Antibodies directed against soluble tumor necrosis factor receptor I (sTNF-RI), lymphotoxin-α (LTα or TNFSF1) were purchased from R&D systems. Antibodies recognizing phospho p65 (Ser536; P-p65), phospho IκBα (Ser32; P-IκBα), and TNF receptor-associated factor 2 (Traf2) were obtained from Cell Signaling Technology. Anti-fibronectin, RalA (Ras-related protein) and pan-actin were from Becton–Dickinson. Antibodies directed against actinin and collagen VI were purchased from Sigma and Biomol, respectively. Enbrel (Etanercept) and pentoxifylline (PTX) were obtained from Amgen and Sanofi-Aventis, respectively. The MIP-2α ELISA kit was purchased from R&D systems.

The knockdown of TNF-RI was accomplished using the MISSION shRNA system from Sigma–Aldrich. shRNA lentiviral plasmids (in pLKO.1-puro) with the clone IDs NM_011609.2-1538s1c1, NM_011609.2-996s1c1, and NM_011609.2-1652s1c1 were employed to generate lentiviral particles, which were used to infect MSC (MSC + TRkd) as described previously [4]. The efficacy of knockdowns was monitored by Western blot analysis before intravenous injection into recipient mice. The non-target shRNA control SHC002, which does not target a mouse or human mRNA was used as a negative control.

The generation of MCP-1 transgenic mice on FVB/N background, hereafter referred to as “TG”, has been described previously [16]. Wild type (WT) FVB/N mice were used as control in morphological and Western blot analyses. Housing and handling of mice were in accordance with the American Physiological Society guidelines for animal welfare and the Bioethical Committee of the District of Darmstadt, Germany. Long-term renewing adult bone marrow-derived multipotent mesenchymal stem cells (MSC) were isolated and transplanted into tail veins of recipient hosts as described previously [4]. Treatment with Enbrel was accomplished by subcutaneous injection at 10 μg/g body weight in saline solution while PTX was injected intraperitoneally at 8 μg/g body weight in saline solution. WT and TG mice were analyzed at 4 months of age by MRI followed by the first injection of MSC, Enbrel or PTX 2 weeks later. Injection of saline solution was used as a sham control. Three additional injections of MSC, Enbrel or PTX were done at weekly intervals as outlined in (Fig. S1A). At 6 month of age, mice were analyzed by MRI, western blot and immunofluorescence.

To induce acute lung injury, wild type mice C57/Bl6 mice were subjected to intratracheal instillation of 50 or 100 µg Lipopolysaccharide (LPS). Bronchoalveolar lavage fluid (BALF) was collected at indicated time points. Albumin leakage was determined as described previously [12]. MSC, Enbrel and PTX were applied either intratracheally (i.t.) or intraperitoneally (i.p.). Polymorphonuclear leukocyte numbers were determined after Pappenheim staining of cytospin preparations as described previously [12].

Cardiac MRI measurements were performed on a 7.0 T Bruker Pharmascan, equipped with a 300 mT/m gradient system, using a custom-built circularly polarized birdcage resonator and the Early Access Package for self-gated cardiac Imaging (Intragate, Bruker) [18]. MRI data were analyzed using Qmass digital imaging software (Medis, Leiden, Netherlands) as described previously [20].

Cryosections for histological and immunohistochemical analyses were prepared from snap frozen hearts and subjected to Masson’s trichrome according to the manufacturer’s instructions (Sigma–Aldrich). Immunofluorescence was accomplished using 7 μm thick cryosections after 4 % paraformaldehyde fixation for 10 min. After overnight incubation with primary antibodies, slides were washed with PBS and incubated with anti-rabbit Alexa 488- or Alexa 594 -conjugated secondary antibodies (Alexa), washed with PBS and counterstained with FITC- or TRITC-conjugated phalloidin (Sigma) to visualize actin [20]. Nuclei were visualized by incubation with DRAQ5™ (Alexis biochemicals) for 15 min. Omission of the first antibody served as negative control.

