Anti-Fibrotic Effect of Human Wharton’s Jelly-Derived Mesenchymal Stem Cells on Skeletal Muscle Cells, Mediated by Secretion of MMP-1

Extracellular matrix (ECM) components play an important role in maintaining skeletal muscle function, but excessive accumulation of ECM components interferes with skeletal muscle regeneration after injury, eventually inducing fibrosis. Increased oxidative stress level caused by dystrophin deficiency is a key factor in fibrosis in Duchenne muscular dystrophy (DMD) patients. Mesenchymal stem cells (MSCs) are considered a promising therapeutic agent for various diseases involving fibrosis. In particular, the paracrine factors secreted by MSCs play an important role in the therapeutic effects of MSCs. In this study, we investigated the effects of MSCs on skeletal muscle fibrosis. In 2–5-month-old mdx mice intravenously injected with 1 × 105 Wharton’s jelly (WJ)-derived MSCs (WJ-MSCs), fibrosis intensity and accumulation of calcium/necrotic fibers were significantly decreased. To elucidate the mechanism of this effect, we verified the effect of WJ-MSCs in a hydrogen peroxide-induced fibrosis myotubes model. In addition, we demonstrated that matrix metalloproteinase-1 (MMP-1), a paracrine factor, is critical for this anti-fibrotic effect of WJ-MSCs. These findings demonstrate that WJ-MSCs exert anti-fibrotic effects against skeletal muscle fibrosis, primarily via MMP-1, indicating a novel target for the treatment of muscle diseases, such as DMD.


Introduction
Skeletal muscle fibrosis, one of the typical symptoms of Duchenne muscular dystrophy (DMD), is an abnormal accumulation of extracellular matrix (ECM) components in muscle tissue and is particularly observed in patients with the end-stage of DMD [1,2]. In fibrotic skeletal muscle, the expression level of pro-fibrotic factors is increased, resulting in excessive deposition of ECM proteins, such as fibronectin. As a result, muscle function is impaired and the regeneration ability of skeletal muscle is diminished, which leads to muscle dysfunction [3][4][5]. Thus, this pathological symptom is considered a hallmark of muscular dystrophies.

Anti-Fibrotic Effects of WJ-MSCs on an Mdx Model
The effects of WJ-MSCs on DMD pathology were evaluated in the gastrocnemius muscles of an mdx murine model. The level of fibrosis, presenting as the relative area stained with Sirius red solution, was increased in the mdx group as compared to the control group ( Figure 1a, 514 ± 5% of the control group value). Additionally, accumulation of calcium/necrotic fibers was increased in the mdx group as compared to the control group (Figure 1b, 661 ± 68% of the control value).
In mdx mice intravenously (I.V) injected with WJ-MSCs, the severity of fibrosis and accumulation of calcium/necrotic fibers was decreased ( Figure 1, 280 ± 5% of control value and 265 ± 28% of the control value, respectively). These data imply that WJ-MSCs have a positive effect on degenerative phenotypes of mdx mice. myotubes treated with H2O2. WJ-MSC co-cultures showed attenuated fibrosis. Western blotting analysis showed significantly decreased fibronectin (Figure 2d, 78 ± 14% of control group value) and increased MHC (Figure 2e, 73 ± 9% of control group value) levels in myotubes after co-culturing with WJ-MSCs. These data demonstrate that H2O2 treatment induced fibrosis in myotubes, but that coculturing with WJ-MSCs can ameliorate this degeneration. To elucidate the mechanisms involved in these changes, we next analyzed the proteins secreted from WJ-MSCs. Data are expressed mean ± SEM (n = 3). Bars with different superscripts are significantly different (one-way ANOVA followed by Duncan multiple range test, p < 0.05). (b) Alizarin red staining was used to detect intracellular calcium and necrotic fiber levels in skeletal muscle. Intracellular calcium deposition and necrotic fiber accumulation was increased in the mdx group as compared to the control group. However, intravenous injection of WJ-MSCs significantly alleviated this phenotype. Data are expressed as mean ± SEM (n = 3). Bars with different superscripts are significantly different (one-way ANOVA followed by Duncan multiple range test, p < 0.05).

