Serum IGFBP-2 in Systemic Sclerosis as a Protective Factor of Lung Dysfunction

Background: Systemic sclerosis (SSc) is a rare connective tissue disease associated with rapid evolving interstitial lung disease (SSc-ILD), driving its mortality. Specic biomarkers associated with the progression of this lung disease are highly needed. We aimed to identify specic biomarkers of SSc-ILD to predict the evolution of the disease. Methods: We compared prospectively serum levels of several biomarkers associated with lung brosis in SSc patients (n=102), among which SSc-no ILD (n=63) and SSc-ILD (n=39), compared to healthy subjects (HS) (n=39). We also performed a longitudinal study in a subgroup of 28 patients analyzing biomarkers variations and pulmonary function tests over a period of 2 years. Furthermore, we performed in vitro analysis to study the impact of Insulin like Growth Factor Binding Protein (IGFBP)-2 on brotic activity of human lung broblasts. Results: Serum levels of IGFBP-1, IGFBP-2, interleukin-8 and matrix metallopeptidase-9 were signicantly increased in SSc patients compared to HS while IGF-1 and IGFBP-3 were reduced. The variation of IGFBP-2 between baseline and 2-year follow-up was positively correlated with pulmonary function (assessed by carbon monoxide transfer coecient (KCO)) at 2-year follow-up (r=0.6, p<0.001). Receiver operating characteristic curve analysis enabled us to identify that baseline IGFBP-2<105 ng/ml was associated with a better outcome (low risk to display KCO<70% predicted) at 2-year follow-up (area under the curve=0.75 at 75% sensibility and 68% specicity, p<0.05). In vitro functional study showed that IGFBP-2 signicantly reduced broblast proliferation and pro-brotic activity. Conclusions: We showed for the rst time that serum levels of IGFBP-2 might predict the evolution of SSc-ILD. Baseline IGFBP-2 above 105 ng/ml might be a prognostic factor of alveolo-capillary

As it is described in many other ILDs, SSc-ILD is associated with multiple clinical phenotypes. It can progress with a wide spectrum of course ranging from a rapid progressive lung disease as well as a slow evolving ILD but also a non-progressive ILD. Contrary to what is seen in IPF [13], treatment is mainly based on an aggressive immunosuppressive therapy speci cally proposed in the progressive forms of SSc-ILD. One of the major problem clinicians have to deal with, as in other ILD, is to identify patients at increased risk of progression for early intervention [5,[14][15][16][17]. In this context, biomarkers are highly needed in order to help clinicians for diagnosis, prognosis and the evaluation of the therapeutic response.
The ideal biomarker would be a good predictors in SSc-ILD [15,18].
To date, the most frequently used diagnostic biomarkers for SSc are serum autoantibodies. Indeed, more than 90% of SSc patients harbor antinuclear antibodies (ANA) in their serum [19][20][21]. Some of these are highly speci c for SSc, including anti-Scl-70 (also called anti-topoisomerase I) and anti-centromere (anti-CENP-B) antibodies [22,23]. While it is clear that anti-Scl-70 is associated positivity with a greater risk of lung brosis [24], it remains unclear whether it is associated with more progressive ILD. Although ANA are historical biomarkers available for Ssc, they are not able to predict the occurrence of ILD. So far, several serum biomarkers, including surfactant protein-D (SP-D) [25,26], Krebs Von Den Lungen 6 (KL-6) [27,28] and chemokine ligand-18 (CCL18), have been identi ed to be associated with SSc-ILD.
The transforming growth factor beta (TGF-β) is known to be involved in the pathophysiology of many lung brotic diseases by stimulating the deposition of collagen and increasing lung remodeling [29,30]. TGF-β is also widely known as an inducer of broblast activity. Besides TGF-β, previous studies identi ed that insulin-like growth factor-binding proteins (IGFBPs) were also clearly associated with IPF [31] and of interest as new potential biomarkers [32].
Insulin-like growth factors (IGFs) are peptides growth factors playing a crucial role in normal growth and development in many tissues as well as in maintaining cellular homeostasis by regulating several cellular processes (such as cell proliferation, differentiation, metabolism) [33]. IGFs are found in serum and other bio uids associated with IGFBPs. Through the sequestration of IGFs, IGFBPs regulate IGF stability, transport and tissue distribution, thus determining the bioavailability and biological activity of IGFs [34]. It has been shown that IGFBPs can inhibit or potentiate IGF activities in vitro depending on cell type and experimental conditions [35]. The main IGFBPs described in ILD context are IGFBP-1 and IGFBP-2. It has been shown that IGFBP-1 and -2 were able to improve tissue repair in in ammatory conditions by increasing broblast proliferation and migration [36]. In SSc context, IGF-1 and IGFBP-3 are involved in abnormal repair process [37,38]. Further, elevated levels of IGFBP-2 were found in broncho-alveolar lavage uid of ILD patients [39], but also in serum and sputum of IPF patients [16,40].
We focused our study on the quanti cation of several serum growth factors including IGFs and IGFBPs in order to identify a potential new biomarker for SSc and sought to determine whether these biomarkers might in uence the evolution of the disease.

