Article Text

Download PDFPDF

The dual inhibitor of lipoxygenase and cyclooxygenase ML3000 decreases the expression of CXCR3 ligands
  1. C Ospelt1,
  2. M Kurowska-Stolarska1,
  3. M Neidhart1,
  4. B A Michel1,
  5. R E Gay1,
  6. S Laufer2,
  7. S Gay1
  1. 1
    Center of Experimental Rheumatology, University Hospital Zurich and Zurich Center for Integrative Human Physiology (ZIHP), Switzerland
  2. 2
    Institute of Pharmacy, Eberhard-Karls University, Tuebingen, Germany
  1. Caroline Ospelt, MD, Center of Experimental Rheumatology, University Hospital, Gloriastrasse 25, CH-8091 Zurich, Switzerland; Caroline.Ospelt{at}


Objective: To find previously unknown properties of ML3000, a competitive inhibitor of the cyclooxygenase and the lipoxygenase (LO) pathway.

Methods: Gene expression of ML3000 treated and untreated rheumatoid arthritis synovial fibroblasts were measured with Affymetrix gene arrays. Downregulation of chemokine (C-X-C motif) ligands CXCL9, CXCL10 and CXCL11 was verified with Real-time polymerase chain reaction, CXCL10 protein levels were determined with ELISA. Rheumatoid arthritis synovial fibroblasts were treated with the cyclooxygenase inhibitor naproxen, the 5-LO inhibitor BWA4C and the 5-lipoxygenase-activating protein (FLAP) inhibitor MK886, and consecutive changes in CXCL10 protein levels measured. 5-LO expression was determined by polymerase chain reaction and Western blot.

Results: In synovial fibroblasts and monocyte-derived macrophages ML3000 inhibited the tumour necrosis factor induced expression of CXCL9, CXCL10 and CXCL11, which are all ligands of the chemokine receptor CXCR3. No effect was observed in monocytes. Whereas inhibition of the cyclooxygenase pathway or the FLAP protein showed no effect, blockade of 5-LO significantly downregulated CXCL10 protein levels. 5-LO mRNA was detected in monocytes and in monocyte-derived macrophages. All tested cell types expressed 5-LO protein.

Conclusions: ML3000 effectively downregulates CXCR3 ligands. This study confirms that a thorough analysis of the impact of a drug on its target cells cannot only reveal unexpected properties of a substance, but also helps to understand the underlying molecular mechanisms. Accordingly, our data provide the basis for further clinical studies testing the application of ML3000 in diseases such as rheumatoid arthritis or multiple sclerosis.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

ML3000 is an analogue of arachidonic acid and thereby blocks the conversion of arachidonic acid to prostaglandins by cyclooxygenases (COX) and to leucotrienes by 5-lipoxygenases (5-LO).1 As ML3000 tightly fits in the active site of the enzymes COX 1, 2 and 5-LO, the recruitment and processing of arachidonic acid is inhibited.2 This dual inhibition leads to an equal reduction of prostaglandins and leucotrienes, which are potent mediators in a variety of inflammatory diseases. Accordingly, ML3000 showed good efficacy against inflammation and hyperalgesia in various animal models of joint inflammation.3 4 Furthermore, it has been suggested that by single inhibition of the COX pathway, the formation of leucotrienes is enhanced due to a shunt mechanism in the arachidonic acid metabolism, resulting in increased gastrointestinal damage and ulceration.1 In accordance with this hypothesis, the balanced reduction of prostaglandins and leucotrienes by ML3000 lead to a better gastrointestinal tolerability compared with classical non-steroidal anti-inflammatory drugs (NSAIDs) in human healthy volunteers.5 Recently, it was also described that a balanced, dual inhibition could have a more favourable cardiovascular profile than selective COX-2 inhibitors.6

In previous reports, we suggested the use of molecular biology methods in drug development and assessment.7 8 The analysis of changes in gene expression before and after treatment can possibly reveal molecular mechanisms behind a clinically observed side-effect or point to unexpected properties of a drug that might establish new indications.