Figure 1.
Wharton's jelly-mesenchymal stem cells (WJ-MSCs) alleviate the fibrosis phenotype in an mdx murine model. (a) The relative fibrosis area was increased in the mdx group as compared to the control group. Collagen accumulation was significantly decreased in the WJ-MSC-injection group. Data are expressed mean ± SEM (n = 3). Bars with different superscripts are significantly different (one-way ANOVA followed by Duncan multiple range test, p < 0.05). (b) Alizarin red staining was used to detect intracellular calcium and necrotic fiber levels in skeletal muscle. Intracellular calcium deposition and necrotic fiber accumulation was increased in the mdx group as compared to the control group. However, intravenous injection of WJ-MSCs significantly alleviated this phenotype. Data are expressed as mean ± SEM (n = 3). Bars with different superscripts are significantly different (one-way ANOVA followed by Duncan multiple range test, p < 0.05).

Effects of WJ-MSCs on an In-Vitro H 2 O 2 -Induced Fibrosis Model
To investigate the role of WJ-MSCs on the fibrosis observed in mdx mice, we established an in-vitro skeletal muscle fibrosis model. There are many possible hypotheses about the cause of fibrosis in DMD, such as aging, trauma, or inflammation [4], but reactive oxygen species (ROS)-induced oxidative stress is considered as a major cause of this feature of DMD. To achieve oxidative stress-induced fibrosis, hydrogen peroxide (H 2 O 2 ) of various concentrations were used to treat myotubes. We found that Figure 2. Wharton's jelly-mesenchymal stem cells (WJ-MSCs) alleviate hydrogen peroxide (H2O2)induced fibrosis. (a) Myotubes were treated with hydrogen peroxide (H2O2, 2 mM) to induce a fibrotic phenotype, and were co-cultured with WJ-MSCs for 24 h. Scale bar: 100 μm. (b) Representative Sirius red-staining images of myotubes under each condition. A picro-Sirius red staining kit was used to visualize deposition of collagens I and III. Quantitative measurements were made using Image J. Data are expressed mean ± SEM (n = 3). Bars with different superscripts are significantly different (oneway ANOVA followed by Duncan multiple range test, p < 0.05). Scale bar: 100 μm. (c-e) Protein levels of fibronectin and myosin heavy chain (MHC) were determined by immunoblotting and were normalized to beta-actin expression. Data are shown as mean ± SEM (n = 4). Bars with different superscripts are significantly different (one-way ANOVA followed by Duncan multiple range test, p < 0.05).

Identification of Proteins Secreted from WJ-MSCs in a Co-Culture System
Since WJ-MSCs are known to have a paracrine effect, we investigated proteins secreted from WJ-MSCs in a co-culture system, as described above. After co-culturing WJ-MSCs with myotubes, conditioned media were collected and concentrated 10-fold. A RayBio Biotin Label-Based Human Antibody Array 507 (RayBiotech Life, Peachtree Corners, GA, USA) was used to analyze the levels of secreted proteins present in each of the conditioned media. We found several upregulated proteins in conditioned media of the co-culture group (Figure 3a). Among these, we focused on MMP-1, the protein most strongly upregulated in the co-culture group, as compared to WJ-MSCs alone. MMP-1 is also known to have anti-fibrotic and muscle regeneration effects. To confirm the results of the antibody array, we performed an enzyme-linked immunosorbent assay (ELISA) to measure the levels (b) Representative Sirius red-staining images of myotubes under each condition. A picro-Sirius red staining kit was used to visualize deposition of collagens I and III. Quantitative measurements were made using Image J. Data are expressed mean ± SEM (n = 3). Bars with different superscripts are significantly different (one-way ANOVA followed by Duncan multiple range test, p < 0.05). Scale bar: 100 µm. (c-e) Protein levels of fibronectin and myosin heavy chain (MHC) were determined by immunoblotting and were normalized to beta-actin expression. Data are shown as mean ± SEM (n = 4). Bars with different superscripts are significantly different (one-way ANOVA followed by Duncan multiple range test, p < 0.05).

