Increased Serum Procollagen type I C-terminal Propeptide Levels are Associated with Left Ventricular Dysfunction in Patients Treated with Anthracycline-based Cancer Chemotherapy


 Background Anthracycline-based cancer chemotherapy (ACC) has been related to myocardial interstitial fibrosis (MIF), a lesion contributing to left ventricular dysfunction (LVD). We investigated whether biomarker-assessed MIF was associated with LVD in patients with breast cancer receiving ACC and in patients with ACC-induced heart failure (ACC-HF). Moreover, chemotherapy agent’s pro-fibrotic activity was evaluated in human cardiac fibroblasts (HCFs). Methods Echocardiography, serum biomarkers of collagen deposition (procollagen type-I C-terminal-propeptide [PICP]) and crosslinking (collagen type-I C-terminal-telopeptide/matrix metalloproteinase-1 ratio), and biomarkers of myocardial and vascular stress (galectin-3, sST2, amino-terminal pro-brain natriuretic peptide [NT-proBNP], hs-troponin-T and vascular cell adhesion molecule-1 [VCAM-1]) were assessed in 70 breast cancer patients at baseline, and during ACC at 3 and 6 months. Subclinical cardiotoxicity was defined as a global longitudinal strain (GLS) relative reduction>15%. In addition, PICP and NT-proBNP were determined in 347 patients with different HF etiologies, 37 with ACC-HF of whom 33 had 12-month-follow-up LVEF assessment. HCF activation was examined after incubation with chemotherapy agents for 24 hours. Results In breast cancer patients, six-months-ACC reduced GLS and upregulated PICP, hs-troponin-T, NT-proBNP and VCAM-1 (P<0.01) versus baseline. At 6 months, elevation of PICP was higher in patients with subclinical cardiotoxicity versus the remaining patients (P for interaction<0.001). PICP levels were directly associated with a relative reduction in GLS (P<0.001). In ACC-HF patients, baseline LVEF was inversely associated with PICP and NT-proBNP (P<0.01). After 12 months, LVEF did not change in patients with higher basal PICP (3rd tertile) but improved in the remaining patients (P<0.001). Doxorubicin, cyclophosphamide and trastuzumab stimulated collagen synthesis in HCFs, including PICP extracellular cleavage. Conclusion These results indicate that increased levels of PICP, a biomarker related to MIF, are associated with early LVD in ACC-treated patients with breast cancer and with established LVD in ACC-HF patients, hindering LV functional improvement after 12 months. Chemotherapy can directly activate collagen metabolism and increase PICP extracellular presence in HCFs. Therefore, PICP emerges as a promising biomarker of ACC-induced cardiotoxicity.

Participants were studied before chemotherapy (baseline) and at standardized intervals after epirubicin/cyclophosphamide treatment at 3 months (visit 1) and after taxane treatment, with or without anti-HER2 therapy, at 6 months (visit 2). Transthoracic echocardiography, biochemical analyses and circulating biomarker determinations were performed at baseline, visit 1 and visit 2 (Fig. 1A). Two dimensional echocardiographic, pulsed Doppler, tissue Doppler imaging and three dimensional echocardiographic studies were performed in all patients using a Philips Epic 7 echocardiography system with a X5-1 transducer. The echocardiography studies' data were stored digitally in the Dicom server. End diastolic mid LV diameter, interventricular septal thickness and LV posterior wall thickness were measured as linear measurements with mode M. LV hypertrophy was diagnosed when the LV mass index (LVMI) was > 115 g/m2 for men and > 95 g/m2 for women. 18 LVEF was measured using the biplane Simpson's method. LV longitudinal myocardial strain and Automated Function Imaging were assessed by an experienced investigator using the Automated Cardiac Motion Quanti cation (aCMQ) on Q-Lab software installed. For the Speckletracking analysis the software automatically traced the myocardial motion on the previously acquired images of four chamber, two chamber and three chamber standard apical views. The operator adjusted manually the myocardial limits if automated tracking was considered inaccurate. Global longitudinal strain (GLS) was obtained from an average of regional longitudinal strain measured from a 17 myocardial segments model. 19 All measurements were made by a single observer who was blinded to all clinical and biomarker data. The intraobserver coe cients of variation for GLS and LVEF were 3.1% and 4.4%, respectively.
Early subclinical cardiotoxicity was de ned as a relative decrease in 2D global longitudinal strain (GLS, as assessed by speckle-tracking analysis) > 15% versus baseline, 20 and established cardiotoxicity as an asymptomatic decrease in LVEF > 15% or an asymptomatic decrease of LVEF > 10% to an absolute value < 53% versus baseline. 20,21 Study 2: Patients with HF

