Global DNA hypomethylation in rheumatoid arthritis synovial fibroblasts (RASFs) contributes to their intrinsic activation. The aim of this study was to investigate whether increased polyamine metabolism is associated with a decreased level of S-adenosyl methionine (SAM), causing global DNA hypomethylation.
Synovial fibroblasts were isolated from synovial tissue obtained from 12 patients with RA and from 6 patients with osteoarthritis (OA). The cells were stained for S-adenosyl methionine decarboxylase (AMD), spermidine/spermine N1-acetyltransferase (SSAT1), polyamine-modulated factor 1–binding protein 1 (PMFBP1), solute carrier family 3 member 2 (SLC3A2), DNA methyltransferase 1 (DNMT-1), α9 integrin, and β1 integrin and analyzed by flow cytometry. Nuclear 5-methylcytosine (5-MeC) was measured by flow cytometry, the expression of diacetylspermine (DASp) in cell culture supernatants and cell extracts was determined by enzyme-linked immunosorbent assay, and SAM expression in cell extracts was measured by fluorometry.
The expression of SSAT1, AMD, and PMFBP1 was significantly increased in RASFs compared with OASFs. The expression of DASp in cell culture supernatants and the expression of SLC3A2 were significantly elevated in RASFs. The levels of SAM in cell culture extracts, as well as the levels of DNMT-1 protein and 5-MeC, were significantly reduced in RASFs. Parameters of polyamine metabolism were negatively correlated with the expression of SAM, DNMT-1, and 5-MeC.
These data clearly show that intrinsic elevations of PMFBP1 and SSAT1 enhance the catabolism and recycling of polyamines in RASFs and suggest that high consumption of SAM via this pathway is an important factor contributing to global DNA hypomethylation in these cells.
In rheumatoid arthritis (RA), a cytokine-independent pathway appears to be responsible for the ongoing joint destruction mediated by synovial fibroblasts. The pathologic changes associated with this process include excessive hyperplasia and infiltration of inflammatory cells into the synovial tissue. The activated phenotype of RA synovial fibroblasts (RASFs) appears to be an intrinsic property of these cells, because our group was able to demonstrate that RASFs co-implanted with human cartilage into SCID mice are invasive even in the absence of other cells of the human immune system (1). This concept is reflected in vitro by, for example, the increased production of matrix-degrading enzymes and adhesion molecules by RASFs.
In vitro studies suggested that the “imprinted” aggressive phenotype of RASFs is in large part due to global genomic hypomethylation (2). In somatic cells, DNA methylation is achieved by DNA methyltransferase 1 (DNMT-1) converting cytosine into 5-methylcytosine (5-MeC) (3). We previously demonstrated that the level of DNMT-1 is deficient in activated RASFs compared with that in osteoarthritis (OA) synovial fibroblasts (OASFs) under the same conditions, e.g., in the presence of proinflammatory cytokines such as interleukin-1β (IL-1β) or tumor necrosis factor α (TNFα) (2). The level of DNMT-1 transcript was normal, but the protein level was lower, suggesting enhanced degradation in proteosomes. In addition, we reported that inhibition of DNMT-1 activity by long-term treatment of normal synovial fibroblasts with a nontoxic dose of 5-azacytidine induced an activated phenotype similar to that observed in RASFs, i.e., with increased expression adhesion molecules, receptors for growth factors, and cytokines as well as matrix-degrading enzymes (2).
In the present study, we investigated the cause of DNA hypomethylation in RASFs. S-adenosyl methionine (SAM) is the methyl donor during DNA methylation (3). We hypothesized that excessive consumption of SAM by other pathways, in particular the metabolism of polyamines, might contribute to the observed global DNA hypomethylation (Figure 1). Global hypomethylation was the first described epigenetic change in RASFs. However, it is only one of several mechanisms of epigenetic regulation involved in RA that also involve histone modifications and microRNA (4). Together, these epigenetic changes might cause stable activation of RASFs and alter the inflammatory response, thereby promoting the development of chronic disease.
