Tryptophan-Containing Dual Neuroprotective Peptides: Prolyl Endopeptidase Inhibition and Caenorhabditis elegans Protection from β-Amyloid Peptide Toxicity

Neuroprotective peptides represent an attractive pharmacological strategy for the prevention or treatment of age-related diseases, for which there are currently few effective therapies. Lactoferrin (LF)-derived peptides (PKHs) and a set of six rationally-designed tryptophan (W)-containing heptapeptides (PACEIs) were characterized as prolyl endopeptidase (PEP) inhibitors, and their effect on β-amyloid peptide (Aβ) toxicity in a Caenorhabditis elegans model of Alzheimer’s disease (AD) was evaluated. Two LF-derived sequences, PKH8 and PKH11, sharing a W at the C-terminal end, and the six PACEI heptapeptides (PACEI48L to PACEI53L) exhibited significant in vitro PEP inhibition. The inhibitory peptides PKH11 and PACEI50L also alleviated Aβ-induced paralysis in the in vivo C. elegans model of AD. Partial or total loss of the inhibitory effect on PEP was achieved by the substitution of W residues in PKH11 and PACEI50L and correlated with the loss of protection against Aβ toxicity, pointing out the relevance of W on the neuroprotective activity. Further experiments suggest that C. elegans protection might not be mediated by an antioxidant mechanism but rather by inhibition of Aβ oligomerization and thus, amyloid deposition. In conclusion, novel natural and rationally-designed W-containing peptides are suitable starting leads to design effective neuroprotective agents.


Introduction
Neurodegenerative diseases account for a significant proportion of morbidity and mortality in the developed countries. Moreover, these disorders are becoming more frequent due to the increased life expectancy [1]. Alzheimer's disease (AD), the most common type of senile dementia, is a neurodegenerative disorder with enormous social and economic impact [2]. Although the cause(s) of AD is still controversial, it is accepted that the accumulation of amyloid β peptide (Aβ) in brain plaques triggers downstream neurotoxic events, leading to neuronal dysfunction, cell death,

Lactoferrin Hydrolysates Inhibit Prolyl Endopeptidase Activity
LF hydrolysates (LFH) generated by pepsin, proteinase K, or trypsin and subjected to ultrafiltration through a 3 kDa cut-off membrane showed PEP inhibitory activity (Table 1). At the maximum concentration tested (2 mg/mL), proteinase K LFH provoked the highest inhibitory effect (41% of inhibition) while pepsin and trypsin hydrolysates inhibited PEP by approximately 20%. The proteinase K LFH (0.36-5.7 mg/mL) provoked significant concentration-dependent inhibition of PEP (Figure 1), reaching 65% inhibition at the highest concentration assayed (5.7 mg/mL). Non-hydrolysed LF (0.7 and 1.4 mg/mL) did not show any inhibitory effect on PEP activity. 81.5 ± 0.6 ** 1 Data are expressed as the percentage of PEP residual activity with respect to a control without peptide (100%) and are the mean ± standard deviation (SD) of three replicates. * Significant inhibition with respect to control (p < 0.05); ** Significant inhibition with respect to control (p < 0.01) (Student's t-test on absolute values of PEP activity).

Figure 1.
The concentration-dependent effect of lactoferrin hydrolysate (LFH) generated by proteinase K on prolyl endopeptidase (PEP) residual activity. Data are expressed as the mean ± SD of three replicates. * Significantly different from the control (p < 0.01; one-way ANOVA followed by a Dunnett multiple comparison test).

Lactoferrin-Derived Peptides from Proteinase K Hydrolysate Are Prolyl Endopeptidase Inhibitors
Proteinase K LFH is a complex hydrolysate from which 37 peptides were identified by high performance liquid chromatography tandem mass spectrometry (HPLC MS/MS) [38]. Eleven of the most abundant peptides identified in the hydrolysate were chemically synthesized and their in vitro PEP-inhibitory activity was tested ( Table 2). Six out of eleven peptides (PKH3, PKH4, PKH5, PKH6, PKH9, and PKH10) included one P residue in the sequence but as summarized in Table 2, all of them failed to inhibit PEP. Only two sequences, the hexapeptide PKH8 (NEGLTW) and the decapeptide PKH11 (SVDGKEDLIW), which share a W residue at the C-terminal end, exhibited significant inhibitory activities of 12% and 22%, respectively, in the conditions tested. Further experiments were carried out to determine the inhibitory potency of the most promising PEP-inhibitory peptide, PKH11, which showed an IC 50 value of 2.1 ± 0.2 mg/mL (Figure 2A).