MSC were grown in SILAC medium for at least five cell divisions as described previously to ensure complete labeling of cells [17]. Cells were washed six times in serum-free medium, and incubated in serum-free medium. The conditioned medium containing the secreted proteins was collected, filtered using a 0.45 μm filter (Millipore, Bedford, MA) and concentrated using a 3000 Da molecular mass cutoff spin column, Centriprep (Millipore). Processing of protein samples and mass spectrometry was performed as described [7] using an LTQ-Orbitrap (Thermo Fisher Scientific) combined with an Agilent 1200 nanoflow HPLC system. The mass spectrometer was operated in the data-dependent mode to automatically measure full MS scans and MS/MS spectra. Peptides were identified by searching in the International Protein Index sequence database (mouse IPI, version 3.24) using MaxQuant (version 1.5.1.0), which performs peak lists, SILAC quantification and false positive rate determination. Go-terms were implemented using Perseus (version 1.5.0.15).

Protein extracts were prepared using standard procedures [20]. Ten micrograms of protein extracts was separated on 4–12 % SDS–PAGE gradient gels and processed as previously described [40]. Immunoreactive proteins were quantified with the Quantity One software (BioRad) after visualization on the VersaDoc system (BioRad). Relative concentrations of secreted cytokines were analyzed using the RayBio cytokine antibody array 3.1 as described by the supplier (RayBiotech, Norcross, USA).

The following tests were used to investigate statistical significance: (1) for three groups one-factor ANOVA and subsequent Bonferroni multiple comparisons or Kruskal–Wallis test and subsequent Dunn’s multiple comparisons in the case of missing normality or heterogeneous variance of the data; (2) for two groups unpaired t test or Mann–Whitney U test in the case of missing normality or heterogeneous variance of the data. For the survival analysis the Kaplan–Meier method was used. Differences in survival between groups were tested using the log-rank test and Pearson’s χ ² test, respectively. p values less than 0.05 were considered statistically significant. No statistical method was used to predetermine sample size.

Intrigued by our previous observation that MSC efficiently home into diseased hearts characterized by strong inflammatory reactions [4] we asked whether local accumulation of MSC modulates inflammatory signaling processes in the myocardium. We decided to focus our investigations on the αMyHC-MCP-1 model. αMyHC-MCP-1 mice develop inflammatory cardiomyopathy in a time dependent manner [16] and attract MSC into the heart after intravenous administration [4]. Cardiac functions of αMyHC-MCP-1 mice are still within the physiological range at 4 months but deteriorate rapidly thereafter, leading to congestive heart failure followed by premature death usually at 6 months [16, 34]. To test the ability of MSC to improve the condition of αMyHC-MCP-1 mice, we performed four weekly injections of MSC starting 2 weeks after an initial assessment of the cardiac function by magnetic resonance imaging (MRI) at 4 months of age (Fig. S1A). Two month after initiation of the regimen cardiac functions the same animals were analyzed again by MRI followed by morphological and biochemical analyses. Sham-treated αMyHC-MCP-1 mice (n = 5) showed left atrial (LA) cavity dilatation (Fig. 1a, b), thrombus formation in the LA (Figs. 1a, 2) and a decrease of the ejection fraction (EF) from 54 % at 4 month to 32 % at 6 month (Fig. 1c). In contrast, animals treated four times with 5.0 × 10⁵ MSC (n = 12) did not show a decline of ejection fractions as determined by MRI (Fig. 1c). Moreover, dilation of the LA and thrombus formation were significantly reduced in treated compared to control animals (Figs. 1a, b, 2). Importantly, we also observed a significant extension of the lifespan of αMyHC-MCP-1 mice. All sham-treated animals died before the age of 27 weeks whereas some MSC-treated αMyHC-MCP-1 mice survived until the age of 36 weeks. The average lifespan increased from 24 ± 0.5 (untreated) to 28 ± 1.5 weeks (treated) (n = 8 for sham-treated and n = 12 for MSC-treated mice, p < 0.01) (Fig. 1d).

