Antagonism Between Saturated and Unsaturated Fatty Acids in ROS Mediated Lipotoxicity in Rat Insulin-Producing Cells

The prevalence of type 2 diabetes mellitus increases in parallel to the increasing rate of obesity [1], which is one of the major risk factors for diabetes development and usually accompanied by hyperlipidaemia [2,3,4]. Hyperlipidaemia, in turn, is typically associated with elevated non-esterified fatty acid (NEFA) plasma concentrations, which are supposed to contribute to insulin resistance, impaired insulin secretion due to a progressive decline in pancreatic β-cell function and mass, a phenomenon called lipotoxicity [5,6,7]. In vitro studies have shown that long-chain saturated NEFAs can induce apoptosis [8,9,10,11]. This has been considered, at least in part, to be mediated by the induction of the unfolded protein response (UPR) [12]. However, other observations indicate that the induction of this ER-stress response may not be solely responsible for lipotoxicity [13]. Other studies suggest that the generation of ROS during the metabolism of NEFAs can mediate their deleterious action. This concept comprises ROS formation at complex I and III in the electron transport chain as a result of an overloaded mitochondrial fatty acid metabolism [14,15,16,17]. In a recent study we put forward a new concept in which the β-oxidation of saturated long-chain fatty acids in the peroxisome was suggested to be responsible for ROS formation and death of pancreatic β-cells [9]. Interestingly, unsaturated long-chain fatty acids have been reported not to be toxic [8,18,][19]. Moreover, unsaturated fatty acids efficiently antagonized the toxic action of saturated NEFAs [19,20,21,22]. Therefore, we analysed in the present study the mechanism underlying this protective action with special emphasis upon the role of peroxisomal ROS formation.

RPMI 1640 tissue culture medium, penicillin and streptomycin were purchased from Biochrom (Berlin, Germany). Tissue culture plates were obtained from Greiner Bio-One (Frickenhausen, Germany). Custom designed PCR primers and Taqman probes were synthesized from Life Technologies (Karlsruhe, Germany). Palmitoleic acid was obtained from Larodan (Malmö, Sweden). All other chemicals were purchased from Sigma-Aldrich (Taufkirchen, Germany).

RINm5F insulin-producing cells were cultured in RPMI 1640 medium supplemented with 10 mM glucose, 10% (v/v) foetal calf serum (Biowest, Nuaillé, France), penicillin, and streptomycin in a humidified atmosphere at 37°C and 5% CO₂, as described previously [23]. RINm5F cell clones that overexpressed catalase either in the mitochondria or in the peroxisomes were generated as described previously [24,25]. Cellular expression of catalase was analysed by enzyme activity measurement.

NEFAs (Sigma, St. Louis, MO, USA) were dissolved in 90% ethanol by heating to 60°C and used at different concentrations in RPMI 1640 (PAN, Aidenbach, Germany) with 1% foetal calf serum and a final BSA/NEFA ratio of 2%/1 mM (defined NEFA free BSA, MP Biomedicals, Eschwege, Germany). All untreated wells received the same amount of solvent and BSA. This procedure did not cause a significant decrease in viability in the absence of fatty acids [26].

Pancreatic islets were isolated from 250-300 g adult male Lewis rats by collagenase digestion, separated by Ficoll gradient, and handpicked under a stereo microscope. Isolated islets were cultured on extracellular matrix (ECM)-coated plates (35mm) (Novamed, Jerusalem, Israel, the ECM being derived from bovine corneal endothelial cells) in RPMI 1640 medium containing 5 mM glucose, 10% FCS, penicillin, and streptomycin at 37 °C in a humidified atmosphere of 5% CO₂[9]. The islets were cultured for 7-10 days on the ECM plates to adhere and spread before they were infected with HyPer-Peroxi lentivirus or treated with palmitic acid.

RINm5F insulin-producing cells were seeded at 2.5 RR 10⁴ cells/well in 100 µl culture medium on 96-well plates and allowed to attach for 24 h before they were incubated at 37°C with NEFAs for 24 h. Cell viability was then determined by a microplate-based MTT assay (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide, Serva, Heidelberg, Germany) [27].