Subject characteristics
In this study, we prospectively recruited patients with SSc (SSc-no ILD and SSc-no ILD) and healthy subjects from our ambulatory care policlinic at CHU Liege. The blood of the patients was collected at time of diagnosis of SSc in our center . The diagnosis of SSc was made according to the international recommendations of ACR/Eular [7]. SSc-ILD was de ned by a combination of speci c HRCT images of at least 10% of all parenchyma (reticulations, honey combing and/or ground glass opacities) with clinical signs (velcros or crackels) or symptoms (cough, shortness of breath) and alteration of pulmonary function tests. We excluded all other causes of interstitial lung disease (such asbestosis, idiopathic pulmonary brosis, idiopathic non speci c interstitial pneumonia, hypersensitivity pneumonitis or toxic pneumonitis). All cases were validated after a multidisciplinary discussion in order to con rm the presence or absence of SSc-ILD. Then, we performed a longitudinal study, resampling blood 2 years after the rst analysis (n=28) (Fig. 1). HS were recruited by advertisement in our policlinic waiting room. They all denied any respiratory disease and had normal spirometric values with FEV1 > 80% predicted and FEV1/FVC ratio > 70%. The protocol was approved by the ethics committee of CHU of Liège, and all subjects gave written consent before their enrollment (Belgian number: B707201422832 ; ref : 2014/302).

Pulmonary function tests
All tests were performed according to the recommendations of the European Respiratory Society (ERS). The results were expressed in percent predicted. The total lung capacity (TLC) was measured by body plethysmography and expressed in percent predicted. The diffusion capacity of CO (DLCO) and the report DLCO/AV (alveolar volume) were measured by the single-breath carbon monoxide gas transfer method and expressed in percent predicted (SensorMedics2400He /CO Analyzer System, Bilthoven, Netherlands).

Fibroblast isolation
Primary control human lung broblast (HLF) were obtained from around 500 mg specimen of healthy lung from surgery indicated for lung tumor resection in University Hospital from Liège, Belgium. The specimen were cut in part of 3 to 5 mm size and transferred in a gentle MACS™ C-tube (Myltenyi biotech™, Benelux) containing a mixture of HBSS with red phenol (Lonza™, Verviers, Belgium), inactivated FCS 5% (Invitrogen™), collagenase A 1mg/ml (Roche™) and Dnase I 0.05mg/ml (Roche™). The tissue was dissociated with a gentle MACS™ Dissociator (Myltenyi biotech, Benelux) according to manufacturer's recommendations and digested in the enzymatic mixture during one hour at 37 °C under agitation. After a supplementary step of dissociation, the cell suspension was ltered on a 70μm cells strainer (Corning Falcon®) and washed in 50ml PBS (Lonza™) + 10mM EDTA. Then cells were plated in a 75 cm ask and cultured in DMEM (Gibco®) containing 2mMGlutamine (Gibco®), 10% inactivated Foetal Calf Serum (FCS) (Invitrogen®), 0.1% fungizone (Gibco®) 100U/ml penicillin and100mg/ml steptomycin (Gibco®) in a humidi ed incubator at 37°Cand 5% CO2. The medium was changed twice a week and after 14 days, cells reached con uence. Cells were harvested by treatment with trypsin /EDTA (Lonza™). All experiment were performed with cell between passage 3 and 5