Next, we investigated the morphology of the myocardium of αMyHC-MCP-1 mice at 6 month of age in MSC-treated and control animals. Masson´s trichrome staining revealed a diminished fibrosis in the LA and in the left ventricle (LV) of MSC-treated mice (Fig. 2). Confocal imaging of sections stained with an antibody directed against collagen VI indicated reduced collagen deposition and improved tissue architecture in treated αMyHC-MCP-1 mice, which went along with reduced expression of ANP (Figs. 2, 3, S1B). Both the LA and LV of treated αMyHC-MCP-1 mice were characterized by moderately reduced infiltration with CD45-positive inflammatory cells (Fig. S2).

Re-establishment and maintenance of blood supply are of pivotal importance for damaged tissues. We therefore investigated the distribution of capillary vessels in MSC-treated and control mice. Staining with BS-1, a marker for endothelial cells, revealed an increase of the capillary density in the LA but not the LV of treated mice (n = 5 for control, n = 12 for treated mice) (Figs. S3, S4). The uneven increase of capillary density suggests that this effect is most likely caused by indirect mechanisms and not by differentiation of MSC to endothelial cells or direct stimulation of endothelial cell growth since we did not find a preferential accumulation of transplanted MSC within the left atrium (data not shown).

Our previous studies revealed that genetically labeled MSC injected into αMyHC-MCP-1 mice either formed small clusters or were scattered throughout the myocardium [4]. Careful morphological analysis indicated that virtually none of these cells expressed cardiomyocyte markers such as sarcomeric actinin and troponin, although a minor fraction was positive for the endothelial cell marker BS-1 (Fig. S4 and data not shown). The relatively low number of BS-1 positive cells suggested that most, if not all, of the positive effects of MSC were mediated by the local release of bioactive molecules. We therefore decided to investigate the secretome of MSC using SILAC-based mass spectrometry. Culture supernatants were sampled at 4, 12, and 18 h resulting in identification of a total of 3066 proteins with at least 2 peptides and 1 unique peptide (Table S1). Next, we performed a Gene Ontology (GO)-annotation analysis to identify proteins that were actively released from MSC and not passively due to death of cells during the sampling period. Based on the annotation we identified 585 proteins, which are localized at the plasma and basement membrane, present in the cytosol, or secreted, including several C–C– and C–X–C–motif chemokines, SDF1 and SDF2, metalloproteinase inhibitors, and TNF-R superfamily member 11B (Fig. S5; Table S1). Of these, 101 proteins were differentially enriched (>1.5-fold change) in the secretome of MSC vs the mesenchymal cell line 10T1/2 while the remaining 484 proteins were expressed at similar or even higher levels by 10T1/2 cells. Notably, six chemokine related proteins showed increased secretion from MSC compared to 10T1/2 cells (Fig. S5; Table. S1). Despite the excellent sensitivity of modern mass spectrometry, it is still not possible to detect all low abundant proteins. We therefore complemented our mass spectrometry-based secretome measurements using protein cytokine arrays. Analysis of conditioned cell culture supernatants from MSC with the mouse cytokine array 3.1 (RayBiotech) that detects 62 cytokines validated the release of 6 proteins (Axl, Cx3cl1, Igfbp5, Timp1, Vcam1, Cxcl12) identified in the mass spectrometry experiment (Table S1) and additionally recognized TARC/CCL17 (thymus- and activation-regulated chemokine) and sTNF-RI (soluble TNF receptor I) (Fig. S5).

The improvement of cardiac functions of αMyHC-MCP-1 suffering from inflammatory DCM by MSC suggested that MSC pathogenetically influence relevant processes in the diseased heart probably by a mechanism that controls inflammatory responses. Such postulate would also fit to the proposed immunomodulatory function of MSC [5, 15]. One of the potentially most interesting molecules in this respect is sTNF-RI, which was identified during our secretome analysis of MSC (Fig. S5 and data not shown). sTNF-RI is the soluble form of the receptor of tumor necrosis factors, which is present at the surface of many cells. sTNF-RI binds TNF-α and lymphotoxin-α (LTα or TNFSF1) thereby inhibiting their activities, although it has been shown that sTNF-RI also stabilizes TNF and increases its half life under certain conditions [1, 2].