RINm5F insulin-producing cells were seeded at 1 aa 10⁶ cells on 6 cm tissue culture dishes and treated after 24 h with NEFAs. The cells were trypsinized after 24 h exposure to different NEFAs and stained for active caspases with the FITC-conjugated monoclonal active caspase-3 antibody apoptosis kit I (BD Biosciences Pharmingen, San Diego, CA). For the detection of caspase-8, caspase-9 or caspase-12 staining kits from PromoKine (Heidelberg, Germany) were used according to the manufacturer's protocols. The cells were analysed by flow cytometry at the FL-1 channel (488/527 nm) for active caspase-3, -9, and -12 or at the FL-2 channel (488/575 nm) for active caspase-8 with the CyFlowML cytometer (Partec, Münster, Germany). The portion of caspase-negative cells within the untreated cell population was gated and, on the basis of this gate, the portion of cells in which the caspases were activated after NEFA treatment could be determined.

Small RINm5F cell pellets treated for 24 h with different fatty acids were fixed in 2% para-formaldehyde and 2.5% glutaraldehyde in 0.1 M cacodylate buffer, pH 7.3, postfixed in 1% OsO₄ and finally embedded in Epon (Agar Scientific, Cambridge, UK) for electron microscopic analysis to identify changes in the different cell organelles [28]. Thin sections were contrast-stained with saturated solutions of lead citrate and uranyl acetate.

To detect overall oxidative stress, 2.5 T 10⁴ cells were seeded on 96-well black plates and cultured for 24 h. The cells were then pre-incubated with 10 µM DCF-DA (2,7-dichlorofluorescein diacetate, Sigma-Aldrich) for 30 min at 37°C. Thereafter, medium containing DCF-DA was replaced with fresh medium with or without NEFAs. After 24 h incubation, the plates were analysed at 480/520 nm excitation/emission using the fluorescence reader Victor² 1420 Multilabel Counter (Perkin Elmer, Wiesbaden, Germany) [29]. Thereafter cell viability was determined by MTT assay and DCF fluorescence was normalized to the viability.

To construct the expression vector for the HyPer-Peroxi protein that allowed detection of peroxisome-generated H₂O₂, the peroxisome target signal 1 (PTS-1) [30,31] was fused to the 3'-end of the HyPer cDNA by PCR using composite primer (HyPer-PTS1-XbaI-fw (5'-TATCTAGACGCCACCATGGAGATGGCAA-3') and HyPer-PTS1-Bsp119I-rv (5'-GCTTCGAATTACAGCTTGGAAACCGCCTGTTTTAAAAC-3')) and the pHyper-dCyto plasmid as template. Thereafter, the HyPer-Peroxi cDNA was subcloned into the XbaI/Bsp119I site of the pLenti6/V5-MCS plasmid [32].

To express the HyPer-Peroxi protein in the insulin-producing RINm5F cell line, lentivirus was prepared [33]. 5 . 10⁶ 293FT cells were transfected with the packaging plasmid pPAX2 (37.5 µg), the envelope plasmid pcDNA3-MDG (7.5 µg), and the transfer plasmid pLenti6/V5-MCS-HyPer-Peroxi (25 µg) by calcium phosphate precipitation. The virus particles were harvested from the culture medium 48 h later and purified by ultracentrifugation (70,000 g, 2 h). The virus titers (3-5 3 10⁷ infectious particles) were quantified by Taqman qPCR assay as described [34].

The RINm5F-control, RINm5F-Cat, and RINm5F-MitoCat [24] cell lines were infected with HyPer-Peroxi lentivirus at a MOI of 10. The cells were selected for HyPer expression using blasticidin (1 µmol/L).

RINm5F cells (1 1 10⁵) that overexpressed HyPer-Peroxi were seeded onto black 24-well glass-bottom plates (Zellkontakt, Nörten-Hardenberg, Germany). Cells were cultured for 24 h and then exposed to 100 µM palmitic acid, 100 µM oleic acid or a mixture of palmitic and oleic acid (100 µM each) for another 24 h. Live cell imaging was performed using a CFP-YFP dual filter (excitation, 427 nm and 504 nm; emission, 520 nm) with a cell^R/Olympus IX 81 inverted microscope system (Olympus, Hamburg, Germany). CellR software (Olympus) was used for imaging and analysis.

