0 vs 54.2 years, p < 0.001), they had a shorter observation period (4.9 vs 7.5 years, p < 0.001) and higher mortality (38% vs 17%, p < 0.001). There were no differences with respect to sex, disease subtype, antibodies (ATA and ACA) or disease duration at serum sampling.

Serum markers associated with LV systolic dysfunction.
To assess the second outcome, association of circulating biomarkers and cardiac dysfunction including LV systolic dysfunction, LV diastolic dysfunction and RV systolic dysfunction in SSc, we analyzed all patients with available sera and echocardiographic data. LV systolic dysfunction assessed by GLS was available in 111/188 (59%), of whom 30/111 (27%) had systolic dysfunction defined as GLS > − 17.0% (Fig. 2). Compared to patients with normal GLS, the patients with low GLS were more often male (43% vs 11%, p < 0.001), but did not differ regarding age, subtype, ACA, ATA, precapillary PH, disease duration or observation period. By adjusting for age and sex in logistic regression, systolic dysfunction by GLS showed significant association to two of the disease-linked serum markers displayed in Fig. 1; angiopoietin 2 (ANGPT2) and OPN (Table 2).

Serum markers associated with left ventricular diastolic dysfunction.
Echocardiographic evaluation of LV diastolic dysfunction by e' , E/e' , left atrial volume index and tricuspid regurgitant velocity according to recommendations, was performed within three years of serum sampling in 135 of the study cohort patients with available echo (72%). Among these, 37/135 (27%) presented with diastolic dysfunction. There were no differences between patients segregated by diastolic function with respect to disease duration, sex, SSc subtype, ACA   Fig. 1. Sub-group analyses of serum marker levels were performed on patients with echocardiography < 6 months from serum sampling. This group showed, as SSc patients in general, highly significantly altered levels of OPN (n = 85), TRAIL (n = 111) and ANGPT-2 (n = 111), compared to controls (p < 0.001 for all three serum markers). However, this group included too few patients with outcomes of cardiac dysfunction to perform multivariable analyses.
In our previous publication, we have presented two multivariable prediction models for SSc mortality 10 . When we added serological markers separately to Model A (including diastolic dysfunction, sex, age at echocardiography, RV systolic function, diffusion capacity of the lungs, NT-proBNP and extent of skin affliction), ANGPT-2 and OPN remained independent predictors of mortality in their respective models (Table 3), yet excluding NT-proBNP from Model A due to insignificant p-value. When we added cardiac markers separately to Model B (including precapillary PH, sex, age at echocardiography, NT-proBNP and extent of skin affliction), ANGPT-2 and TRAIL remained independent predictors of mortality, retaining significance for all parameters of the original model (Table 4).

Serum markers and relation to pulmonary disease. By logistic regression, neither OPN nor TRAIL
showed association to ILD. OPN and TRAIL showed no correlation with mPAP, PVR, cardiac output (CO) nor cardiac index (CI). Angiopoietin showed a numerical association with ILD (p = 0.16) and a weak correlation with mPAP, PVR and CI, but not with CO.