Identification of Proteins Secreted from WJ-MSCs in a Co-Culture System
Since WJ-MSCs are known to have a paracrine effect, we investigated proteins secreted from WJ-MSCs in a co-culture system, as described above. After co-culturing WJ-MSCs with myotubes, conditioned media were collected and concentrated 10-fold. A RayBio Biotin Label-Based Human Antibody Array 507 (RayBiotech Life, Peachtree Corners, GA, USA) was used to analyze the levels of secreted proteins present in each of the conditioned media. We found several upregulated proteins in conditioned media of the co-culture group (Figure 3a). Among these, we focused on MMP-1, the protein most strongly upregulated in the co-culture group, as compared to WJ-MSCs alone. MMP-1 is also known to have anti-fibrotic and muscle regeneration effects. To confirm the results of the antibody array, we performed an enzyme-linked immunosorbent assay (ELISA) to measure the levels of human MMP-1 protein in each of the conditioned media. Amounts of human MMP-1 in co-culture media were increased as compared to WJ-MSC only media (Figure 3b, 3965.9 ± 0.8% of control group value). value, +MMP-1: 105 ± 8% of control group value). Furthermore, the relative protein expression level of fibronectin was decreased (Figure 4c,d, H2O2: 149 ± 10% of control group value, +MMP-1: 119 ± 9% of control group value), and that of MHC was increased (Figure 4c and e, H2O2: 56 ± 4% of control group value, +MMP-1: 80 ± 2% of control group value). Taken together, we confirmed that WJ-MSCs secreted MMP-1 protein in co-culture systems and that this protein has an alleviating effect on H2O2induced fibrosis.

MMP-1 Is a Key Factor for the Anti-Fibrotic Effect of Human WJ-MSCs
To determine whether MMP-1 secretion was the major mechanisms involved in the anti-fibrotic effect of WJ-MSCs, MMP inhibitors (tissue inhibitor of metallopeptidase-1 [TIMP-1] and GM6001) were used to treat cells. After treatment with MMP inhibitors, WJ-MSCs were immediately cocultured with myotubes exposed to H2O2. As observed above, WJ-MSCs attenuated fibrosis induced by H2O2 (Figure 5b, H2O2: 117 ± 3% of control group value, +WJ-MSC: 87 ± 5% of control group value). However, when TIMP-1 or GM6001 was used to treat cells, no anti-fibrotic effects were observed in either group (Figure 5b, +WJ-MSC+TIMP-1: 121 ± 3% of control group value, +WJ-MSC+GM6001: 121 ± 3% of control group value). The relative protein expression level of fibronectin was changed following a similar pattern as seen with Sirius red staining (Figure 5c,d, H2O2: 183 ± 11% of control group value, +WJ-MSC: 125 ± 7% of control group value, +WJ-MSC+TIMP-1: 225 ± 35% of control group value, +WJ-MSC+GM6001: 177 ± 13% of control group value). In contrast, the expression of MHC was increased in the WJ-MSC co-culture group as compared with the treated group. However, when MMP inhibitors were used, this effect was abrogated (Figure 5c and e, H2O2: 11 ± 4% of control group value, +WJ-MSC: 37 ± 17% of control group value, +WJ-MSC+TIMP-1: 8 ± 3% of control group value, +WJ-MSC+GM6001: 11 ± 5% of control group value). These results indicated that WJ-MSCs reduced fibrosis caused by oxidative stress mainly via the secretion of MMP-1 protein. was increased in co-cultured media. MMP-1: matrix metalloproteinase-1, INHBA: inhibin, beta A, IGFBP7: insulin like growth factor binding protein 7, PDGFA: platelet derived growth factor subunit A, and THBS1: thrombospondin 1. (b) Concentration of secreted MMP-1 in each of the conditioned media was measured by using a human MMP-1 ELISA kit. A four-fold increase was observed in the co-cultured media as compared to the WJ-MSC-only media. Data are expressed as mean ± SEM (n = 5, ** p < 0.01).