To determine H₂O₂ production, changes in the fluorescence ratios of RINm5F, RINm5F-Cat and RINm5F-MitoCat cells that overexpressed HyPer-Peroxi were quantified spectrofluorometrically. Those cells were seeded at 2.5 - 10⁴ cells per well onto black 96-well plates and cultured for 24 h. The fluorescence ratio (excitation 427 nm and 475 nm, emission 520 nm) was measured immediately before and after 24 h treatment with NEFAs.

RINm5F cells were cultured for 24 h and afterwards exposed to 100 µM of different unsaturated NEFAs with different chain lengths and different numbers of double bonds for another 24 h. Cells were trypsinised and fixed in paraformaldehyde for 15 min at room temperature. Thereafter cells were stained with Oil Red O solution (Sigma-Aldrich, Munich, Germany) and washed twice with PBS. Lipid droplet formation was analysed by flow cytometry or fluorescence microscopy. For the flow cytometry analyses the fluorescence of 2 m 10⁴ cells was measured at the FL-2 channel (488/575 nm) with the CyFlowML cytometer (Partec, Münster, Germany). For the fluorescence microscopy analyses the area within the cells was quantified by the use of the Olympus xcellence software (Olympus, Hamburg, Germany) at 546 nm excitation and 580 nm emission. For each incubation condition five to seven coincidentally selected images, each containing 30 to 100 cells, were used to quantify the proportion of the lipid droplet area to the total cell area with the phase analysis module of the xcellence software.

Data are expressed as means ± SEM. Statistical analyses were performed using ANOVA plus Dunnett's or Bonferroni's test for multiple comparisons, unless otherwise stated (Graphpad, San Diego, CA).

Palmitic acid, the physiologically most abundant saturated NEFA, had a strong cytotoxic effect on RINm5F insulin-producing cells, whereas unsaturated fatty acids showed only a minor toxic effect (Fig. 1A). Moreover, unsaturated NEFAs were protective against the toxicity of long-chain saturated NEFAs. The toxicity of saturated NEFAs is known to be chain-length dependent [9], whereas the structural requirements for the protective effect of unsaturated NEFAs and the mechanism of protection are still unknown. Therefore we analyzed the protective effect of unsaturated NEFAs that differ in the chain length of the carbon scaffold (Fig. 1B). Palmitoleic (C16:1) and oleic (C18:1) acid exhibited a strong protective effect against the toxicity of palmitic acid (200 µM) with comparable EC₅₀ values, of 21 ± 2 µM for palmitoleic acid and 22 ± 2 µM for oleic acid. In contrast, the shorter monounsaturated NEFA myristoleic acid (C14:1) had a significantly lower protective effect. Even at a concentration of 200 µM myristoleic acid, viability increased only to 47 ± 3%. With further shortening of the chain length the protective effect was even more reduced. Hence, the viability reached only 18 ± 2% after addition of 200 µM dodecenoic acid (C12:1). In summary, the protective effect against palmitic acid induced toxicity decreased with shortening of the chain length, with oleic and palmitoleic acid providing the greatest protection (Fig. 1B).

Unsaturated NEFAs that differ in the number of double bonds, oleic acid (C18:1), linoleic acid (C18:2), and ,□-linolenic acid (C18:3), exhibited only a minimal toxic effect upon insulin-producing cells (Fig. 1A). Nevertheless, all unsaturated C18 NEFAs showed a comparable, significant protection against the toxicity of palmitic acid (C16:0) irrespective of the number of double bonds.

To gain insight into the mechanism of lipotoxicity and the protective effect of unsaturated NEFAs in insulin-producing cells pristanic acid, a branched chain saturated NEFA, which is exclusively metabolized by peroxisomal β-oxidation, was used. Pristanic acid, like palmitic acid, also showed a distinct toxic effect on insulin-producing cells (47 ± 2% viability at 200 µM pristanic acid). Oleic acid protected against pristanic acid induced toxicity in almost the same concentration-dependent manner as against palmitic acid induced toxicity. These results indicate that the toxicity of saturated fatty acids depends upon the metabolism in peroxisomes (Fig. 1C).

The comparison of the toxic effect of palmitic acid with that of the respective non-metabolisable methyl derivative revealed that its metabolism is essential for the toxic action. The viability of RINm5F cells after treatment with 200 µM palmitic acid decreased significantly to 12 ± 1%, whereas the viability after incubation with 200 µM methylpalmitic acid (palmitic acid methyl ester) was not significantly reduced in comparison to untreated cells (95 ± 7%).