Discussion
LV and RV systolic and diastolic dysfunction are frequent cardiac complications of SSc and are associated with substantial reduced survival. Today, cardiac involvement often remains subclinical until severe organ manifestation is evident, and is often diagnosed at an advanced disease stage. Biomarkers revealing patients at high risk of cardiac dysfunction and increased mortality are scarce to date and therefore in demand 10,11 .
In the present study, we investigated candidate serum markers from two custom-tailored large panels for cardiopulmonary disease as a proof of concept study. We identified several markers associated with SSc in general, and several associated with cardiac disease. Our findings show an independent association of ANGPT2 with LV and RV systolic dysfunction, and mortality strongly linking this molecule to cardiac dysfunction in SSc. Table 3. Prediction of mortality by angiopoietin 2 and OPN in multivariable cox regression. ANGPT2 angiopoietin 2, CI confidence interval, DLCO diffusion capacity of the lungs for carbon monoxide, HR hazard ratio, mRSS modified Rodnan skin score, OPN osteopontin, TAPSE tricuspid annular plane systolic excursion. HR of ANGPT2 and OPN represent an increase of one standard deviation. www.nature.com/scientificreports/ In addition, OPN, endostatin and TRAIL were associated with cardiac dysfunction, and notably, TRAIL was associated with diastolic dysfunction as well as mortality. ANGPT2, OPN and TRAIL have all been previously shown altered in SSc. In addition, they have been reported altered in general cardiac disease 41,43,44 . These markers were therefore of particular interest for evaluation of cardiac disease associated with SSc. To our knowledge, this is the first study to present data on the association between biomarkers and cardiac dysfunction in SSc, potentially also representing novel targets for therapy in these patients. Here we show strong associations between these three circulating markers with LV and/or RV systolic dysfunction, and for TRAIL also with LV diastolic dysfunction. The three markers were also independently associated to mortality. In addition to prognostic and predictive potential, a greater understanding of these serum markers' physiologic characteristics may provide insight into the insufficiently comprehended pathophysiology of SSc. We do not expect these markers to replace the well appreciated NT-proBNP, and we have therefore neither fully investigated their association to NT-proBNP. However, this work is considered to be hypothesis generating, with a potential to enlighten and expand our knowledge on cardiac SSc. Given the retrospective nature of this study, it was not possible to relate the aforementioned markers to clinical cardiac symptoms.
ANGPT2 is produced by endothelial cells 45 and presumed to impair angiogenesis and promote cardiac fibrosis 43 . ANGPT2 is an antagonist of ANGPT1, an anti-inflammatory protein stimulating vascular integrity and homeostasis 45 . Our study supports earlier reports on higher serum concentrations of ANGPT2 in SSc patients in general, but without assessment to cardiac dysfunction 46 . ANGPT2 has been suggested a potential therapeutic target against pathologic angiogenesis and inflammation 45 . It is unclear whether ANGPT2 contributes to SSc pathophysiology or whether increased levels may reflect cardiac tissue at distress, or both. Future studies on whether ANGPT2 may predict cardiac disease, and whether regulation of ANGPT2 expression may have a therapeutic potential, are in demand.
TRAIL is a cytokine inducing endothelial cell apoptosis and smooth muscle proliferation 27 . Serum levels are shown reduced in rheumatoid arthritis patients with heart failure 47 . Among patients with acute coronary syndrome, low TRAIL concentration is reported a strong predictor of poor prognosis 44 . Further, it is suggested that inhibition of TRAIL may reverse PH 27 . While our cohort presented reduced levels of TRAIL in cardiac SSc, an earlier small scale study (n = 30) reported increased serum values of TRAIL among SSc patients compared to healthy controls, with no relation to cardiac involvement 48 . Different results may also reflect dual effects of apoptosis. Thus, while apoptosis could attenuate inflammation at an early stage of disease, it could enhance cardiac failure in a fibrotic myocardium.
OPN is a cytokine recruiting macrophages and T-cells to site of inflammation. In rodents, OPN is reported upregulated in cardiac fibrosis 25 and to promote collagen synthesis 49 , possibly at least at an early stage protecting the LV from dilation and systolic dysfunction. OPN has further been suggested as a therapeutic target 41 . These findings are in concordance with our data, showing higher values in patients with reduced LV and/or RV systolic function. One may speculate whether regulation of OPN may reduce the burden of cardiac and/or pulmonary fibrosis in SSc. Future studies on the therapeutic potential of OPN regulation are in demand.