Biosynthesis of polyamines is tightly regulated by enzymes, including ornithine decarboxylase (ODC), which regulates the conversion of ornithine into putrescine, and spermidine synthase (SdS) and spermine synthase (SmS), which regulate the biosynthesis of spermidine and spermine, respectively (5). In this regard, SdS and SmS need decarboxy-SAM that is provided by SAM after enzymatic conversion mediated by S-adenosyl methionine decarboxylase (AMD). SAM is synthesized from the essential amino acid L-methionine. Putrescine holds the AMD domains in a more active state, inducing realignment of charged residues to the active site and thus enhancing the conversion of SAM (6).
In addition, a recycling pathway that reconverts polyamines into putrescine is under the control of spermidine/spermine N1-acetyltransferase (SSAT1) (5). Moreover, expression of the SSAT1 gene is regulated by polyamine-modulated factor 1–binding protein 1 (PMFBP1) and nuclear factor E2–related factor 2 (7). Various stimuli, including growth factors, hormones, and cytokines, activate both ODC and SSAT1, dramatically increasing the need for SAM, which, under such conditions, is predominantly used in the biosynthesis and recycling of polyamines. Polyamines have an important role in cell proliferation, migration, and transcription, and in stabilization of nucleic acids and proteins. SSAT1 also interacts with the cytoplasmic domain of α9 integrin, which plays a role in cell migration (8).
Proinflammatory cytokines increase ODC and SSAT1 activities in synovial fibroblasts, triggering the synthesis of polyamines and their recycling (9). However, the level of putrescine in RASFs is also elevated in the absence of stimulation. Dysfunction in the regulation of ODC does not cause elevated levels of putrescine, which, instead, could result from enhanced recycling of polyamines, a process that is controlled by SSAT1. During recycling of polyamines, intermediaries are produced in the form of diacetylpolyamines, which can be transported outside of cells through an exporter system involving solute carrier family 3 member 2 (SLC3A2)/CD98 (10). Urinary diacetylspermine (DASp) is a marker for cancer and liver diseases. The levels of urinary acetylpolyamines are also increased in patients with RA compared with patients with OA or healthy control subjects (11, 12).
Increased SSAT1 activity produces more putrescine and diacetylpolyamines (5, 13). High putrescine levels, in turn, enhance the synthesis of AMD, which is responsible for the conversion of SAM into AMD. Thus, overexpression of AMD and/or SSAT1 may result in a decrease in SAM, which is the methyl donor for the cell.
In the present study, we show that an intrinsically activated PMFBP1/SSAT1-dependent pathway could be an important factor leading to global DNA hypomethylation in RASFs.
PATIENTS AND METHODS
Cells were obtained from 12 patients with RA (6 men and 6 women, mean ± SD age 63 ± 7 years, disease duration 16 ± 6 years), all of whom fulfilled the American College of Rheumatology criteria for the classification of RA (14) and were rheumatoid factor positive. The patients were treated with the following disease-modifying antirheumatic drugs: steroids (n = 4), methotrexate (n = 3), leflunomide (n = 2), hydroxychloroquine (n = 1), and sulfasalazine (n = 1). Cells were also obtained from 6 patients with OA (3 men and 3 women, mean ± SD age 66 ± 6 years) and were used as normal control fibroblasts.
RASFs and OASFs were isolated from synovial tissue obtained during joint replacement surgery. The cells were cultured in Dulbecco's modified Eagle's medium (DMEM) including 10% fetal calf serum (FCS) and were used between passage 4 and passage 6; the reagents were obtained from Life Technologies. The procedure was approved by the local ethics committee (University Hospital Zurich).
Characterization of cells by flow cytometry.
Confluent cultures of synovial fibroblasts were passaged (1:4), kept for 3–4 days before being detached with Accutase (Innovative Cell Technologies), stained, and analyzed by flow cytometry. The following primary antibodies were diluted 1:10 in DMEM including 10% FCS and kept at 4°C: mouse polyclonal antibodies to PMFBP1 (Abcam), mouse monoclonal antibodies to SLC3A2/CD98 (clone 4F2; Abcam) and to DNMT-1 (clone 60B1220; Abcam), rabbit polyclonal antibodies to SSAT1 and AMD1 (Abcam), and chicken polyclonal antibodies to SmS and to α9 integrin (Abcam). The secondary antibodies that were used included fluorescein isothiocyanate (FITC)–conjugated goat polyclonal antibodies to mouse IgG/IgM (1:100; BD PharMingen), FITC-conjugated goat polyclonal antibodies to the rabbit IgG F(ab′)2 fragment (1:100; Abcam), and FITC-conjugated goat polyclonal antibodies to chicken IgY (heavy and light chains) (1:25; Abcam). The expression of β1 integrin was determined using phycoerythrin-conjugated mouse anti-human CD29 antibodies (1:10; BD PharMingen). To determine the expression of AMD, SSAT1, PMFBP1, SmS, and DNMT-1, the cells were treated with FACS Permeabilizing Solution 2 (BD PharMingen) during staining with the first and second antibodies. The cells were analyzed using a FACSCalibur (BD Biosciences). Murine IgG1, murine IgG2a, and chicken IgY served as isotype controls for the primary antibodies. The side scatter/forward scatter gate on the flow cytometer was set to analyze large cells and to exclude small cells and debris. Values were expressed as the mean fluorescence intensity (MFI).