A flow cytometry based caspase-3 assay was performed to determine the cell death mode of lipotoxicity. Caspase-3 was significantly activated in RINm5F insulin-producing cells after 24 h treatment with palmitic acid (100 µM) (30 ± 4% vs. 6 ± 1% in control cells, n=4; p < 0.01). By contrast, oleic acid (100 µM) (5 ± 1%; n=4) did not activate caspase-3. Moreover, oleic acid entirely prevented palmitic acid induced caspase-3 activation (6 ± 1%; n=4).

To obtain information upon the morphological changes induced by NEFAs, an ultrastructural analysis by electron microscopy was performed (Fig. 2). After a 24 h incubation with palmitic acid (100 µM) damage to the rough endoplasmic reticulum (ER) was clearly detectable, while at the same time the mitochondria remained intact (Fig. 2B). The cisternae of the ER were elongated and dilated containing dark deposits. When the cells were incubated with oleic acid (100 µM), lipid droplets accumulated in the cytoplasm, whereas the ER remained unaffected (Fig. 2C). Only minor accumulation of lipid droplets and damage to the ER were observed after treatment with a mixture of palmitic and oleic acid (Fig. 2D) when compared to control (Fig. 2A). There were no mitochondrial alterations detectable under any of the different experimental conditions.

To characterize the underlying mechanism of NEFA induced cell death the gene expression of the ER-stress marker genes X-box binding protein 1 (XBP1), the spliced form of the X-box binding protein 1 (XBP1s), CHOP (C/EBP (CCAAT/enhancer-binding protein)-homologous), and the ER chaperone Bip were analysed. Palmitic acid (100 µM) induced the expression of the ER stress marker genes CHOP and Bip and also splicing of XBP-1 was markedly induced after palmitic acid incubation with a maximum after 12 h, whereas expression of XBP-1 was not affected (Fig. 3). Oleic acid (100 µM) had no effect on the gene expression of any of the analyzed ER stress marker genes (Fig. 3). Moreover, it completely prevented the palmitic acid mediated ER stress gene induction (Fig. 3).

To gain a closer insight into the role of different cellular compartments on the NEFA induced apoptosis the activation of death-receptor specific caspase-8, mitochondria specific caspase-9 and ER specific caspase-12 were analysed. Gene expression of ER stress marker genes was elevated after 12 h incubation with palmitic acid (100 µM), but did not result in an immediate activation of caspase-12. However, the caspases were significantly induced after 18 h with a further increase at 24 h (Fig. 4). Remarkably, the induction of all analysed caspases was not different at any time point, indicating that all caspases could be activated by the same upstream effector. In contrast to palmitic acid, oleic acid (100 µM) did not induce any of these three caspases -8, -9 and -12, but it reversed the effects of palmitic acid (data not shown).

ROS generation observed with the DCF assay was increased approximately 7-fold when RINm5F control cells were treated with palmitic acid (100 µM) in comparison to untreated cells, while palmitic acid incubation of RINm5F cells overexpressing catalase in the peroxisome resulted in a merely 2-fold increase in DCF fluorescence (Fig. 5). When catalase was overexpressed in the mitochondria, the ROS production was 4-fold elevated. The pronounced decrease of DCF fluorescence by catalase overexpression indicates that H₂O₂ is the main reactive oxygen species formed during incubation with palmitic acid. By contrast ROS generation was only slightly increased after exposure to oleic acid (100 µM). Moreover, oleic acid was able to prevent palmitic acid induced ROS generation (Fig. 5).

Reduction of DCF by overexpression of catalase suggests that H₂O₂ is generated due to palmitic acid treatment. To obtain direct evidence for the H₂O₂ formation in a defined subcellular localization the H₂O₂ sensor protein HyPer was expressed in primary rat islet cells and RINm5F cells. The HyPer-Peroxi protein containing the signal peptide sequence for peroxisomal import allows ratiometric measurement of H₂O₂ by fluorescence microscopy in this organelle. In the false colored images, green staining corresponds to low, yellow to medium, and red to high H₂O₂ concentrations. After 24 h of exposure to palmitic acid (100 µM), RINm5F-HyPer-Peroxi cells showed a strong shift towards red fluorescence in comparison to untreated cells (Fig. 6A and Fig. 6B), whereas the cells exposed to oleic acid (100 µM) showed a fluorescence signal similar to that of untreated cells (Fig. 6C). Moreover, oleic acid prevented the strong fluorescence shift towards red observed after palmitic acid treatment (Fig. 6D).