Human Recombinant MMP-1 Protein Has Protective Effects in an In-Vitro H 2 O 2 -Induced Fibrosis Model
MMP-1 is known to have anti-fibrotic effects, but it has not been confirmed whether MMP-1 is effective in an H 2 O 2 -induced fibrosis model. Therefore, to verify the effect of MMP-1 protein on H 2 O 2 -induced fibrosis, we added human recombinant MMP-1 protein to myotubes treated with H 2 O 2 and demonstrated that MMP-1 treatment alleviated H 2 O 2 -induced fibrosis (Figure 4a). We also confirmed the alleviating effect of MMP-1 on fibrosis by using Sirius red staining. Collagen deposition was decreased in myotubes and was treated with MMP-1 (Figure 4b, H 2 O 2 : 133 ± 5% of control group value, +MMP-1: 105 ± 8% of control group value). Furthermore, the relative protein expression level of fibronectin was decreased (Figure 4c,d, H 2 O 2 : 149 ± 10% of control group value, +MMP-1: 119 ± 9% of control group value), and that of MHC was increased (Figure 4c,e, H 2 O 2 : 56 ± 4% of control group value, +MMP-1: 80 ± 2% of control group value). Taken together, we confirmed that WJ-MSCs secreted MMP-1 protein in co-culture systems and that this protein has an alleviating effect on H 2 O 2 -induced fibrosis.

MMP-1 Is a Key Factor for the Anti-Fibrotic Effect of Human WJ-MSCs
To determine whether MMP-1 secretion was the major mechanisms involved in the anti-fibrotic effect of WJ-MSCs, MMP inhibitors (tissue inhibitor of metallopeptidase-1 [TIMP-1] and GM6001) were used to treat cells. After treatment with MMP inhibitors, WJ-MSCs were immediately co-cultured with myotubes exposed to H 2 O 2 . As observed above, WJ-MSCs attenuated fibrosis induced by H 2 O 2 (Figure 5b, H 2 O 2 : 117 ± 3% of control group value, +WJ-MSC: 87 ± 5% of control group value). However, when TIMP-1 or GM6001 was used to treat cells, no anti-fibrotic effects were observed in either group (Figure 5b, +WJ-MSC+TIMP-1: 121 ± 3% of control group value, +WJ-MSC+GM6001: 121 ± 3% of control group value). The relative protein expression level of fibronectin was changed following a similar pattern as seen with Sirius red staining (Figure 5c,d, H 2 O 2 : 183 ± 11% of control group value, +WJ-MSC: 125 ± 7% of control group value, +WJ-MSC+TIMP-1: 225 ± 35% of control group value, +WJ-MSC+GM6001: 177 ± 13% of control group value). In contrast, the expression of MHC was increased in the WJ-MSC co-culture group as compared with the treated group. However, when MMP inhibitors were used, this effect was abrogated (Figure 5c,e, H 2 O 2 : 11 ± 4% of control group value, +WJ-MSC: 37 ± 17% of control group value, +WJ-MSC+TIMP-1: 8 ± 3% of control group value, +WJ-MSC+GM6001: 11 ± 5% of control group value). These results indicated that WJ-MSCs reduced fibrosis caused by oxidative stress mainly via the secretion of MMP-1 protein.
measurements were made using Image J. Data are expressed mean ± SEM (n = 3). Bars with different superscripts are significantly different (one-way ANOVA followed by Duncan multiple range test, p < 0.05). Scale bars: 100 μm. (c-e) Representative images of Western blot analysis to determine protein expression levels of fibronectin and myosin heavy chain (MHC). Beta-actin was used as a loading control. Data are expressed as mean ± SEM (n = 3). Bars with different superscripts are significantly different (one-way ANOVA followed by Duncan multiple range test, p < 0.05).

Figure 5. MMP-1 is required for the anti-fibrotic effect of human Wharton's jelly-mesenchymal stem cells (WJ-MSCs). (a) MMP inhibitors (TIMP-1 or GM6001
) were used to block the effects of MMP-1 secreted from WJ-MSCs. Fibrosis in myotubes was induced by H 2 O 2 (2 mM) and WJ-MSCs were co-cultured with the myotubes, with or without MMP inhibitors. Anti-fibrotic effects of WJ-MSCs were not observed when myotubes were treated with MMP inhibitors. Scale bars: 100 µm. TIMP-1: metallopeptidase inhibitor-1. (b) Collagen deposition was significantly decreased in the +WJ-MSC group as compared to the treated group. However, these effects were reversed in the MMP inhibitor-treated group. Quantitative measurements were made using Image J. Data are expressed as mean ± SEM (n = 3). Bars with different superscripts are significantly different (one-way ANOVA followed by Duncan multiple range test, p < 0.05). Scale bars: 100 µm. (c-e) Relative protein expression level of fibronectin and myosin heavy chain (MHC) was also observed. Fibronectin expression level was significantly decreased in the +WJ-MSC group as compared to the treated group. However, when myotubes were treated with MMP inhibitor, this level was significantly increased. In contrast, relative MHC expression level was increased in the +WJ-MSC group as compared to the Treat treated group, but it was decreased in +WJ-MSC+TIMP-1 and +WJ-MSC+GM6001 group. Beta-actin was used as a loading control. Data are expressed as mean ± SEM (n = 3). Bars with different superscripts are significantly different (one-way ANOVA followed by Duncan multiple range test, p < 0.05).