Study population
The source of the study population from which this cohort is derived has been described elsewhere. 22 In summary, all patients were ambulatory HF patients treated at a multidisciplinary unit from the Division of Cardiology at the Hospital Universitari Germans Trias i Pujol (Badalona, Spain). The principal referral criterion was HF according to the European Society of Cardiology guidelines irrespective of etiology, at least one HF hospitalization, and/or reduced LVEF. The present cohort inclusion criteria were baseline echocardiogram and biomarker measurement.
Three hundred and forty-seven HF patients at baseline were categorized according to etiologies as follows: 65 patients with ischemic heart disease (IHD), 85 with diabetic cardiomyopathy (DCM), 47 with hypertensive heart disease (HHD), 45 with valvular heart disease (VHD), 68 classi ed as "Others" which included alcoholic, hypertrophic and non-compacted cardiomyopathies and 37 with ACC-HF of whom twenty-four (64.9%) had been diagnosed with breast cancer. In these patients, anthracyclines (median [IQR] cumulative doses for epirubicine and doxorubicine: 540 [450-630] and 300 [278-413] mg/m2, respectively) were given a median (IQR) of 7.0 (1.7-12.6) years ago, and the presence of cardiovascular comorbidities was further examined to exclude causes other than ACC therapy as the main driver of symptomatic HF.

Circulating Biomarkers
Blood samples were taken from the antecubital vein at the time of entry in the study, kept in ice for 1-2 hours maximum before processing, aliquoted and kept at -80ºC until measurements. All biomarkers were measured in serum. NT-proBNP was measured using an ELISA (Roche Diagnostics). The inter-assay and intra-assay coe cients of variation were less than 10%. The lower limit of detection was 5 pg of NT-proBNP per mL. Hs-TnT was measured using a highly sensitive assay (Troponin T hs STAT, Roche Diagnostics). The lower detection limit of the assay was 3 ng/l and the inter-assay coe cient less than 10%. Galectin-3 was measured by ELISA (BG Medicine). The inter-assay and intra-assay coe cients of variations were 8.0% y 4.4%, respectively. The lower limit of detection was 1.32 ng/mL. VCAM-1 was measured by using an alphaLISA (Perkin Elmer). The inter-assay and intra-assay coe cients of variation were 7.7% and 5.3%, respectively. The lower limit of detection was 0.92 pg/ml. Soluble ST2 was measured by using an EIA (ST2 Presage®, Critical Diagnostics). The inter-assay and intra-assay coe cients of variation were 6.6% and 6.1%, respectively. The lower limit of detection was 1.8 ng/ml. CITP was measured by radio-immuno assay (Orion Diagnostica). The inter-assay and intra-assay coe cients of variation were 10.2% and 9.3%, respectively. The lower limit of detection was 0.6 µg/L. MMP-1 was measured by an alphaLISA (PerkinElmer). The inter-assay and intra-assay coe cients of variation were 10.7% and 1.6%, respectively. The lower limit of detection was 82.6 pg/mL. CITP and MMP-1 values were expressed in molarity and their ratio was calculated in each patient as previously reported. 17 PICP was measured using the EIA MicroVue CICP (Quidel Corporation). The inter-assay and intra-assay coe cients of variation were 9.1% and 4.1%, respectively. The lower limit of detection was 0.2 ng/mL.
Primary cell cultures of HCFs were isolated from atrial samples obtained as discarded surgical tissue by mechanical and tissue enzymatic digestion with collagenases and characterized (i.e. vimentin staining) as previously described. 28 In addition, all cell cultures were tested for brotic response (increased mRNA procollagen type I expression) to the transforming growth factor-beta 1. All patients gave written informed consent to donate the tissue, and the Institutional Review Committee approved the study protocol. The study conformed to the principles of the Helsinki Declaration. Cells were left to expand until they reached 60% con uence. Cells were then starved in reduced serum medium for 24 h prior to stimulation with chemotherapy agents for another 24  Extracellular PICP content PICP in the supernatants of HCFs was quanti ed using the alphaLISA PIP (PerkinElmer). The inter-assay and intraassay coe cients of variation were < 1%. The lower limit of detection was 12 pg/mL. Data were normalized by viable cell number and expressed as fold change versus unstimulated cells.
LOX activity LOX activity in the supernatants of HCFs was measured with a commercially available uorimetric assay (AAT Bioquest) following the manufacture's indications as previously described . 29 Data were normalized by viable cell number and expressed as fold change versus unstimulated cells.