Measurement of nuclear 5-MeC by flow cytometry.
Cells (106) were fixed in 1% paraformaldehyde for 10 minutes at 37°C and kept on ice for 10 minutes before the addition of 88% methanol/12% phosphate buffered saline (PBS) and incubation overnight at −20°C. The nuclei were centrifuged at 3,000g for 5 minutes, washed twice with PBS containing 0.01% Tween 20 (PBST) and 1 μg/ml bovine serum albumin (BSA), and treated with 1N HCl for 40 minutes at 37°C. The acid was neutralized by one washing step with 0.1M borate buffer, pH 8.5, and 2 washing steps with PBST–BSA. The nuclei were incubated for 20 minutes at 37°C with a blocking solution (PBS–BSA containing 10% FCS and 10 μg/ml RNase). Next, the nuclei were incubated with murine monoclonal anti–5-MeC antibodies (Imgenex) for 1 hour at 37°C, washed twice with PBS, and incubated with FITC-conjugated anti-mouse antibodies (BD Biosciences) for 1 hour at 22–24°C. Finally, the samples were stained with propidium iodide before analysis by flow cytometry (FACSCalibur). The background fluorescence (i.e., without anti–5-MeC antibodies) was subtracted, and values were expressed as the MFI.
Confluent cell cultures were passaged (1:4), kept for 3 days in DMEM containing 10% FCS, detached using Accutase (PAA Laboratories), centrifuged, and lysed (using cell lysis buffer from Cell Signaling Technology). The protein concentration in the cell extract was assessed using a Pierce BCA Protein Assay Kit, and the extracts were diluted with distilled water to 100 μg/ml.
Enzyme-linked immunosorbent assay (ELISA) for diacetylspermine.
DASp was measured in cell culture supernatants and cell extracts using a commercially available ELISA kit (BioVendor). The values for cell supernatants were expressed as nanomolar per 106 cells, and the values for cell extracts were expressed as nanomolar per 100 μg of protein.
Fluorometry for S-adenosyl methionine.
For the SAM assay, reagents from the commercially available methyltransferase HT Activity Kit (Assay Designs), including the histone methyltransferase SET-7/9 and the peptide substrate TAF-10, were used. The reaction mixture was prepared as recommended by the manufacturer of the kit, as follows: 200 μl of SAM-free methyltransferase reaction buffer concentrate was added to 1,700 μl of transferase assay buffer. Serial dilutions of SAM (80 nM) were prepared in the reaction mixture; cell extracts were diluted 1:5 in the reaction mixture. SET-7/9 (0.5 μM) was suspended in transferase activity buffer, while TAF-10 (15 μM) was suspended in DMSO. SET-7/9 (25 μl) was added to each well of a black microtiter plate; 100 μl of TAF-10 and 25 μl of standard or diluted cell extract was added in duplicate to the plates. The plates were incubated at 22–24°C for 30 minutes. Finally, 50 μl of 1 mMN-ethylmaleimide was added to stop the reaction. The emission at 520 nm was read using an FLx800 Fluorescence Microplate Reader (BioTek Instruments). Average relative fluorescence was plotted against SAM concentrations to generate a standard curve. For cell extracts, the background (without SET-7/9) was subtracted and the value multiplied by the dilution factor. Values were expressed as nanomolar per microgram of protein. This assay was not influenced by the amount of S-adenosyl homocysteine in the sample.
Differences between RASFs and OASFs were examined using the Mann-Whitney U test. Wilcoxon's signed rank test was used to evaluate the effects of treatment. Spearman's rank correlation was used to describe possible relationships between the variables. P values less than 0.05 were considered significant.