Tryptophan-Containing Synthetic Heptapeptides Inhibit Prolyl Endopeptidase
The potential role of a W residue at the C-terminal end of LF-derived PEP-inhibitory peptides has also been pointed out in angiotensin-converting enzyme (ACE) inhibitory peptides, in which W at the C-terminus is associated with high inhibitory potency [20]. This prompted us to evaluate a set of six synthetic heptapeptides, which share two W residues at positions 3 and 7 (Table 3) and previously described as ACE inhibitors [39]. The inhibitory effects of 1 mg/mL heptapeptides on PEP activity are summarized in Table 3. The six heptapeptides showed significant enzyme inhibition with values ranging from 29% to 94%. Remarkably, PACEI50L exhibited a potent PEP-inhibitory activity with an IC 50 value as low as 0.33 ± 0.02 mg/mL ( Figure 2B). Ac-RKWHLFW-NH 2 71.6 ± 4.9 ** (a) PACEI52L Ac-RKWLFHW-NH 2 65.9 ± 3.6 ** (a) PACEI53L Ac-RKWLHFW-NH 2 71.5 ± 4.5 ** (a) 1 All the peptides were acetylated at the N-terminus (Ac) and amidated at the C-terminus (NH 2 ). 2 Data are expressed as the percentage of PEP residual activity with respect to a control without peptide (100%) and are the mean ± SD of three replicates. ** Significant inhibition with respect to control (p < 0.01) (Student's t-test on absolute values of PEP activity). Data with the same letter do not differ at the 95% level of confidence (Tukey's honestly significant difference (HSD) procedure).

Tryptophan Residues Are Important for Prolyl Endopeptidase Inhibition in PACEI50L and PKH11
To characterize the role of the C-terminal W in PEP-inhibitory activity, this amino acid residue was exchanged to alanine (A) in the selected sequences PACEI50L and PKH11 and thus the sequence variants RKWHFLA (PACEI50L-v1) and SVDGKEDLIA (PKH11-v1) were generated. A second PACEI50L derivative (PACEI50L-v2; RKAHFLA) was obtained by replacing both W residues at position 3 and 7 with A. We next analyzed the impact of the amino acid exchanges on PEP inhibition ( Figure 2). The exchange of W to A moderately reduced PEP inhibition by PKH11-v1, in which the inhibitory potency (IC 50 = 3.0 ± 0.3 mg/mL) was significantly lower than the one showed by the natural peptide PKH11 (IC 50 = 2.1 ± 0.2 mg/mL), as can be seen in Figure 2A. By contrast, a severe reduction of the inhibitory activity of both PACEI50L variants was observed ( Figure 2B). At the highest concentration tested (3.3 mg/mL), PACEI50L-v1 inhibited PEP by 20% while PACEI50L-v2 did not exhibit a clear inhibitory effect ( Figure 2B).

Prolyl Endopeptidase-Inhibitory Peptides Alleviate Amyloid β Peptide-Induced Paralysis in the Transgenic C. elegans
To study the functional effect of PEP-inhibitory peptides in an in vivo AD model, we used the transgenic C. elegans strain CL4176 engineered to express human Aβ 1-42 . The active peptides PACEI50L and PKH11 were assayed at doses of 0.5, 0.1, and 0.02 µg/mL. Statistical analysis showed significant differences among paralysis curves of control and treated worms at doses of 0.5 and 0.1 µg/mL, thus, indicating a delay in paralysis ( Table 4). The addition of PACEI50L and PKH11 at 0.5 and 0.1 µg/mL to the C. elegans medium provoked a reduction in the percentage of paralyzed worms at the end of the assay (49 h) and, in some cases, also a delay in the onset paralysis ( Table 4). The dose of 0.02 µg/mL did not provide any effect (data not shown). Moreover, the most effective dose was 0.1 µg/mL for both PACEI50L and PKH11 peptides (Trial 2, PACEI50L 0.1 versus PACEI50L 0.5, p = 0.004; Trial 2, PKH11 0.1 versus pKH11 0.5, p = 0.006; paired log Rank survival test). The average paralysis curves of 0.1 µg/mL PACEI50L and PKH11 treatments are shown in Figure 3, and the statistical analyses of the curves are summarized in Tables 5 and 6. In the case of peptide PACEI50L, a decrease of 25.2% of paralyzed worms compared with control fed nematodes at 49 h was observed ( Figure 3A and Table 5) whereas a delay in the onset paralysis and a reduction of 20% of final paralyzed worms were noticed for PKH11 treatment ( Figure 3B and Table 6).
We also evaluated the activity of both peptide variants derived from PACEI50L at the effective dose of 0.1 µg/mL. Both modified peptides significantly changed their activity when compared with the original PACEI50L peptide, and no protective effect upon paralysis was detected (Tables 4  and 5, and Figure 3A). In the case of PACEI50L-v2, the final percentage of non-paralyzed worms was even 5.6% lower than the control conditions. Finally, the analysis of the modified peptide PKH11-v1 (0.1 µg/mL) resulted in a partial reduction of the efficacy upon paralysis (Tables 4 and 6, and Figure 3B), in correlation with the in vitro data of PEP inhibition.