Statistical analysis
Non-normally distributed variables were examined after logarithmic transformation. Differences between two groups of subjects were tested by Student's t test for unpaired data once normality was demonstrated. Otherwise, a nonparametric test (Mann-Whitney U test) was used. Differences among three groups or more were tested by one-way ANOVA followed by Bonferroni's post-hoc comparisons tests in case normality was demonstrated. Otherwise, a Kruskal-Wallis test was performed followed by Mann-Whitney U test with Bonferroni correction. Categorical variables were analyzed by the chi-squared test or Fisher's exact test when necessary. Baseline associations were examined by linear regression. Mixed-effects regression modelling was used for longitudinal and association analyses. Individual linear mixed-effects models with a random intercept to account for intraparticipant correlation of repeated measures were used to estimate changes across visits. In patients with breast cancer, subclinical cardiotoxicity at visit 2, visit, their interaction term, and baseline values of the dependent variable were added as xed effects to examine whether the development of subclinical cardiotoxicity was accompanied by greater changes in the parameters of interest as compared with its absence over time. Visit as random slope was allowed if the model was signi cantly better tted compared to a model with only random intercept. The variance component structure was speci ed according to the best t model. The likelihood ratio test and Akaike information criterion were used to select the model with the best t.
To examine the independent association between PICP and GLS, confounding variables with P values < .10 in univariate analyses, including baseline GLS values, were considered in the multivariable analysis. Multicollinearity was de ned as variance in ation factor (VIF) > 5 with model reduction in case any variable showed evidence of multicollinearity. The heteroscedasticity and normality of the residuals were examined by scatterplots and by the Skewness/Kurtosis test, respectively. Logistic regression analyses were used to assess the relationships of the biomarker levels at baseline, or biomarker changes versus baseline or between visits, with the presence of subclinical cardiotoxicity. Optimal cut-off values for predicting the development of subclinical cardiotoxicity at visit 2 were determined by performing receiver-operating characteristic (ROC) curve analysis followed by the calculation of the Younden's J statistic. The Benjamini and Hochberg multiple test correction (false discovery rate of 5%) was applied to all biomarker analyses. or median (interquartile range), and categorical variables as numbers and percentages. Statistical signi cance was set as a 2-sided P of 0.05. The statistical analyses were performed by using SPSS (15.0 version) and STATA (13.0 version) software.