Increased catabolism and/or recycling of polyamines in RASFs.
In RASFs, compared with OASFs, the levels of DASp in cell extracts (DASp-ex) were stable, whereas the concentration of DASp in the culture supernatant (DASp-sup) was increased (Figures 2A and B and Table 1). Thus, the DASp-sup:DASp-ex ratio was significantly increased in RASFs, reflecting active excretion (Table 1). In addition, compared with OASFs, RASFs expressed more SLC3A2 on the cell surface (Figure 2C). DASp-ex expression was inversely related to the concentration of DASp-sup (Figure 1D). When only RASFs (in which the levels of DASp-sup were increased) were considered, the inverse correlation became highly significant (r = −0.83, P < 0.001). Furthermore, SLC3A2 expression was significantly correlated with DASp-sup (Figure 2E). In addition, DASp-sup showed an inverse and significant correlation with the amount of intracellular DNMT-1 (Figure 2F). These findings do not mean that DASp in the cell culture supernatant has a direct effect on the expression of DNMT-1 but suggest a link between the metabolism of polyamines and DNA methylation.
Table 1. Levels of enzymes and substrates involved in polyamine metabolism and DNA methylation in OASFs and RASFs*
Values are the mean ± SD. Synovial fibroblasts were obtained from 6 patients with osteoarthritis (OASFs) and from 12 patients with rheumatoid arthritis (RASFs). SSAT1 = spermidine/spermine N1-acetyltransferase; MFI = mean fluorescence intensity; AMD = S-adenosyl methionine decarboxylase; PMFBP1 = polyamine-modulated factor 1–binding protein 1; DASp-sup = diacetylspermine (DASp) in cell culture supernatants; SLC3A2 = solute carrier family 3 member 2; DASp-ex = DASp in cell extracts; SmS = spermine synthase; SAM = S-adenosyl methionine; DNMT-1 = DNA methyltransferase 1; 5-MeC = 5-methylcytosine.
By Mann-Whitney U test.
17.17 ± 3.54
34.19 ± 8.21
23.51 ± 3.69
35.88 ± 5.34
49.69 ± 7.15
79.01 ± 27.66
DASp-sup, nM/106 cells
10.73 ± 2.08
15.56 ± 3.25
15.29 ± 4.47
24.15 ± 6.38
1.94 ± 0.50
4.92 ± 3.31
DASp-ex, nM/100 μg protein
5.77 ± 0.83
4.93 ± 2.20
12.54 ± 1.94
13.50 ± 2.68
SAM, nM/μg protein
3.10 ± 0.25
0.95 ± 0.94
13.34 ± 1.01
7.21 ± 1.33
2.52 ± 0.37
0.98 ± 0.26
β1 integrin, MFI
16.28 ± 2.58
25.48 ± 3.90
α9 integrin, MFI
17.96 ± 3.09
17.85 ± 3.98
In RASFs, compared with OASFs, the intracellular levels of AMD, SSAT1, PMFBP1, and SLC3A2, as measured by flow cytometry in permeabilized cells, were significantly increased (Figure 3 and Table 1). The levels of SmS in permeabilized cells were not different in RASFs compared with OASFs. These data suggest an intrinsic increase in the catabolism and/or recycling of polyamines and increased transport outside of the cells in the form of diacetylpolyamines.
Deficient levels of SAM and DNMT-1, and global hypomethylation in RASFs.
We developed a fluorometric assay for SAM (Figure 4A). In RASFs, compared with OASFs, the levels of SAM in cell culture extracts were significantly decreased (Figure 4B). The level of SAM in RASFs was 3.3-fold (−69%) lower than that in OASFs. The levels of SAM were negatively correlated with parameters of polyamine metabolism, such as SSAT1, and were positively correlated with parameters of DNA methylation, such as 5-MeC (Figures 4C and D, respectively).
We further analyzed the relationship between these variables, using all values obtained in OASFs and RASFs. In general, parameters of polyamine metabolism were negatively correlated with SAM, DNMT-1, and 5-MeC. This observation suggested that increased consumption of SAM by enhanced catabolism and/or recycling of polyamines is an important factor related to global DNA hypomethylation in RASFs.