Discussion
In most skeletal muscle diseases, including DMD, fibrosis is a major pathogenetic mechanism [2,28,29]. Accumulation of calcium and necrotic fibers is induced by ROS stress, causing fibrosis [30][31][32]. Skeletal muscle fibrosis and accumulation of calcium/necrotic fibers are increased in DMD mice [7,[33][34][35]. Several studies have suggested potential anti-fibrotic therapy approaches for DMD [10,36,37]. As WJ-MSCs exert anti-apoptotic effects on skeletal muscle cells [27], we here investigated whether these cells could have anti-fibrotic effects, and also sought to elucidate the main paracrine factor underlying such effects. We showed that intravenous injection of WJ-MSCs into 2-5-month-old C57BL mdx mice significantly decreased fibrosis intensity and accumulation of calcium/necrotic fibers as compared to wild-type mice. We confirmed the effect of WJ-MSCs in an H 2 O 2 -induced fibrosis myotubes model and demonstrated that the paracrine factor MMP-1 is critical for these anti-fibrotic effects. For the first time, we revealed that WJ-MSCs could reduce skeletal muscle fibrosis.
A recent study reported that H 2 O 2 , a ROS with a long half-life, diffuses both within cells and across cell membranes [38][39][40][41]. H 2 O 2 can damage cellular bio-molecules, such as lipids and protein.
It induces oxidative stress and results in chronic fibrosis in DMD patients [6,42,43]. Based on these characteristics, an H 2 O 2 -induced fibrosis model was established as an in-vitro system in this study. We revealed that myotubes exposed to H 2 O 2 showed typical patterns of fibrosis, including increased expression of fibronectin. Using this in-vitro model, it was possible to study the mechanism underlying the effects of the WJ-MSCs on fibrosis.
A reduction of fibrosis was observed during fibrosis induction in myotubes co-cultured with WJ-MSCs. Furthermore, MMP-1 protein was identified as a paracrine factor through antibody array screening. In order to verify that MMP-1 was the critical factor involved in the anti-fibrotic effects of WJ-MSCs, myotubes were treated with recombinant MMP-1 protein under conditions of oxidative stress. Treatment with recombinant MMP-1 markedly reduced fibrosis in oxidative stress-induced myotubes. We further confirmed that MMP-1 is a key factor by using MMP inhibitor in our in vitro model. Similar to our results, recent studies have reported that MMP-1 has anti-fibrotic effects in a liver fibrosis model and a lacerated skeletal muscle model [44][45][46][47]. Furthermore, the increase in ECM components and upregulation of related genes in DMD patients were reported based on the analysis of differentially expressed gene profile [48]. The amount of secreted protein associated with the reported genes was found to be decreased in the co-culture conditioned media of this study, which showed that our experimental design well reflected the DMD fibrosis model. Our findings demonstrated that human WJ-MSCs play an anti-fibrotic role in skeletal muscle fibrosis via MMP-1.
We also showed increased levels of MHC in this model, but the mechanism for this observation was not identified in this study. Therefore, further study is needed to characterize the effect of MMP-1 treatment on the regenerative function of fibrotic muscle.
DMD is characterized by muscle inflammation and progressive deterioration of muscle mitochondria and function. The absence of dystrophin resulted in excessive calcium penetration into the sarcolemma (the cell membrane). Alterations in calcium and signaling pathways also cause water to enter into the mitochondria, resulting in bursting. Previous studies have documented that MSC-secreted trophic factors highly depend on external environment and status of MSC [49,50]. Some pro-inflammatory cytokines, such as IL-6, TNF-α, can induce MSC transfer of mitochondria to recuse injured cells including retinal cells, cardiomyocytes, and airway epithelia cells [51][52][53]. It appears a pro-inflammatory environment can enhance MSC-mitochondrial transfer and MSC mitochondrial transfer to T cells can in turn educate immune cells (CD4 T cells, i.e.) [54]. In this connection, MSC paracrine functions and mitochondrial transfer capacity are interactive and linked together to promote tissue regeneration.
It is well known that MSCs of different origins display distinct potential of cell proliferation and survival capacity. For example, MSCs from birth-associated tissues, preferably parts of the placenta and Wharton's jelly, and pluripotent stem cell-derived MSCs, are highly proliferative and survive longer than adult tissues derived MSCs after transplantation [55][56][57][58].
Taken together, our study suggested that reduction of fibrosis by WJ-MSCs could be a novel treatment strategy for DMD.