Analysis of longitudinal changes with cancer therapy
All patients Effects of chemotherapy on biochemical and echocardiographic parameters, and on biomarkers are shown in Table  S1. Of interest, left atrial volume index (LAVI) and LV end-diastolic and systolic volume indexes as determined by 3D echocardiography increased at visit 2 versus baseline, and a relative reduction of GLS at visit 2 was observed compared with baseline and visit 1 (Table S1). In addition, levels of NT-proBNP, hs-TnT, sST2 and VCAM-1 were increased at visit 1 as compared with baseline (Table S1, Fig. 1B). At visit 2, NT-proBNP, hs-TnT and VCAM-1 remained elevated and PICP increased as compared with baseline (Table S1, Fig. 1B).
Patients categorized according to the presence of subclinical cardiotoxicity at visit 2 Fourteen patients (20.0%) exhibited subclinical cardiotoxicity at visit 2. Table 1 shows the baseline clinical characteristics of patients classi ed according to the absence or presence of subclinical cardiotoxicity at visit 2. No clinically relevant differences were observed between the two groups of patients at baseline (Table 1).
Differences at visit 1 and visit 2 in hematological, renal and liver function parameters between patients with and without subclinical cardiotoxicity are shown in supplemental Table 2. Of interest, patients with subclinical cardiotoxicity at visit 2 exhibited lower hemoglobin and hematocrit values at visit 1 than the remaining patients (P for interaction ≤ 0.018).
Differences at visit 1 and visit 2 in echocardiographic parameters and biomarkers between patients with and without subclinical cardiotoxicity at visit 2 are shown in Table 2. Of interest, patients with subclinical cardiotoxicity exhibited higher PICP levels at visit 2 as compared with the remaining patients (P for interaction < 0.001) ( Table 2 and Fig. 2A).
No other interactions were found signi cant for the remaining biomarkers (Table 2).

Analyses of associations
Univariable linear relationships between levels of the biomarkers that showed variations across visits and all echocardiographic parameters were evaluated in all patients (Table S2). Of interest, NT-proBNP, hs-TnT, VCAM-1 and PICP exhibited a direct association with a relative reduction of GLS along chemotherapy (Table S2).
To identify factors independently associated with the relative reduction of GLS, we conducted multivariable linear mixed regression analyses including all characteristics associated with this echocardiographic parameter (de ned as P ≤ 0.10 in univariable analysis) as shown in Table 3. The nal model revealed that, along with NT-proBNP, a progressive increment in PICP levels is an independent signi cant predictor of the relative reduction of GLS over the course of chemotherapy (Fig. 2B, Table 3). Abbreviations as in Table 1.
The associations of biomarker values at baseline or interval changes across visits with subclinical cardiotoxicity at visit 2 were estimated by univariable logistic regression analyses (Table S3). As shown in the table, only the increments in PICP between baseline and visit 2 and between visit 1 and visit 2 were signi cantly associated with subclinical cardiotoxicity (Table S3). Speci cally, there was more than a 3-fold increase in the risk of subclinical cardiotoxicity for each increase in 1 SD of PICP change at visit 2 versus baseline (which corresponded to a PICP increment of 58%) or versus visit 1 (which corresponded to a PICP increment of 65%) (  (Fig. 3), with cut-off values of 65.3% (79% sensitivity, 82% speci city) and of 105% (64% sensitivity, 87% speci city) respectively. In addition, PICP increments at visit 2 vs baseline predicted cardiotoxicity as assessed by a decrease in LVEF (AUC = 0.805 [95%CI: 0.586-1.0], P = 0.008).

Analysis of associations
Univariable linear relationships showed that NT-proBNP and PICP exhibited inverse associations with LVEF in ACC-HF patients (Table 5). Of interest, PICP association with LVEF in these patients was independent of NT-proBNP levels ( Table 5). In addition, whereas NT-proBNP was inversely associated with LVEF also in HF patients of DCM and HHD origin, no other association was found between PICP and LVEF in the remaining HF etiologies (Table 5). Analysis of LVEF longitudinal changes over one year LVEF was measured at 12 months after recruitment in 33 ACC-HF patients. These patients were categorized according to baseline PICP tertiles in patients with low (1st tertile: <100 ng/mL, n = 11), medium (2nd tertile: 100-129 ng/mL, n = 11) and high (3rd tertile: >129 ng/mL, n = 11) PICP levels. As shown in Fig. 4, baseline LVEF-adjusted analyses revealed that, after one year, LVEF improved in patients with low or medium PICP but remained unchanged in those with high PICP levels at baseline.