Peptides Do Not Protect C. elegans from Oxidative Stress
The protective effect upon oxidative stress in the wild-type strain N2 of C. elegans was assayed with PACEI50L, PKH11, and the corresponding peptide variants at a concentration of 0.1 µg/mL. Figure 4 shows that none of them was able to increase the worm survival after induction of oxidative stress with 2 mM H 2 O 2 , in contrast to the significant protection caused by vitamin C. These results indicate that peptides tested at 0.1 µg/mL, which is the most effective dose reducing C. elegans paralysis, did not show any antioxidant effect. Nematodes were cultured in control conditions (NGM), with vitamin C (10 µg/mL) and with peptides (0.1 µg/mL). * Significantly different from the control (p < 0.05; one-way ANOVA followed by Tukey's HSD); NS: not significant (p > 0.05).

Discussion
Currently, peptides have a wide range of applications in medicine and biotechnology. Moreover, multifunctionality is a common trait of many peptides, which might function as polypharmacological sequences. The present study characterizes new dual neuroprotective peptides that show in vitro PEP-inhibitory properties and C. elegans protection against Aβ 1-42 -associated toxicity in vivo.
The neuroprotective peptides include natural sequences derived from the milk protein LF and sequence-related rationally-designed synthetic heptapeptides.
LFHs obtained with three different proteases showed a moderate ability to inhibit PEP that was not shown by non-hydrolysed LF, suggesting that LF-derived peptides possess the inhibitory activity. The hydrolysates generated with three different proteases showed different PEP-inhibitory activities pointing out the importance of peptide sequences for enzyme inhibiting activity. In our study, the most potent LFH was generated using proteinase K, which exhibits specificity for peptide bonds adjacent to the carboxylic group of aliphatic and aromatic amino acids [40]. Moreover, the inhibitory activity of the hydrolysate was comparable to the previously reported IC 50 values of a sodium caseinate hydrolysate (0.77 mg/mL) [29] and different fish protein hydrolysates (1.10-4.21 mg/mL) [23]. LFHs obtained with proteinase K were previously described as in vitro inhibitors of the endothelin-converting enzyme (ECE) and ex vivo as inhibitors of ECE-dependent vasoconstriction [38]. Our results describe a new biological activity related to neuroprotection for LF-derived peptides.
PEP inhibitors developed as potential therapeutic drugs are substrate-like compounds containing one P or one P-analogue residue [10]. Most of the PEP-inhibitory peptides derived from natural sources described to date contain at least one and up to six P residues in their sequences, and from 3 to 18 amino acid residues in length [22,27,29]. In spite of the fact that six of the eleven LF-derived peptides evaluated here contain one P residue in their sequences, none of them displayed PEP-inhibitory activity, as shown for some P-containing peptides derived from collagen or corn γ-zein [27]. By contrast, the only two LF-derived peptides that provoked a modest in vitro enzyme inhibition (PKH8, NEGLTW; PKH11, SVDGKEDLIW) do not contain any P residue in their sequence, suggesting that both LF-derived peptides do not act as substrate-type inhibitors. Non-P-containing peptides derived from α-casein were also described as PEP inhibitors [30].
Since the two LF-derived peptides with PEP-inhibitory activity described here shared a C-terminal W, we hypothesized its potential key contribution to enzyme inhibition. Our hypothesis was confirmed by the ability of the six sequence-related synthetic heptapeptides (PACEI48L to PACEI53L) to inhibit PEP. PACEI heptapeptides are the second generation of angiotensin converting enzyme inhibitory peptides based on two hexapeptides leads, PACEI32L (RKWHFW) and PACEI34L (RKWLFW) [41]. Heptapeptides were designed by combinations of F, H, and L residues in positions 4-6, all of them share R, K, and W residues at the N-terminus and W residue at the C-terminus of a given heptapeptide (Table 3) [39]. Among these peptides, the highest inhibitory activity was recorded for PACEI50L with H, F, and L residues at positions 4, 5, and 6, respectively. The activity of this heptapeptide is sequence-specific, as demonstrated by the swapping of H and F residues (sequences PACEI50L and PACEI48L) which drastically reduced the PEP-inhibitory activity. These results confirm and extend previously reported data on how minor amino acid exchanges affect biological properties [41,42].
Further confirmation of the key role of W residue at the C-terminus was provided by the evaluation of PACEI50L and PKH11 variants that contain W to A substitutions. Our results showed that the effect of W residue on PEP-inhibitory activity is dependent on the peptide sequence since a severe reduction of the biological activity was observed for the two variants of PACEI50L while only a 40% reduction of the inhibitory potency was recorded for the PKH11 variant. Our results also pointed to the relevance of both W residues in the PACEI50L sequence since the total loss of PEP inhibition was achieved only with the double substitution of W residues at positions 3 and 7. In agreement with these results, the most potent PEP-inhibitory peptides named 13L and 9L identified from a cocoa hydrolysate contained W in their sequence (13L, DNYDNSAGKWWVT; 9L, NYDNSAGKW) and the lower IC 50 value (0.19 mg/mL) corresponded to the sequence 13L [31]. Additional studies are required to explain the role of W residues in PEP-inhibitory peptides.