In the present study, we analysed the impact of ML3000 on gene expression in cultured, activated synovial fibroblasts (SF) of patients with rheumatoid arthritis (RA). We found that the expression of the chemokines CXCL10 (IP-10), CXCL11 (I-TAC) and CXCL9 (Mig) was downregulated after treatment. All three chemokines are ligands of the chemokine receptor CXCR3. They have been described to play a part in a variety of inflammatory and autoimmune diseases and have repeatedly been discussed as attractive new drug targets.911 The regulation of these chemokines has never been described for either a COX, 5-LO or a dual inhibitor. We found that this property of ML3000 is linked to the inhibition of 5-LO and is influenced by the transcriptional regulation of 5-LO in different cell systems.


Cell culture and media

Synovial tissues from RA and osteoarthritis (OA) patients undergoing joint replacement surgery were minced and digested in 150 mg/ml dispase II (Roche, Mannheim, Germany) at 37°C for 60 min. All patients with RA fulfilled the American College of Rheumatology criteria for the classification of RA. The SFs were then grown in DMEM (Gibco Invitrogen, Basel, Switzerland) supplemented with 10% heat inactivated fetal calf serum (FCS), 50 IU/ml penicillin-streptomycin, 2 mM l-glutamine, 10 mM HEPES and 0.2% fungicide (all Gibco Invitrogen). For experiments cultured SFs were used after four to nine passages.

Peripheral blood mononuclear cells were isolated from Buffy coats of healthy volunteers with Ficoll-Paque PLUS (Amersham Biosciences, Uppsala, Sweden) gradient centrifugation. Peripheral blood monocytes were then positively separated with CD14 microbeads (Miltenyi Biotec, Auburn, CA, USA) according to the manufacturer’s protocol. Monocyte-derived macrophages (MDMs) were generated by treatment of the isolated peripheral blood monocytes with 15 ng/ml macrophage colony-stimulating factor (R&D Systems, Minneapolis, MN, USA) every 48 h for 7 days. MDMs and monocytes were cultivated in RPMI 1640 (Gibco Invitrogen) supplemented with 10% FCS, 50 IU/ml penicillin–streptomycin, 2 mM l-glutamine, 10 mM HEPES and 0.2% fungicide.

All cell cultures were maintained in a 5% CO2 humidified incubator at 37°C.


ML3000 (Merckle GmbH, Blaubeuren, Germany) was reconstituted with DMSO to a stock concentration of 45 mg/ml, naproxen to a concentration of 150 mg/ml, BWA4C to 1.4 mg/ml and MK886 to 2.5 mg/ml (all Sigma, Basel, Switzerland). The drugs were further diluted with 2% human serum albumin (1:50), 0.5% human serum albumin (1:10), and with the corresponding medium without antibiotics, fungicides and with 0.5% FCS to the final working concentration. For control, DMSO alone was diluted correspondingly. Viability of RA SFs (RASF) after treatment with the above-mentioned drugs was assessed by trypan blue exclusion and by luminometric measurement of adenosine triphosphate with the ViaLight MDA Plus kit according to the manufacturer’s instructions (Cambrex, Taufkirchen, Germany).

In the chosen working concentrations all the drugs showed comparable effectiveness; ie, 3 μg/ml ML3000, 0.05 μg/ml MK886 and 0.03 μg/ml BWA4C similarly reduced the production of leucotriene B4 in MDMs by 60–65%, and 3 μg/ml ML3000 and 30 μg/ml naproxen similarly reduced prostaglandin E2 production in SFs by 85–90%. Leucotriene and prostaglandin production was measured after stimulation with the calcium ionophore A23187 (Sigma) (2.5 μM, 15 min) using ACE Competitive Enzyme Immunoassay (Cayman Chemical, Ann Arbor, MI, USA) according to the manufacturer’s instructions. Absorption was measured at 405 nm and data were analysed using the data analysis spreadsheet available from the Cayman Chemical web page (, accessed 10 June 2007). In the concentrations used, ML3000 specifically inhibits of COX and 5-LO, but has no effect on the activity of 12-LO or 15-LO (communication with Stefan Laufer).

Cultured cells were pre-stimulated with 10 ng/ml tumour necrosis factor (TNF)-α or 100 U/ml interferon (IFN) γ (both R&D Systems) for 24 h and then treated with the corresponding drugs in addition with TNF-α or IFN-γ as indicated for 48 h. During treatment cells were cultivated in their respective medium without antibiotics or fungicide, supplemented with 0.5% FCS.