To investigate whether TNF-related processes are active in αMyHC-MCP-1 mice, we analyzed several components of the TNF-pathway by Western blot analysis. We detected a strong expression of LTα in the LA and LV of αMyHC-MCP-1 mice, which was completely absent in wild-type control mice in the LA and LV (Fig. 3a, b). Interestingly, we also found a significant downregulation of sTNF-RI, which was strongly expressed in normal wild-type hearts but was only faintly present in transgenic animals in the LA and LV (Fig. 3a, b). Moreover, phosphorylation of NFκBp65, one of the main intracellular signal transducers of TNF signaling, was strongly upregulated in the LA and LV of transgenic mice indicating activation of the inflammatory TNF-pathway (Fig. 3a, b). This observation is consistent with the strong induction of the NFκBp65 targets Traf2 (TNF receptor-associated factor-2) and fibronectin (Fig. S6A, B). Furthermore, we detected increased phosphorylation of the NFκBp65 inhibitor IκBα at S32, which leads to its proteasome-mediated degradation causings derepression of NFκBp65 signaling (Fig. S8A, B). Enhanced activation of inflammatory responses in transgenic animals went along with deposition of collagen VI in the LA and LV corroborating our histological findings (Fig. 2). Statistical analysis indicated that changes in the activation of the TNF-pathway were more pronounced in the atrium than in the ventricle (Fig. 3a, b).

We next examined whether transplantation of MSC affects the TNF/NFκB system. Western blot analysis of the myocardium of αMyHC-MCP1 mice, which received MSC injections, showed a strong inhibition of phosphorylation (activation) of NFκBp65 at Ser536 compared to sham-treated controls (Fig. 4) (n = 3 for sham controls, n = 7 for treated mice, p < 0.05). Similarly, treatment with MSCs reduced levels of phosphorylated IκBα in αMHC-MCP1 hearts, which prevents its degradation thereby inhibiting NFκBp65 signaling. In addition, we observed decreased expression of the NFκBp65 targets fibronectin (Fn1) and Traf2 by immunofluorescence staining and Western blot analysis (Fig. S8) which further corroborated the inhibition of NFκBp65 signaling. Transplantation of MSC also increased levels of sTNF-RI in the hearts of αMyHC-MCP1 mice, which nearly reached levels seen in wild-type controls (Fig. 4). Interestingly, the expression of LTα dropped after transplantation of MSC (Fig. 4) suggesting that MSC inhibited TNF-feedback loops, which are believed to contribute to increased amounts of LTα in inflamed organs [35]. The inhibition of NFκBp65 activation in MSC transplanted αMyHC-MCP1 mice was associated with decreased deposition of collagen VI and fibronectin (Fig. 4, S6). In contrast, MSC infected with a lentivirus expressing a siRNA directed against TNF-RI failed to block NFκBp65 activation, collagen VI deposition, LTα expression, and failed to restore sTNF-RI expression (Fig. S7).

The relation between sTNF-RI release by MSC, inhibition of NFκBp65 activation, and inhibition of LTα expression in hearts of αMyHC-MCP-1 mice strongly suggested that disruption of TNF signaling by sTNF-RI is at least partially responsible for the positive effects of MSC on cardiac function and survival. To prove this hypothesis we tried to mimic the effects of MSC by administration of Enbrel (also known as Etanercept), a recombinant protein that consists of two extracellular domains of the TNF-RII receptor (p75) fused to IgG1, and pentoxifylline (PTX), which increases the intracellular concentration of cAMP cyclic adenosine monophosphate resulting in decreased TNF-α production. Treatment of αMyHC-MCP-1 mice with Enbrel led to significant extension of the average lifespan from 25 weeks in the sham-treated group to 34 weeks in the Enbrel group (n = 12 control mice, n = 10 Enbrel-treated mice, p < 0.001) (Fig. 5a). In contrast, no extension of the lifespan was monitored in the PTX group (n = 8 pentoxifylline-treated mice) (Fig. 5a). αMyHC-MCP-1 mice treated with either Enbrel or PTX showed a reduced degree of thrombus formation and LA dilatation at 6 months of age, although this effect was more pronounced in the Enbrel compared to the PTX group (Fig. 5a). Masson’s trichrome staining revealed a reduced fibrosis in the LA and LV (Fig. 5b) and a reduced collagen deposition after both therapies (n = 13 for each group). Similar to the LA and LV of MSC-treated mice (Fig. S2), we also observed a reduced accumulation of CD45-positive inflammatory cells in the LA and LV of Enbrel- and PTX-treated animals (Fig. S8). Moreover, we found an increase of the capillary density in the LA but not in the LV of Enbrel-treated mice (Fig. S3). No increase in the capillary density was detected in the PTX group (n = 13 for each group) (Fig. S3). Surprisingly, assessment of cardiac function by MRI did not reveal a consistent increase of the ejection fraction in Enbrel- or PTX-treated animals compared to the control group (data not shown).