Ethics Statement
This study was approved by the Institutional Animal Care and Use Committee of the Samsung Biomedical Research Institute (SBRI) at Samsung Medical Center. The SBRI is an accredited facility of the Association for Assessment and Accreditation of Laboratory Animal Care International, and abides by the Institute of Laboratory Animal Resources guidelines. In accordance with the guidelines approved by the institutional review board of Samsung Medical Center, umbilical cords were collected with informed consent from pregnant mothers (IRB#2016-07-102).

Cell Culture
Wharton's jelly-derived mesenchymal stem cells (WJ-MSCs) were cultured according to the standard operating procedures of the Good Manufacturing Practice facility at ENCell Co. Ltd. (Seoul, Korea).

MMP Inhibitors Application
To determine whether the effect of WJ-MSCs on fibrosis is MMP-1-dependent, human recombinant metallopeptidase inhibitor-1 (TIMP-1) (100 ng/mL) or GM6001 (50 µM) (R&D Systems, Minneapolis, MN, USA), as an MMP inhibitor was used. After treatment with H 2 O 2 for 24 h, WJ-MSCs were co-cultured with or without MMP inhibitors for 24 h in serum-free medium.

MSCs Administration
Mdx (C57BL/10ScSn-Dmdmdx/J(mdx) (JAX#001801)) and wild-type (C57BL/10ScSnJ (JAX#000476)) mice were purchased from Jackson Laboratory (Bar Harbor, ME, USA), and the offspring of the mice were bred and maintained according to the protocols recommended by the Jackson Laboratory.
Two-to five-month-old mdx mice were placed in a restraining device and injected with a total of 1 × 10 5 WJ-MSCs (suspended in 100 µL of phosphate buffered saline) through the lateral tail vein. Mice were sacrificed using isoprene 7 days after administration.

Antibody Arrays and Enzyme-Linked Immunosorbent Assay
Proteins secreted in the conditioned media were analyzed using RayBio Biotin Label-Based Human Antibody Array 507 (#AAH-BLG-1-2, RayBiotech Life, Peachtree Corners, GA, USA). All slides were scanned using a GenePix 4100A scanner and analyzed using GenePix Pro 7.0 software (Molecular Devices, San Jose, CA, USA). All data were normalized using internal controls. To determine the concentration of MMP-1 protein secreted in conditioned media, an enzyme-linked immunosorbent assay (#ELH-MMP1-1, RayBiotech Life, Peachtree Corners, GA, USA) was performed according to the recommended protocol.

Histological Analysis
For staining, myotubes were fixed with 4% paraformaldehyde after co-culturing with WJ-MSCs or treatment with MMP-1 protein. Picro-Sirius red staining (Abcam, Cambridge, UK) was performed to detect collagen deposition in each group, and the image was analyzed using light microscopy, with quantification using Image J (NIH, New Bethesda, MD, USA).
The gastrocnemius muscles of mice were fixed in 4% paraformaldehyde and then embedded in paraffin, sectioned at 4 µm, and stained with Sirius red staining solution or Alizarin red staining solution. Sirius red staining was used to detect fibrosis, and Alizarin red to detect large accumulations of calcium/necrotic fibers, according to standard procedures. Analysis was conducted using a ScanScope AT (Leica Microsystems, Buffalo Grove, IL, USA), and the images were quantified using Image J.

Statistical Analysis
All results were analyzed by t-tests or one-way ANOVA followed by Duncan's multiple comparison using SPSS software (version 23.0; IBM, Armonk, NY, USA). The differences were considered statistically significant at p < 0.05, and data are expressed as mean ± SEM.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.