Affymetrix gene chip

Total RNA from TNF-α stimulated RASF, incubated with or without ML3000 (10 μg/ml) was isolated with the RNeasy MiniPrep Kit (Qiagen, Basel, Switzerland) including treatment with RNase-free DNase. The quality of the RNA was checked with the Agilent 2100 Bioanalyser (Agilent Technologies, Palo Alto, CA, USA). Double-stranded cDNA was synthesised from 5 μg RNA using the SuperScript cDNA synthesis customer kit (Invitrogen, Basel, Switzerland). Convertion to biotin-labelled cRNA was done with the BioArray HighYield RNA labeling kit (Enzo Diagnostics, Farmingdale, NY, USA). The labelled RNA was hybridised on human genome U133A oligonucleotide probe arrays (Affymetrix, Santa Clara, CA, USA), according to standard protocols. Data were normalised (set measurements <0.01 to 0.01 and per chip normalisation to 50th percentile) and analysed with GeneSpring microarray analysis software (Silicon Genetics, Redwood City, CA, USA). Filters were set on a fivefold regulation, flags had to be present or marginal in at least one of two compared samples.

Real-time polymerase chain reaction

Total RNA was isolated with the RNeasy MiniPrep kit, including treatment with RNase-free DNase (Qiagen) and reverse transcribed using random hexamers and multiscribe reverse transcriptase (both Applied Biosystems, Rotkreuz, Switzerland). Non reverse transcribed samples were used as negative controls. Quantification of mRNA was performed by single-reporter Real-time polymerase chain reaction (PCR) using the ABI Prism 7700 Sequence Detection system (Applied Biosystems). Eukaryotic 18S rRNA levels, measured with a pre-developed primer/probe system (Applied Biosystems) were used as endogenous control for relative quantification. The differences of the comparative threshold cycle (Ct) values of sample and 18S cDNA were calculated (dCt). Relative expression levels were calculated following the formula ddCt = dCt (sample stimulated)– dCt (sample unstimulated), relative expression was calculated using the expression 2−ddCt. Only samples with a difference of at least four cycles between cDNA and non-reverse, transcribed samples were considered for calculations.

Primers were designed with Primer 3 software (, accessed 1 June 2007). Dissociation curve analysis for each SYBR green primer pair and reaction was performed to verify specific amplification. CXCL9: forward primer 5′-GGC ATC ATC TTG CTG GTT CT-3′, reverse primer 5′-TCA CTA CTG GGG TTC CTT GC-3′; CXCL10: forward primer 5′-AAC CAG AGG GGA GCA AAA TC-3′, reverse primer 5′-CTG TGT GGT CCA TCC TTG G-3′; CXCL11: forward primer 5′-TGG CAA CAG TGC ACA TAT TTC A-3′, reverse primer 5′-CAA ATT AAG ACC GGT GCT GCT A-3′.

Enzyme-linked immunoabsorbent assay

CXCL10 protein was detected in the supernatant using DuoSet ELISA Development Systems (R&D Systems) according to the manufacturer’s instructions. Absorption was measured at 450 nm and data were analysed using Revelation v4.22 software (Dynex Technologies, Denkendorf, Germany).

Conventional polymerase chain reaction

Total RNA was isolated as described above for real-time PCR. Conventional PCR was performed on a GenAmp PCR System 9700 (Applied Biosystems) with the following primer pairs and protocols. 5-LO:12 forward primer 5′-TAC ATC GAG TTC CCC TGC TAC-3′, reverse primer 5′-GTT CTT TAC GTC GGT GTT GCT-3′; β-microglobulin forward primer 5′-AAG ATT CAG GTT TAC TCA CGT C-3′, reverse primer 5′-TGA TGC TGC TTA CAT GTC TCG-3′; 94°C 5 min, 35 cycles with 94°C 30 s, 57°C (for 5-LO)/55°C (for β-microglobulin) 30 s, 72°C 30 s and a final elongation of 5 min with 72°C. Reaction products were separated on a 1% agarose gel and signals were visualised using ethidium bromide. As a negative control, PCR was carried out in the absence of cDNA for each set of primers.