Protein measurement
Cell culture 60000 cells were plated into12 well plate and cultured in 800µl DMEM supplemented with 2mMGlutamine, 0.1% fungizone (Gibco®) 100U/ml penicillin and100mg/ml steptomycin (Gibco®) and 10% inactivated FCS. When cells reached 80% of con uence, the medium was removed and replaced by medium with 2% inactivated serum or without serum alone or in presence of TGF-β 5ng/ml (R&D Systems Europe, Abingdon, UK). After 48h while culture mediums were recovered and conserved at -20°C for protein measurement, cells were used after trypsinisation for FACS analysis and an Viability and Proliferation measurement 8000 cells were plated into 96 well plate and cultured in 200µl DMEM supplemented with 2mM Glutamine, antibiotics and 10% inactivated FCS. When cells reached 80% of con uence, cells were deprived of serum during 24h then cultured in medium free serum. After 48h, HLF viability and proliferation were evaluated. BrdU (5-bromo-2-deoxyuridine) cell Proliferation Assay Kit were performed according to instructions provided by the manufacturer Cell signaling Technology®,Europe : At the end of the culture , 20µl BrDU were added into wells. After 4h of incubation (37°c -5% CO2) cells were xed and labeled with an BrdU antibody. The revelations were performed with the system Biotin/HRP. The BrdU incorporation was detected by measuring the absorbance at 450nm

Statistical analysis
Demographic and functional data were expressed as mean ± standard deviation (SD). The biomarkers levels were expressed as median (IQR). Comparisons between groups were performed by Dunn's test of multiple comparisons following a signi cant Kruskal-Wallis test, or by Mann-Witney or unpaired "t" test (according to the distribution of the variable) for pairwise comparison. Correlations between variables were performed using Spearman's rank correlation test. A p<0.05 was considered as signi cant. Statistical analysis and graph were performed with Prism Graph Pad software® v6. San Diego.

Results
Study population, patient characteristics, and clinical data We prospectively recruited patients with SSc (SSc-no ILD, n=63; SSc-no ILD, n=39) from our ambulatory care policlinic at CHU Liege and compared them to healthy subjects (HS) (n=39) (Fig. 1). Demographic, functional and treatment characteristics of the subjects are given in Table 1. The average age of patients compared to HS was similar. FEV1 was moderately lowered in the SSc-no ILD and SSc-ILD groups compared to HS (p<0.05 and p<0.05, respectively). SSc-ILDs present lower levels of FEV1, FVC, TLC and DLCO compared to SSc-no ILD patients (p<0.05; p<0.001; p<0.0001 and p<0.001, respectively). The blood cell count analysis showed that patients with SSc-no ILD and SSc-ILD exhibited higher levels of neutrophils (p<0.0001 and p<0.0001, respectively) and reduced levels of lymphocytes (p<0.001 and p<0.001, respectively) compared to HS. There was not difference between SSc-no ILD and SSc-ILD patients. Of note 30% of patients were receiving maintenance treatment with immunosuppressive drugs and 27% were receiving oral corticosteroids.
Then, we compared the levels of serum biomarkers between the two sub-group of SSc patients (SSc-ILD and SSc-no ILD) and HS (concentrations are listed in Table 2). There is a signi cant increase of the levels of IGFBP-2, IL-8 and MMP-9, as well as a decrease of IGFBP-3 in SSc-no ILD and SSc-ILD patients compared to HS, respectively. Of note, the level of IGFBP-1 was only increased in SSc-no ILD patients compared to HS (p<0.05). On the other side, the level of IGF-1 was reduced and CRP increased only in SSc-ILD patients compared to HS (p<0.05 and p<0.0001, respectively). Then, we focused our analysis on the difference between SSc-no ILD and SSc-ILD. Interestingly, we observed a signi cant reduction of the levels of IGFBP-1 and IGFBP-3 in SSc-ILD compared to SSc-no ILD (p<0.01 and p<0.05, respectively). We did not nd any signi cant relation between biomarkers and therapies at baseline (immunosuppressive agent or systemic corticosteroids).