C. elegans is a suitable in vivo model for research on the molecular biology and genetics of different diseases as well as for drug and bioactive compound screenings [43]. C. elegans models present alternative approaches to understanding neurodegenerative diseases for which there are currently few effective therapies [6]. Here the transgenic C. elegans model of AD, which develops a paralysis phenotype, was used to study the effects of peptides. PACEI50L and PKH11 treatments ameliorated Aβ-induced paralysis suggesting a potential in vivo protection from Aβ 1-42 toxicity, as described for cocoa and maize peptides [31,32] and also for a sequence purified from scorpion venom [33]. Remarkably, and as observed for PEP-inhibitory activity, peptide variants of PACEI50L completely lost paralysis suppression in C. elegans demonstrating the positive role of W residues in the context of the heptapeptide sequence evaluated.
The effect of different compounds on Aβ 1-42 -mediated paralysis in transgenic C. elegans has also been associated with antioxidant effects [31][32][33]44]. In contrast to other neuroprotective peptides, the reduction of Aβ-toxicity by PACEI50L and PKH11 does not seem to operate through an antioxidant mechanism since both peptides did not produce significant C. elegans protection upon oxidative stress under the conditions tested. Conversely, the antioxidant properties of maize and scorpion venom peptides might underline, at least in part, their protective effect observed in C. elegans, since they inhibited the production of reactive oxygen species [32,33]. Additionally, in a similar experiment to the one described here, the peptide 13L from cocoa provided protection against oxidative stress, suggesting that the antioxidant activity might contribute to the protection against Aβ 1-42 -induced damage [31]. Therefore, our results suggest alternative mechanisms for PACEI50L and PKH11.
Aβ 1-42 contains hydrophobic motifs in the central (residues 17-21) and carboxy-terminal (residues 29-42) regions of the peptide [45]. Amyloidogenic peptide aggregation seems to be primarily driven by these hydrophobic domains [46]. Here, docking analysis showed that the potential hydrogen bond interactions between the active PACEI50L and Aβ 1-42 are focused on amino acids within both hydrophobic motifs (L17, M35, and G38) or close to them (K28). Particularly M35 and its hydrophobic surroundings seem to be important for the oxidative, neurotoxic, and aggregation properties of the peptide [47,48] whereas G residues can stabilize amyloidogenic structures by means of hydrogen bonds [49]. The conformation adopted by Aβ 1-42 seems to be an important factor in amyloid formation since the Aβ 1-42 peptide with α-helical or random coil structure aggregates slowly while Aβ with β-sheet conformation aggregates rapidly [45]. L17 included in the central hydrophobic stretch 17-21 contributes to adopt a β-sheet conformation that facilitates monomeric interaction, β-sheet oligomers, and amyloid fibrils [50]. Moreover, amino acids 17-20 served as a template for designing synthetic peptides able to inhibit fibrillogenesis in a rat brain model of amyloidosis [35]. Our docking results suggest that molecular interactions through hydrogen bonding between PACEI50L and residues L17, M35, and G38 from Aβ 1-42 might impede monomeric interactions and thus, Aβ oligomerization. Docking analysis revealed the loss of hydrogen bonds between non-active PACEI50L variants and these three residues in Aβ 1-42 . Whether PACEI50L might interfere with the folding of Aβ 1-42 to form aggregates as suggested by in silico analysis requires further in vitro and in vivo research.
Our study underscores the important role of W for the bioactivity of neuroprotective peptides. W-containing peptides as those described here display several biological activities including antihypertensive, antioxidant, antidiabetic, and satiating properties [51]. Besides W is the sole precursor of serotonin, which has been reported to have an effect on the psychological/cognitive function in humans. Remarkably, increased dietary W intake reduced intra-neuronal Aβ accumulation in a mouse model of AD, suggesting a neuroprotective role of W through the increase of serotonin levels [52]. Increased W levels also extend longevity in C. elegans and protect from alpha-synuclein and polyglutamine toxicity [53,54], suggesting a protective role against proteotoxicity in aging and age-related diseases. Additionally, in the C. elegans model, supplementation with the W-containing 13L peptide upregulated the W metabolism, including genes involved in the synthesis of serotonin and other neurotransmitters [31]. It is worthwhile to note that milk proteins and, among them, LF, are particularly rich in W in comparison with other dietary proteins opening the way to the future inclusion of LF in dietary recommendations for preventing or postponing AD.
In conclusion, we have identified novel natural and rationally-designed W-containing peptides showing in vitro PEP inhibition and in vivo protection from Aβ 1-42 toxicity, although further research needs to be conducted in murine models to analyze the effectiveness of the peptides. These results add a new application to the antihypertensive proteinase K LFH and the synthetic heptapeptides confirming their multifunctionality. The effect of PACEI50L on delayed paralysis in C. elegans might be mediated, at least in part by the inhibition of Aβ 1-42 oligomerization and thus, amyloid deposition, while the peptide antioxidant activity does not seem to be involved in the protective effect. Our results suggest that W-containing peptides are suitable starting leads to design effective neuroprotective agents. Further improvements to increase in vivo peptide stability based on the use of D-amino acid sequences are in progress. Future efforts are currently directed to clarify the mechanisms underlying the in vivo protective effects of W-containing peptides.