Study 3: In vitro ndings
In a rst step, dose-response curves were performed for doxorubicin, cyclophosphamide, paclitaxel and trastuzumab in HDFs (Fig. 5). Concentrations with signi cant submaximal effects on cell viability and procollagen type-I mRNA expression, and two lower doses, were chosen for each agent to be tested in HCFs. As shown in Table S4, all agents decreased the number of cells and increased SMA mRNA expression as compared with control in HCFs. In addition, doxorubicin, cyclophosphamide and trastuzumab, but not paclitaxel, increased the expression of procollagen type-I mRNA as compared with control ( Fig. 6A-C). Moreover, these three compounds increased the expression of PCP and increased the extracellular protein levels of collagen and PICP (Fig. 6D-F), as well as LOX activity (Table S4). Finally, doxorubicin and cyclophosphamide increased the expression of PCPE, and cyclophosphamide and trastuzumab increased LOX mRNA expression, as compared with control (Table S4).

Discussion
The major ndings of this study are the following: 1) In patients with breast cancer there was a relative GLS reduction accompanied by increments in serum NT-proBNP, hs-TnT, VCAM-1 and PICP after 6 months-chemotherapy, without changes in the remaining biomarkers; 2) PICP increment after chemotherapy was more pronounced in the subset of patients who developed subclinical cardiotoxicity; 3) Levels of PICP were directly and independently associated with the relative GLS reduction over the course of chemotherapy; 4) Levels of PICP were inversely associated with LVEF in ACC-HF patients; 5) High levels of PICP at baseline were associated with lack of LVEF improvement after 12 months in ACC-HF patients; and 6) the cardiotoxic agents doxorubicin, cyclophosphamide and trastuzumab stimulated the differentiation of HCFs to a myo broblast collagen-synthesizing phenotype. To our knowledge, this is the rst study evaluating the effects of cancer chemotherapy on myocardial collagen, as assessed non-invasively by using circulating biomarkers associated with histologically-proven MIF. In addition, to the best of our knowledge, this is the rst time that the effects of chemotherapy agents on collagen metabolism have been evaluated in HCFs.
It has been widely demonstrated that NT-proBNP and hs-TnT levels increase during cancer therapy, although their utility for cardiotoxicity prediction has been recently questioned. 21,30−32 In this study, we have reported chemotherapyinduced increments in these biomarkers associated with a reduction of LV contractility, although the association shown by hs-TnT was not independent of confounding factors. On the other hand, con icting results have been shown regarding longitudinal changes in circulating VCAM-1, galectin-3 and sST2 during cancer treatment, 30,32,33 thus further studies are necessary to con rm the in uence of oncologic therapy on the serum levels of these proteins. The current study is the rst to report that anthracycline-based-chemotherapy for breast cancer increases the circulating levels of a biomarker that re ects increased collagen deposition in the myocardium, 16 being this increase particularly evident in patients developing subclinical cardiotoxicity. In addition, we have found an association between biomarker-assessed myocardial collagen content and a reduction of LV contractility along chemotherapy, independently of relevant clinical factors including chemotherapy cumulative doses and biomarkers of cardiomyocyte damage. Importantly, this association between biomarker-assessed MIF and decreased LV contractility was further con rm in patients with ACC-HF. Even more, longitudinal pilot observations in these patients suggest that the degree of MIF at baseline may in uence evolution of LV function over time.
These ndings reinforce the notion that, not only the cardiomyocyte injury, but also the presence of MIF seems to be relevant for the impairment of LV contractile function in cancer patients treated with anthracycline-based chemotherapy. 7 Mechanistically, these effects combined may result in larger accumulation of collagen encircling cardiac muscle with reduced functionality, further restricting the stretching of muscle bers and impairing the direct cell-to-cell communication necessary for synchronous activation of cardiomyocytes. 34 In this regard, several studies demonstrate a detrimental effect of MIF on the deformation of the LV wall (i.e. reduced GLS), in experimental models 35 and in patients with different cardiomyopathies. [36][37][38][39] Of notice, the PICP levels here observed in breast cancer patients with cardiotoxicity and in patients with ACC-HF were similar to those found in hypertensive heart failure patients with severe MIF. 40 It is also important to acknowledge the potential in uence of a systemic reparative response after ACC treatment on the levels of serum PICP, with activation of brotic processes in organs other than the heart. In this regard, it has been shown that anti-cancer therapies may cause arterial stiffness, with increased collagen production suggested as a potential mechanism, among others, underlying this vascular alteration. 41,42 Since arterial stiffness contributes to LVD, we cannot discard that the association found between PICP with LVD in ACC-treated patients is re ecting the combined interaction of cardiac and vascular brosis.