The H₂O₂ generation in the peroxisomes following exposure to NEFAs was quantified spectrofluorometrically. The H₂O₂ concentration after oleic acid treatment was significantly lower than after palmitic acid treatment (Fig. 7). Moreover, as observed with the DCF assay, oleic acid reduced palmitic acid induced H₂O₂ generation (Fig. 7). Further evidence that the peroxisomes were the subcellular site of H₂O₂ formation arises from the fact that only catalase in the peroxisomes significantly reduced H₂O₂ formation, whereas overexpression of catalase in the mitochondria had no effect. The mitochondrial H₂O₂ generation observed with the HyPer protein located in the mitochondria showed no significant changes following exposure to palmitic acid, oleic acid or a mixture of both (data not shown).

Primary rat islet cells expressing HyPer-Peroxi were treated for 24 h with palmitic acid (500 µM), oleic acid (500 µM) or a mixture of palmitic and oleic acid (500 µM each) (Fig. 8). The hydrogen peroxide production was analysed by fluorescence microscopy and the fluorescence signal of individual cells was quantified with the CellR software (Olympus). After a 24 h incubation with palmitic acid the hydrogen peroxide production increased 10-fold in comparison to untreated cells while oleic acid treatment did not result in a significant increase in hydrogen peroxide production (Fig. 8). Moreover, oleic acid was able to prevent the palmitic acid mediated induction of hydrogen peroxide production (Fig. 8).

Incubation with 100 µM oleic acid lead to significant NEFA accumulation by lipid droplet formation in RINm5F as well as in primary rat islet cells (Fig. 9). In RINm5F cells PA increased the lipid droplet content slightly (by 9%) (Fig. 9A). In primary rat islet cells PA showed no significant effect on lipid droplet content (Fig. 9B). Interestingly the lipid droplet formation induced by OA in RINm5F and primary rat islet cells was significantly reduced in co-incubation experiments with PA (Fig. 9). These quantitative data agree with those of the ultrastructural analyses (Fig. 2).

Hyperlipidaemia is a major risk factor for the development of type 2 diabetes mellitus [35,36]. Elevated levels of non-esterified fatty acids (NEFAs) are under suspicion to mediate β-cell dysfunction and β-cell loss [5,37,38]. However, not only the total concentration of NEFAs is crucial for these deleterious effects but also the ratio of saturated to unsaturated NEFAs [39]. In the present study we analysed the protective effect of different unsaturated NEFAs against the toxicity of palmitic acid, the physiologically most abundant saturated fatty acid. According to earlier observations [8,19] unsaturated fatty acids are able to counteract the ability of saturated fatty acids to induce β-cell apoptosis, but the underlying mechanisms remain unknown. It has been proposed that increased β-oxidation and oxidative phosphorylation cause lipotoxicity by enhanced formation of reactive oxygen species (ROS) in the mitochondria [15,16,40] or alternatively in the peroxisomes [9,41]. For the detoxification of H₂O₂, the oxidoreductase catalase is expressed in the peroxisomes of most tissues [42,43], but not in those of pancreatic β-cells [23,44]. This low catalase enzyme activity could also be found in insulin-producing RINm5F cells. This makes these cells well suited as model cells for studies on H₂O₂-mediated lipotoxicity and the particular sensitivity of insulin-producing cells towards oxidative stress.

Initially we analysed the ROS formation by a DCF assay in RINm5F cells overexpressing the H₂O₂ detoxifying enzyme catalase either in the mitochondria or in the peroxisomes. While palmitic acid caused a strong increase of ROS formation the effect of oleic acid was negligible. Moreover, oleic acid was even able to suppress palmitic acid induced ROS formation. The reduced ROS formation in catalase overexpressing cells indicates that H₂O₂ is the deleterious reactive oxygen species. As shown before [9], this effect was much more pronounced when the catalase was overexpressed in the peroxisomes than in the mitochondria.