Our data showed that a deficiency of SAM in RASFs is an important factor in the context of DNA hypomethylation. The levels of SAM in cell culture extracts, as well as the amounts of DNMT-1 and 5-MeC measured by flow cytometry, were significantly decreased in RASFs compared with OASFs. Regarding DNMT-1 protein and 5-MeC in the nuclei, the present results confirm our previous report (2). In the current study, we observed a positive correlation between the amount of SAM in cell extracts and 5-MeC in the nuclei. An association between SAM deficiency and decreased DNA methylation has also been reported in Alzheimer's disease (15–17). In such studies, the levels of SAM were measured using liquid chromatography tandem mass spectrometry (LC-MS/MS). The fluorometric assay described here is faster and more economical. The levels measured in OASFs (∼2.5–3.5 nM SAM/μg protein) were comparable with those determined by LC-MS/MS in human and murine brains (∼2.0–3.0 nM/μg protein) (18).
In RASFs, DNMT-1 protein is reduced in spite of normal transcript levels, suggesting accelerated degradation (2). The deficiency of SAM in RASFs possibly also favors the degradation of DNMT-1. Indeed, we report here that supplementation with SAM at least partially restored DNMT-1 protein to normal levels. Recently, an age-dependent decrease in SAM due to an increased conversion into homocysteine and a concomitant decrease in DNMT-1 in T cells have been reported (19, 20). The stability of DNMT-1 is dependent on its interaction with its substrates, i.e., SAM and DNA; these induce conformational changes that activate the catalytic center (3) and result in a stabilization signal, avoiding ubiquitination of DNMT-1 (21). In contrast, inactive DNMT-1 is ubiquitinated and degraded in proteosomes (21–23). In addition, increased levels of AMD may result in accumulation of decarboxy-SAM. Although the decarboxy-SAM molecule contains a methyl group, it does not act as a methyl group donor in DNA methylation. Instead, it acts as a competitive inhibitor of DNA methyltransferases (24). Thus, DNMT-1 protein can be affected in different ways.
We searched for an explanation for the 3.3-fold (−69%) decreased expression of SAM in RASFs. Previously, Brooks proposed a hypothesis that some autoimmune diseases occur due to a loss of dose compensation of X-linked polyamine genes at Xp22.1, which impacts intracellular methylation (25). The 2 genes of interest are SmS and SSAT1. Overexpression of SSAT1 increases recycling of polyamines. The resulting decrease in SAM would hamper methylation in the cell, and this could account for the aberrant DNA methylation observed in RASFs. Confirming this hypothesis, we observed that the amounts of AMD and SSAT1 were increased in RASFs. AMD is stabilized by the elevated level of putrescine in the cell (12, 13, 25). In addition, an increase in putrescine can stimulate histone acetylation through enhanced histone acetyltransferase p300/CREB binding protein activity (25). This could contribute to the shifted balance of histone acetylase/histone deacetylase activities toward histone hyperacetylation in RASFs (26). Thus, DNA methylation is not the only mechanism involved in the intrinsic activity of RASFs. Epigenetic gene regulation is complex; it also involves modifications (27) and microRNA (28). These mechanisms interact with each other and may contribute to the activated phenotype of RASFs as well as to the development of chronic disease (4).
Our data suggest that elevated levels of SSAT1 are attributable to the increased expression of PMFBP1. This transcription factor is a tissue-specific gene (e.g., expressed in heart and skeletal muscle but not in brain or colon) (6). Its expression increases in the presence of high levels of spermidine and spermine, activating the recycling and catabolism pathway of these polyamines (6). In RASFs, however, the intracellular levels of spermidine and spermine are “normal” (8). In this context, it is important to note that the majority of polyamines are bound to nucleic acids (especially RNA) and phospholipids, whereas free polyamines account for only a small portion in the cell. The “normal” levels of free polyamines result from a dynamic equilibrium; an interesting question is whether the intracellular levels of free polyamines are affected by the increase in RNA transcripts due to DNA hypomethylation. Furthermore, the PMFBP1 promoter could be affected by DNA hypomethylation (29).