Although several studies have reported pro brotic effects of anthracyclines, cyclophosphamide, paclitaxel and trastuzumab in the myocardium of experimental models of cardiotoxicity, [8][9][10][11][12][13][14] and in rodent cardiac broblasts, 9,11 the role of these agents in HCFs has not been characterized. In this regard, we describe that doxorubicin, cyclophosphamide and trastuzumab stimulated the differentiation to a myo broblast collagen-synthesizing phenotype, including enhanced presence of PICP in the extracellular medium. Therefore, we may speculate that cardiotoxic agents used to treat breast cancer may directly stimulate HCFs differentiation and collagen ber synthesis and deposition. In this regard, it has been suggested that the anthracycline pro brotic effects are mediated through substance P, ROS, STAT3 and atypical G protein Gβ 5 -related mechanisms. 9,11 In addition, cyclophosphamide has been shown to induce human lung broblasts premature senescence by activating the MAP kinase signaling pathway. 43 Whether these pathways are involved in the pro-brotic actions exerted by these agents in HCFs require further studies.
Some limitations of the present study must be acknowledged. First, small sample size may have constrained statistical power to detect differences in the different biomarkers assessed in patients with breast cancer, and preclude from adequately assess the impact of potential confounders in the association and longitudinal analyses in ACC-HF patients. Second, there is lack of follow-up information beyond the considered treatment period in patients with breast cancer, but long-term follow-up is planned and ongoing. Third, serum samples after 1-year follow-up were not available to measure PICP in ACC-HF patients. Fourth, our data cannot be extrapolated to oncologic patients treated with other anti-cancer drugs different from those evaluated in this study. Fifth, because they are descriptive in nature, the associations found between PICP and cardiotoxicity do not establish causality. Finally, we cannot discard extra-cardiac sources of PICP. In this regard, determining the extracellular volume fraction in the myocardium by means of cardiac magnetic resonance could have enriched the description of the in uence of myocardial brosis on PICP serum levels in ACC-treated patients.
In summary, this study shows for the rst time that the circulating biomarker of myocardial collagen type-I deposition, PICP, is elevated in the blood of women receiving breast cancer anthracycline-based chemotherapy. In addition, we report that PICP increment is particularly pronounced in those patients with signs of early LV dysfunction, being the levels of this peptide directly and independently associated with a reduction of LV contractility along chemotherapy.
Moreover, we con rm the association of PICP with LV contractile dysfunction in patients with ACC-HF, with high levels of this peptide associated with a lack of improvement in LV contractility over time. Finally, we show for the rst time that anthracyclines, cyclophosphamide and trastuzumab exert pro brotic actions, particularly increasing collagen type-I synthesis, in HCFs. These results provide additional evidence for MIF as a cardiotoxic side-effect of anthracycline-based breast cancer treatment, and allow us to speculate that this lesion may be a pathophysiological mechanism underlying LVD in these patients. Finally, these preliminary ndings should be considered in the context of a pilot hypothesis-generating study supporting the need of further studies designed to evaluate the prognostic potential of PICP, either by itself or combined with other biomarkers of myocardial remodeling, to monitor cardiotoxicity in larger cohorts of patients under chemotherapy.

Declarations
Page 26/30 Ethical approval: All procedures performed in studies involving human participants were in accordance with the ethical standards of the Clinical Investigation Ethics Committees of the University of Navarra and of the Hospital Universitari Germans Trias i Pujol. All study procedures were in accordance with the ethical standards outlined in the Helsinki Declaration of 1975 as revised in 1983.

Consent for publication:
Not applicable Data availability: The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Con ict of interest:
AB-G has received honoraria for lectures and advisory boards for Novartis, Vifor, Boehringer Ingelheim, Roche Diagnostics and Critical Diagnostics. The remaining authors have nothing to disclose.