NEFAs are not only metabolized through mitochondrial β-oxidation but, in particular long-chain NEFAs, are substrates for peroxisomal β-oxidation [45,46,47]. In contrast to the mitochondrial β-oxidation, the acyl-CoA oxidases in the peroxisomes form H₂O₂ rather than reducing equivalents [48]. H₂O₂ can easily disproportionate by the metal-catalysed Fenton reaction to the highly reactive hydroxyl radical [49,50].

In order to obtain direct evidence for H₂O₂ generation in a defined subcellular localization we used the fluorescence sensor protein HyPer. An increase of the peroxisomal H₂O₂ concentration was detectable in insulin-producing RINm5F cells as well as in primary pancreatic islet cells after exposure to palmitic acid. This increased peroxisomal H₂O₂ production was suppressed in the presence of oleic acid. Additional evidence for the contribution of peroxisomal fatty acid metabolism to the lipotoxicity in insulin-producing cells was provided by the toxicity of pristanic acid, a branched chain saturated fatty acid, which is exclusively metabolized in the peroxisomes [45,47]. Oleic acid protected also against pristanic acid mediated toxicity in a concentration-dependent manner like against that of palmitic acid. The protective effect of unsaturated NEFAs was independent of the number of double bonds, indicating that the mechanism was not dependent on the radical scavenging capability of the double bond.

Another proposed mechanism for lipotoxicity has been the induction of ER stress [12,51], which ultimately resulted in apoptosis through CHOP activation. The damage to the ER observed in the present ultrastructural analyses provides support for such a contention. Palmitic acid induced expression of ER stress marker genes was fully suppressed by oleic acid. The observation that ER specific caspase-12 was activated to almost the same extent and over the same time course as the initiator caspases-8 and -9, questions the role of ER stress as a primary mediator of lipotoxicity. This equal activation including that of the mitochondria specific caspase-9 indicates that the mitochondria are not a primary site for the initiation of oxidative stress mediated lipotoxicity. This conclusion is in accordance with our recent observation [9] that peroxisomally generated H₂O₂ rather than mitochondrially generated H₂O₂ mediates lipotoxicity.

In contrast to palmitic acid, which was found to form TG depots in vacuoles of expanded endoplasmic reticulum confirming earlier observations [19,26,52], oleic acid induced cytosolic lipid droplet formation outside the ER, apparently due to its efficient esterification to triglycerides [19,26,52,53].

Through this sequestration and deposition of triglycerides in lipid droplets [19] unsaturated fatty acids such as oleic acid may not be available for peroxisomal β-oxidation thereby avoiding H₂O₂ formation. If this hypothesis for the protective mechanism would be correct, a co-incubation of palmitic and oleic acid should led to an increase in lipid droplet formation. However, our experiments revealed just the opposite result. Palmitic acid reduced oleic acid induced lipid droplet formation to the level of the control condition. This result is remarkable in view of the fact that in the co-incubation experiments the total NEFA concentration was doubled. It is therefore a question whether this sequestration by long-term storage of fatty acids in the form of triglycerides may be beneficial for pancreatic β-cells. Not least, because a number of diseases affecting other tissues such as non-alcoholic fatty liver disease [54] or cardiovascular disease have been associated to intracellular fat accumulation [55].

In conclusion, palmitic acid induced H₂O₂ formation by peroxisomal β-oxidation resulting in β-cell apoptosis, whereas oleic acid prevented this H₂O₂ formation due to antagonizing palmitic acid induced H₂O₂ formation in peroxisomal β-oxidation in rat insulin-producing RINm5F cells and primary rat islet cells. The results indicate that unsaturated fatty acids in the lipid component of the diet can help to prevent lipotoxicity-mediated deterioration of glucose tolerance ultimately resulting in the manifestation of type 2 diabetes.

DCF-DA (2 ,7 -dichlorofluorescein diacetate); EC₅₀ (half maximally effective concentration); ER (endoplasmic reticulum); H₂O₂ (hydrogen peroxide); MOI (multiplicity of infection); MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide); NEFAs (non-esterified fatty acids); OA (oleic acid); PA (palmitic acid); ROS (reactive oxygen species); TG (triglyceride).

We are grateful to Maren Böger and Martin Wirth for their skillful technical assistance. This work was supported by the European Union (Collaborative Project BetaBAT in the Framework Programme 7, grant agreement 277713).

The authors declare that there is no duality of interest associated with this study.