Western blot

Cells were lysed in 2× Laemmli buffer, denaturated for 3 min at 95°C and separated on 10% SDS–PAGE. After transfer on to Protran nitrocellulose transfer membranes (Schleicher & Schüll, Dassel, Germany) and blocking with 5% non-fat dry milk, membranes were incubated with mouse anti-human 5-LO antibodies (BD Transduction Laboratories, San Diego, CA, USA) overnight at 4°C (dilution 1:250). Signals were visualised by incubation with peroxidase-conjugated goat anti-mouse antibodies (Jackson ImmunoResearch, West Grove, PA, USA) and enhanced chemoluminescence system (Amersham Biosciences, Otelfingen, Switzerland). For normalisation, membranes were stripped and probed with mouse anti-human α-tubulin antibodies (Sigma).

Statistical analysis

Values are presented as mean (SEM); n = number of stimulations. Wilcoxon matched-pairs signed-ranks test was used for statistical evaluation of the data by SPSS software; p<0.05 was considered significant.


Downregulation of CXCR3 ligands by ML3000

To screen for genes that are regulated by treatment with ML3000, we used an Affymetrix high-density oligonucleotide array to compare gene expression in ML3000-treated RASF versus untreated RASF, both stimulated with TNF-α. For the stimulation of SF we chose TNF-α, as this is the main pro-inflammatory cytokine found in joints of patients with RA.

After normalisation and filtering of the data, 28 genes were shown to be more than fivefold downregulated after treatment with ML3000 (data not shown). Most strikingly, seven (25%) of these genes could be associated with IFN regulation. A dose-dependent downregulation of this group was verified with Real-time PCR. Two of the three CXCR3 ligands, namely CXCL10 and CXCL11 were among the genes that were regulated by treatment with ML3000 (fig 1). Using Real-time PCR we also confirmed the downregulation of CXCL9, the third ligand of the CXCR3 receptor (fig 2a). CXCL9 expression was downregulated to a lesser extent than CXCL10 or CXCL11, which could be the reason why it did not appear on the chip. A time course of the downregulation of CXCL9, -10 and -11 is shown in fig 2b. The expression of all three chemokines was strongly induced by TNF-α treatment in SF, in non-stimulated RASF the levels of measured chemokines were under the detection limit. For all further experiments, we used a concentration of 3 μg/ml ML3000, as according to the manufacturing company this concentration approximates the level of ML3000 measured in the joints of patients using oral treatment (unpublished communication).

Figure 1 Dose-dependent downregulation of genes after treatment with ML3000. Rheumatoid arthritis synovial fibroblasts were treated with 1 μg/ml, 3 μg/ml, 9 μg/ml ML3000 or control reagent for 48 h (n = 5). Expression levels were measured by real-time polymerase chain reaction. Values are shown as mean percentage expression (SEM) as compared with control.
Figure 2 (A) Downregulation of CXCL9 mRNA after treatment with ML3000 in stimulated rheumatoid arthritis synovial fibroblasts (RASF; n = 8). (B) Time course of the effect of ML3000 on the expression of CXCL9, 10 and 11 (n = 3). Downregulation of CXCL9, CXCL10 and CXCL11 mRNA after treatment with ML3000 in stimulated (C) OA SF (n = 5), (D) monocytes (n = 5) and monocyte-derived macrophages (MDMs; n = 6). Expression levels were measured by real-time polymerase chain reaction. TNF, tumour necrosis factor; IFN, interferon.

To see if the changes in CXCR3 ligand gene expression were disease specific, we performed the same experiments using SF from patients with OA. No disease-specific regulation by ML3000 was detectable, the downregulation of all three measured chemokines was in the same range as in RASF (fig 2c).

To analyse whether the effect of ML3000 is limited to fibroblasts, we treated other cell types that are known to express CXCL9, CXCL10 and CXCL11. In monocytes and macrophages the main stimulus for the expression of CXCL9, CXCL10 and CXCL11 is IFN-γ. Therefore, we stimulated monocytes and MDMs with IFN-γ and treated them with 3 μg/ml ML3000. The effect of ML3000 between monocytes and MDMs was different. Whereas the downregulation in macrophages (MDMs) was similar to SF, only a small effect could be seen in monocytes (fig 2d).

In the further analysis, we concentrated on CXCL10 with regard to its well described role in a variety of pathological status including RA. Even though also CXCL9 and CXCL11 seem to be important factors, their specific roles are much less defined. Protein levels of CXCL10 were measured after treatment with ML3000. The results confirmed the data obtained at the mRNA level. Whereas 3 μg/ml ML3000 was enough to significantly reduce CXCL10 production in SFs and MDMs, no significant change could be observed in monocytes (fig 3).