To investigate whether improvement of cardiac morphology and lifespan extension after Enbrel treatment was associated with a reduced activation of the NFκB pathway, we performed a Western blot analysis. We detected a significant reduction of NFκBp65 phosphorylation in Enbrel- and PTX-treated animals (n = 6, 5, 4 for Enbrel-treated, PTX-treated, and sham-treated animals, respectively, p < 0.05), which was associated with reduced collagen VI deposition and reduced concentrations of LTα (Fig. 6a, b). Although treatment with Enbrel and PTX mimicked several effects of transplanted MSC, we did not observe an increase of sTNF-RI concentrations, which might indicate that sTNF-RI was directly derived from MSC. Taken together, our experiments revealed that injection of recombinant soluble TNF receptor recapitulated many, but not all effects of MSC suggesting that inhibition of the TNF/NFκB pathway is responsible for a major part of the beneficial effects of MSC in inflammatory DCM.

TNF signaling also plays a major role in the pathophysiology of lung diseases such as acute lung injury (ALI) and chronic obstructive pulmonary disease (reviewed by [27]). To determine whether the release of sTNF-RI by MSC represents a general principle that might explain the beneficial role of MSC in the suppression of inflammatory processes, we induced ALI through intratracheal administration of the bacterial endotoxin Lipopolysaccharide (LPS). In agreement with previous findings we detected a more than 20-fold increase of TNF-α concentrations by ELISA in the bronchoalveolar lavage fluid (BALF) within 6 h after intratracheal instillation of 50 µg LPS compared to sham-treated controls (Fig. 7a). Furthermore, we observed a robust recruitment of polymorphonuclear leukocytes (PMN) (Fig. 7c) into the alveolar space at 6 and 24 h after LPS administration, which was accompanied by leakage of FITC-labeled albumin from the vascular into the alveolar compartment, indicative of barrier function loss [12] (Fig. 7b). The damage of the alveolar epithelial barrier was also reflected by the presence of numerous erythrocytes in the BALF (Fig. 7e). Importantly, intratracheal instillation of 5 × 10⁵ MSC improved LPS-induced alveolar epithelial barrier leakage and reduced accumulation of PMN and the concentration of macrophage inflammatory protein 2α (MIP-2α), a PMN recruiting cytokine, in the BALF (Fig. 7b, c). In contrast, administration of MSC, in which the TNF-RI was knocked down by shRNA (MSC + TRkd), did not correct any parameters of ALI. The number of PMN (3.1 × 10⁶ PMN in MSC-treated animals vs 4.4 × 10⁶ PMN in MSC + TRkd, p < 0.05) as well as alveolar leakage (0.05 AU in MSC-treated mice vs 0.09 AU in MSC + TRkd treated animals; p < 0.05) did not show any significant differences compared to sham-treated mice (Fig. 7b, c). Similarly, the concentration of MIP-2α did not decline in the BALF of mice that received MSC + TRkd (Fig. 7c). To further prove that sTNF-RI is a major mediator of the positive effects of MSC during ALI, we administered Enbrel instead of MSC releasing sTNF-RI. We found that Enbrel recapitulated most effects of MSC in mice with ALI including reduction of albumin leakage, reduction of PMN numbers, and decrease of MIP-2α in the BALF (Fig. 7c, d). We noted that it was more effective to administer Enbrel i.t. rather than i.p., which might be explained by a better bioavailability of the protein within the damaged tissue (Fig. 7d). Taken together, our data strongly suggest that MSC exert protective effects in different inflammatory conditions by release of sTNF-RI.