Correlation between serum biomarkers and pulmonary function tests at baseline
We performed correlation analysis to assess whether biomarkers were associated with pulmonary function tests at baseline. The correlation matrix is shown in table 3. IGFBP-2 was negatively correlated with alveolo-capillar function assessed by carbon monoxide transfer coe cient (KCO) (r=-0.29, p< 0.01) (Fig. 3). In addition, YKL-40 was also negatively correlated with Forced Vital Capacity (FVC) as well as the Diffusion Lung capacity for CO (DLCO) (r=-0.31, p<0.01 and r=-0.24, p<0.05 respectively) ( Table 3).

Longitudinal analysis on serum biomarker variations and pulmonary function tests
To assess whether the variation over the time of the levels of serum biomarkers was associated with pulmonary function declines, we performed a longitudinal study in a subgroup of 28 patients analyzing biomarkers variations and pulmonary function tests over a period of 2 years (Fig. 1). Demographic and biological characteristics of SSc patients at baseline and after 2 years are given in Table 4. The 2-year longitudinal analysis of pulmonary function revealed that only KCO was signi cantly reduced (Baseline: 81 (±14) % and 2-years 73 (±12) %, p<0.001) (Fig. 4a). There was no signi cant variation of the levels of serum biomarkers between baseline and 2-year follow-up (supplementary Table S1). Next, we performed analysis to determine if pulmonary function decline was associated to the variation of serum biomarkers (Table 5). Interestingly, we found a positive correlation between the variation of IGFBP-2 and KCO at 2year follow-up (r=0.6, p<0.001) (Fig. 4b). ROC curve analysis enabled us to identify that baseline IGFBP-2 < 105 ng/ml was associated with a better outcome assessed by KCO at 2 years follow-up in the overall SSc cohort (AUC 0.75 at 75% sensibility and 68% speci city) ( Fig. 4c and Supplementary Fig. S1). Of note, we did not nd any correlations between biomarkers evolution and immunosuppressive therapies. These results suggest a predictive value of IGFBP-2 on the progression of SSc disease.

Impact of IGFBP-2 on brotic and proliferative processes
In order to elucidate the impact of IGFBP-2 on the progression of SSc disease, we investigated its effect on SSc-related processes in vitro.

IGFBP-2 reduces the proliferation of HLF broblasts
Then, we studied the effect of IGFBP-2 on HLF proliferation by BrdU assay. In presence of IGFBP-2 (100ng/ml), we observed a signi cant decrease in BrdU incorporation at 48h, which reveals a reduction of HLF proliferation (p<0.05) (Fig. 5b). As expected, the positive control (TGF-β) increases the proliferation of HLF broblasts.

IGFBP-2 reduces the brotic activity of HLF broblasts
The impact of IGFBP-2 on brotic activity of HLF broblasts was rst assessed by measuring the level of α-SMA via immunocytochemistry (Fig. 5c) and FACS analysis (Fig. 5d), then completed by the quanti cation of pro-collagen I secretion (Fig. 5e). In presence of IGFBP-2, we observed by FACS a decrease of α-SMA levels in HLF broblasts (Fig. 5d). Furthermore, the secretion of pro-collagen I was also signi cantly reduced by 35% in presence of IGFBP-2 (p<0.05) (Fig. 5e). The immunocytochemistry analysis did not show any difference without or with IGFBP-2.
Taken together, these results highlighted the anti-proliferative and anti-brotic properties of IGFBP-2.