Figure 3 Regulation of CXCL10 protein levels as measured by ELISA in supernatants of stimulated rheumatoid arthritis synovial fibroblasts (RASFs; n = 7), osteoarthritis synovial fibroblasts (OASF; n = 5), monocytes (n = 4) and monocyte-derived macrophages (MDMs; n = 6) after treatment with 3 μg/ml ML3000. (A) Values are shown as mean percentage expression (SEM) as compared with control. *p<0.05. (B) Levels of CXCL10 in each individual measurement are shown before and after treatment.

Mechanisms of CXCR3 ligand regulation

To find out whether the blockade of the COX or the 5-LO pathway is responsible for the described effects of ML3000, we treated TNF-α stimulated RASF with the COX inhibitor naproxen, the 5-LO inhibitor BWA4C and with the inhibitor of the 5-LO activating protein (FLAP) MK886. The COX inhibitor naproxen, had no effect on CXCL10 production by TNF-α-stimulated RASF. Treatment with the FLAP inhibitor only slightly reduced the production of CXCL10 protein in some of the tested patient cells; therefore, no statistical significance was reached. In contrast, downregulation of CXCL10 protein levels after treatment with BWA4C was consistent and reached statistical significance (fig 4). The combination of either 5-LO inhibitor, BWA4C or MK886 with naproxen, had the same effect on the production of CXCL10 as treatment with the 5-LO inhibitors alone (n = 2; data not shown). Thereby, we concluded that inhibition of CXCR3 ligand production by ML3000 must be an effect, that is independent of its ability to block the COX pathway, but is linked to its blockade of 5-LO.

Figure 4 Changes of CXCL10 protein levels in the supernatants of stimulated rheumatoid arthritis synovial fibroblasts (RASFs) after treatment with 30 μg/ml naproxen, 0.05 μg/ml MK886, 0.03 μg/ml BWA4C and 3 μg/ml ML3000 as measured by ELISA (n = 5). All values are shown as mean percentage expression (SEM) as compared with control. *p<0.05.

Furthermore, mRNA and protein levels of 5-LO were measured. We could detect a strong band for 5-LO mRNA in monocytes, only a weak band in MDMs and no band was visible in SF (fig 5a). All measured cell types had detectable levels of 5-LO protein, albeit they were noticeably higher in monocytes compared with MDMs or SF (fig 5b). Thus, we assume that different regulation of 5-LO transcription and de novo synthesis in the analysed cell types could influence their sensitivity to treatment with ML3000 in respect to the measured outcomes.

Figure 5 (A) 5-LO mRNA expression in RASF (n = 4), OASF (n = 2), monocytes (n = 5) and MDMs (n = 3). One representative polymerase chain reaction is shown. (B) 5-LO protein (79 kDa) expression in cell lysates of RASF, OASF, monocytes and MDMs. RASF, rheumatoid arthritis synovial fibroblast; OASF, osteoarthritis synovial fibroblast; 5-LO, 5-lipoxygenase; MDMs, monocyte-derived macrophages.


In the present study, we analysed the impact of ML3000 on the transcriptome of SFs and found a previously unknown, anti-inflammatory action. Thereby, we used microarray technology as starting point for further experiments, characterising the newly found properties in-depth. This simplified microarray approach is not motivated by the generation of a complete gene expression profile with clusters of genes that correlate with a specific condition like in large-scale microarray experiments, but serves as a basic screening technique for novel, unexpected properties of a drug.