ANGPT2, OPN and TRAIL were all also strong predictors of mortality in our SSc cohort; even when adjusting for general risk factors of SSc. NT-proBNP lost significance when either ANGPT2 or OPN were included in multivariable model A including age, sex, diffusion capacity of the lungs, modified Rodnan skin score, NT-proBNP, LV diastolic dysfunction and RV systolic function. NT-proBNP is a highly recognized serum marker of heart failure, extensively utilized in clinical cardiology. NT-proBNP levels increase both with elevated cardiac pressures and reduced renal filtration. This may suggest a cardiac, or possibly renal, impact of ANGPT2 and OPN. As these two markers excluded NT-proBNP from prediction Model A, one may speculate whether these markers reveal cardiac, or non-cardiac, relations to mortality, superior to the ability of NT-proBNP. The roles of ANGPT2 and OPN are therefore especially interesting and call for future research approaches. The focus of this study is however not to design a prediction model for primary cardiac affection from SSc, but to cast light on novel promising serum markers.
Echocardiographic findings of cardiac dysfunction may be due to primary cardiac affliction, non-SSc general cardiac disease or PH, the most feared complication of SSc. Increased pulmonary pressures increase RV pressure Table 4. Prediction of mortality by ANGPT2 and TRAIL in multivariable cox regression. ANGPT2 angiopoietin 2, CI confidence interval, HR hazard ratio, mRSS modified Rodnan skin score, NT-proBNP N-terminal prohormone of brain natriuretic peptide, PH pulmonary hypertension, TRAIL tumor necrosis factor-related apoptosis-inducing ligand. The HR of ANGPT2 and TRAIL represent an increase of one standard deviation. www.nature.com/scientificreports/ and may cause septal shift towards the LV, impairing LV filling during diastole. While the reported associations may reflect primary cardiac dysfunction due to SS, we cannot disregard that some of these associations may alternatively reflect precapillary PH. However, in this report, we show ANGPT2 and TRAIL to be independent predictors of mortality, adjusted for precapillary PH in multivariable analyses. Further, severe diastolic dysfunction is a very common cause of PH in the general population. Reports indicate that a considerable amount of patients with PH due to diastolic dysfunction may be misclassified as precapillary PH 50,51 , enlightening some of the association between diastolic dysfunction and precapillary PH. It is therefore of vital importance to evaluate the impact of cardiac SSc on mortality. We were not able to compare levels of the aforementioned serum markers between pre-and postcapillary PH due to a low number of patients with postcapillary PH, which did not allow for meaningful statistical analyses. Our study possesses some limitations worth recognizing. SSc patients may present cardiac dysfunction from non-SSc etiologies, which may dilute the characteristics of SSc-related cardiac dysfunction. Second, as echocardiographies were recorded from 2003, novel parameters as GLS were only evaluable for half of the cohort. Third, the nature of our study did not allow simultaneous serum sampling and echocardiography. In order to assure a temporal relation, only patients with a maximum of three years between echocardiography and sampling were included in analyses on GLS, diastolic dysfunction and TAPSE. Fourth, patients and controls were not completely matched to age and sex, which might have had an effect on serum marker levels. Last, while we consider our well-examined cohort favorable for evaluation of biomarker associations, we lack a validation cohort and validation studies are in demand.
The study also possesses significant strengths. This proof of concept serum marker study is to our knowledge the first study on serum markers of echocardiographic verified cardiac dysfunction in SSc. The study includes data on multiple parameters of a well-characterized cohort. Both panels were custom tailored for specific circulating markers known from cardiopulmonary disease. The levels were compared to a large control group of healthy individuals, analyzed at the time of patient serum analysis. Third, while the rarity of echocardiographic abnormalities limits the number of outcomes for analyzation purposes, the large size of our cohort allowed us to evaluate independent associations between serum markers and cardiac dysfunction using multivariable regression analyses.

Conclusion
In this study, we have shown ANGPT2, endostatin, OPN and TRAIL to be altered in SSc patients with cardiac dysfunction. ANGPT2, OPN and TRAIL were further strong independent predictors of mortality in combination with known risk factors for mortality in SSc. These markers may help enlighten the inadequately understood pathomechanisms of cardiac SSc, and could even potentially prove valuable for the diagnosis and treatment of this detrimental disease.

Data availability
The datasets used during the current study are available from the corresponding author on reasonable request.