Discussion
Our study highlights the involvement of the IGF/IGFBP pathway in SSc disease. SSc patients featured a marked increase in serum levels of IGFBP-1, IGFBP-2, IL-8 and MMP-9 whereas IGF-1 and IGFBP-3 were signi cantly reduced compared to HS. Of interest, IGFBP-2 was correlated to KCO at baseline. Two-year longitudinal analysis determined that IGFBP-2 variation was positively correlated with the KCO measurement. Of great interest, initial levels of IGFBP-2 at baseline under 105 ng/ml were associated with better KCO at 2-years, suggesting that serum levels of IGFBP-2 might predict the evolution of SSc-ILD. In vitro functional analysis highlighted the anti-brotic property of IGFBP-2. Indeed, we identi ed that IGFBP-2 was able to reduce broblastic proliferation as well as global brogenic process (α-SMA and procol-I).
In previous studies, we have identi ed that IGFBP-2 was positively associated with lung brosis in serum and induced sputum of IPF patients [31,40]. Moreover, IGFBP-2 was reduced in IPF patients receiving anti-brotic therapy, although serum levels remained higher in IPF patients than in HS [31]. Other studies focusing on IGFBP-2 in lung brosis identi ed a signi cant increase in broncho-alveolar lavage uid and in lung tissue of interstitial lung diseases without focusing on systemic sclerosis [39]. In this study, we showed that patients suffering from SSc exhibited higher levels than HS but to a lesser extend than patients suffering from IPF as previously shown in one of our study [31,40]. Of interest, we identi ed that level variation of IGFBP-2 was associated with the severity of lung evolution. Indeed, baseline serum level of IGFBP-2 under 105 ng/ml allows identifying patients with a good prognosis based on a better KCO at 2-year follow-up (above 70% pred) (DLCO/VA). This interesting observation suggests the potential prognostic value of baseline IGFBP-2 to identify SSc patients with risk of rapid evolution. Integrating new biomarkers in the follow up of SSc-ILD is challenging taking into account the variability of other clinical markers like symptoms, CRP, DLCO or FVC. Moreover, it is suitable to avoid repeated chest imaging in the follow-up of the patients to limit as much as possible irradiation. The use of serum biomarker IGFBP-2 could be a good candidate to predict the progression of SSc-ILD and need to be explored.
IGFBP-2 was able to reduce signi cantly broblast proliferation and the expression of two pro-brotic factors, pro-col-I and α-SMA, revealing its capacity to ght against brosis. The upregulation of antibrotic IGFBP-2 in SSc patients is probably a process to ght against brogenesis and to slow the progression of the disease. Similarly, other studies focusing on human dermal broblasts identi ed that IGFBP-2 inhibited IGF-induced broblastic proliferation [34,41,42]. Of note, the inhibitory effect of IGFBP-2 was reduced when it was cleaved by metalloproteases like MMP-9 [43], which was also increased in our present study in SSc group. Accordingly, increased MMP-9 levels due to neutrophilic activation could reduce simultaneously the regulatory effect of IGFBP-2.
In our study, serum levels of TGF-β were similar for all groups even though TGF-β is widely known to be associated with the pathophysiology of brosing lung disease [44]. Similarly, our previous study focusing on IPF did not nd any difference in TGF-β levels between HS and patients suffering from IPF leading to the conclusion that serum TGF-β is not a good biomarker of lung brosis [16].
YKL-40 was negatively correlated with pulmonary function tests in our study (FEV1, FVC, DLCO) even if there were no signi cant differences of the serum level between the control group and the SSC group. In the same line, there was no signi cant difference with regard to the follow-up level at 2 years. Of great interest, the basal level of YKL-40 is still negatively correlated with the FVC at 2-year foolw-up. YKL-40 is a chitinase 3-like protein 1, regulates cell proliferation and survival and is produced by activated macrophages. It acts as a growth factor for connective tissue. Con rmatory to previous studies, we identi ed that YKL-40 is associated with the lung function impairment of patients suffering from SSc [45][46][47]. Therefore, these observations need further explorations to see whether YKL-40 could act as a predictor of lung degradation for patients with SSc.
IL-8 was also increased in our study in SSc patients. IL-8 is known to be a strong chemotactic agent for neutrophils and can take part to the pathophysiological process of SSc [48] by recruiting neutrophils in lungs [49]. Of interest, it should be noted that blood neutrophils were increased in SSc compared to HS. In the same line, MMP-9 was also increased in SSc context. MMP-9 is known to be actively secreted by neutrophils [50,51], which are increased in SSc patients.
The impact of maintenance of immunosuppressive drugs on cell count and biomarker levels was not relevant in our study.

Conclusions
We conclude that IGFBP-2 is increased in patients with SSc in comparison to HS and may play a protective role against the progression of the disease by blocking brogenesis process. We certainly need to validate those results in a larger longitudinal trial to con rm the clinical value of these observations.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.