In our screening, we found evidence that ML3000 has an impact on the transcriptional regulation of the chemokines CXCL9, CXCL10 and CXCL11, all three ligands of CXCR3. This effect was measurable on the mRNA as well as on the protein level and was not restricted to SF or stimulation with TNF-α. Chemokines in general play a fundamental role in the migration of leucocytes to the site of inflammation. In the subgroup of CXC chemokines, the CXCR3 ligands take an exceptional position. While most CXC chemokines are clustered at chromosome 4q12–13 and act on neutrophils, CXCL9, CXCL10 and CXCL11 are found at 4q21.21 and mainly attract T lymphocytes.13 High levels of CXCL9, 10 and 11 could be detected in the synovial fluid of patients with RA.1417 They are mainly produced by fibroblasts and macrophages in the synovium and are thought to be involved in the recruitment of T helper 1 cells into the joint (for review18). CXCR3 ligands were also found to play a crucial part in the accumulation of activated lymphocytes in the brain of patients with multiple sclerosis.19 Moreover, it is suggested that T cell attraction in autoimmune liver disease, myasthenia gravis and acute renal allograft rejection is mediated by CXCR3 and its ligands.2022 Thus, this group of chemokines has been suggested as drug targets in a variety of diseases.22 23 In fact, the application of a decoy chemokine receptor DNA for the binding site of CXCR3 was shown to suppress the relapse of experimental autoimmune encephalomyelitis in rats.24 Therefore, our data provide the basis for further experiments testing the application and effectiveness of ML3000 in diseases, such as multiple sclerosis and acute allograft rejection, and thus a new field of indications for ML3000 could be opened up.

We found that the ability of ML3000 to inhibit CXCR3 ligand production does not stem from its ability to block the COX pathway, but seems to be linked to the inhibition of 5-LO. In accordance with our experiments, it was previously described in macrophages and cancer cells that prostaglandin E2 can suppress CXCL10 production and consequently, inhibition of the COX pathway rather enhances production of CXCL10.25 26 Stimulation of SFs with prostaglandin E2 and/or leucotriene B4 had no effect on the expression of either CXCL9, 10 or 11, which remained under the detection limit (data not shown). Additionally there was no difference between the tested 5-LO inhibitors in their potential to inhibit leucotriene B4 synthesis in monocytes. Therefore the downregulation of CXCR3 ligands is probably independent of the role of 5-LO in leucotriene biosynthesis. We assume that inhibition of the production of this specific group of chemokines could be a direct effect of inhibiting other functions of the 5-LO protein. Functions of 5-LO beyond the leucotriene pathway have been suggested before and are supported by studies showing 5-LO to localise at sites of active gene transcription in the nucleus even in cells that are not producing leucotrienes.27 By means of a yeast two-hybrid system 5-LO was found to interact with the transforming growth factor type β receptor-I-associated protein 1 (TRAP-1). It was suggested that via TRAP-1, 5-LO is associated with the activated transforming growth factor-β receptor and modulates its signalling.28 However, up to now the association between the expression of CXCL10 and transforming growth factor β signalling in fibroblasts has not been described yet, and further studies have to be conducted to elucidate these cellular functions of 5-LO.

Regulation, localisation and translocation of 5-LO have been found to be highly variable between different cell types.29 In particular it has been shown that while monocytes express high amounts of 5-LO mRNA, this production is lost during the maturation to MDMs30. In resting cells 5-LO is stored, and only upon activation translocated and bound to membranes.31 These findings can explain why we could only detect mRNA for 5-LO in monocytes, but not in MDMs or SFs. In contrast, stored 5-LO protein could be found in all tested cell types. Inhibition of 5-LO activity by blocking its active centre could possibly be overcome in monocytes via increased de novo production.

In summary, we show that the COX/5-LO inhibitor ML3000 inhibits the production of the CXCR3 ligands CXCL10, CXCL9 and CXCL11. Furthermore, we found evidence for new cellular roles of 5-LO. Our study proves that by analysing the molecular impact of a drug on its target cells, off-target effects and their underlying molecular mechanisms can be revealed. In our case, off-target effects of the analysed drug turned out to be beneficial, as CXCR3 ligands have repeatedly been suggested as possible drug targets for a variety of inflammatory diseases. Accordingly, animal and clinical studies testing the application of ML3000 in diseases such as RA and multiple sclerosis or acute allograft rejection should be taken into consideration. On the other hand, off-target effects are often the cause of adverse side-effects, which makes it even more important to include such molecular studies early on in the development of a drug.


We thank Maria Comazzi and Ferenc Pataky for their excellent technical assistance. We thank the Functional Genomic Center Zurich for their technical support.


View Abstract


  • Funding: The study was financially supported by a research grant from Merckle, Germany. The company did not exert any influence on study design, analysis and interpretation of the data or in the writing of the report, nor in the decision to submit the paper for publication.

  • Competing interests: None.