Previously, it was reported that the expression of putrescine, but not spermidine or spermine, is increased in RA synovial tissue and synovial fluid (12). The inflammatory milieu was considered responsible for this increase in polyamine biosynthesis. However, our data suggest that at least in part these increased levels of putrescine are due to the intrinsic activation of the polyamine recycling pathway controlled by SSAT1. The cells export the intermediaries of this pathway, and DASp can be measured in cell culture supernatants; in vivo they can be found in the circulation and are excreted into the urine. It was previously reported (11) that urinary acetylpolyamines in RA correlated with the degree of joint function damage and radiographic progression. Thus, the presence of these acetylpolyamines in RA tissue and synovial fluid not only is a sign of inflammation but also could reflect the activity of aggressive synovial fibroblasts. Thus, because our data showed that recycling and catabolism of polyamines are activated in RASFs independently from external stimuli, it is possible that DASp exported outside of the cell could be a useful biomarker for the activity of RASFs. Alternatively, our data underline the importance of maintaining a stable intracellular level of DASp. A depletion of polyamines occurring after stable transfection of cells with SSAT1 expression vectors led to inhibition of proliferation and DNA damage (30). This was not the case for RASFs, in which putrescine expression was elevated, but the intracellular levels of spermine and spermidine were “normal” (9, 12).
DASp in cell extracts and DASp in the cell culture supernatant were inversely correlated. In addition, the levels of DASp in the supernatants also correlated with the amount of transporter on the cell surface (SLC3A2). Interestingly, the negative correlation between DASp in the supernatant and intracellular DNMT-1 could be explained by the fact that the metabolism of polyamines affects both parameters, and that the level of SAM is the common denominator between polyamines and DNA methylation. In particular, the increased turnover of polyamines led to a depletion of SAM that in turn affected the stability of DNMT-1.
In RASFs, it is possible that a vicious circle exists between the recycling pathway of polyamines and DNA hypomethylation. According to our working hypothesis (Figure 1), SAM deficiency leads to the degradation of DNMT-1 and to DNA hypomethylation that affects multiple genes. The expression of PMFBP1 and SSAT1 is up-regulated, leading to an increased intracellular level of putrescine, which in turn stabilizes AMD, further increasing the conversion of SAM into decarboxy-SAM. The pool of intracellular SAM decreases and is no longer available as a methyl donor. This vicious circle would lead to progressive DNA hypomethylation and changes in the cellular phenotype and contribute to the aggressive behavior of RASFs. Many genes were up-regulated in RASFs by hypomethylation, including cytokines and their receptors, matrix-degrading enzymes, and adhesion molecules (2). In addition, long interspersed nuclear element 1 (LINE-1)–related proteins, which normally are silenced by methylation in synovial fibroblasts, appear at detectable levels in RASFs because of a defective counterregulatory mechanism (31). It could be contended that DNA methylation can change with the passage number. We used cultures between passages 4 and 6 only. However, global hypomethylation in RASFs, as well as the expression of LINE-1 messenger RNA and proteins, are stable and are also detected in later passages. We analyzed the pooled data of RASFs and OASFs, hypothesizing that similar processes would affect any cells that would be deprived of SAM.
It is important to note that in this study, RASFs and OASFs were not cultured in the presence of the proinflammatory cytokines IL-1β and TNFα. It can be expected that these treatments would render the problem more severe, because they stimulate ODC and SSAT1, in addition to promoting proliferation (data not shown). Further studies will clarify the status of DNA hypomethylation in early arthritis as well as the influence of a current therapeutic regimen (e.g., methotrexate that interferes with folate metabolism) as possible confounding factors related to the intrinsic activation of RASFs.
In conclusion, intrinsic up-regulation of PMFBP1 and SSAT1 enhances the catabolism and/or recycling of polyamines in RASFs. The elevated level of AMD appears secondarily to the increased intracellular levels of putrescine that are due to activation of the recycling pathway of polyamines. Our data are consistent with the hypothesis that excessive consumption of SAM by this intrinsically activated pathway is an important factor related to the induction of global DNA hypomethylation in RASFs. Supplementation with SAM and/or inhibition of the PMFBP1/SSAT1 pathway could be promising new therapeutic strategies in RA. The questions for the future will be whether this treatment without concomitant inhibition of SSAT1 is efficient, and whether stable normalization of the phenotype in RASFs can be achieved.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Neidhart had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Karouzakis, Neidhart.
Acquisition of data. Karouzakis, Neidhart.
Analysis and interpretation of data. Karouzakis, R. E. Gay, S. Gay, Neidhart.