Materials
Bovine LF was provided by FrieslandCampina Domo (Zwolle, The Netherlands). Porcine pepsin and trypsin (type II-S), the bicinchoninic acid (BCA) kit, and the PEP substrate Z-Gly-Pro-p-nitroanilide were purchased from Sigma (Madrid, Spain). PEP was supplied by the Seikagaku Corporation (Tokyo, Japan). Recombinant proteinase K was purchased from Roche (Mannheim, Germany).
Synthetic peptides were purchased at >95% purity from GenScript Corporation (Piscataway, NJ, USA), wherein they were synthesized by solid phase methods using N-(9-fluorenyl) methoxycarbonyl chemistry. PACEI peptides were acetylated at the N-terminus and amidated at the C-terminus. The synthetic peptide concentration was based on the dry weight and purity provided by the manufacturer.
The protein content of LFHs was estimated by the BCA method using bovine serum albumin as the standard [55].

Prolyl Endopeptidase Assay
The PEP activity was determined as previously described [31] with minor modifications. Enzyme assays were performed in 96-well microplates using Z-Gly-Pro-p-nitroanilide as a substrate and measuring the increase in absorbance at 410 nm due to the release of p-nitroaniline. Ten µL of tenfold peptide or hydrolysate solution in 100 mM sodium phosphate buffer, pH 7 (final concentrations from 0.063 to 3.6 mg/mL for peptides and from 0.35 to 5.7 mg/L for LFHs), 30 µL of a 0.1 U/mL PEP solution in 100 mM sodium phosphate buffer (pH 7) and 35 µL of the same sodium phosphate buffer were pre-incubated at 30 • C for 5 min, and the mixture incubated with 25 µL of 2 mM Z-Gly-Pro-p-nitroanilide in 40% 1,4-dioxane for 30 min at the same temperature. The reaction was finished by the addition of 100 µL 10% Triton X-100 in 1 M sodium acetate buffer (pH 4), and the absorbance measured at 410 nm in a microplate reader. Data are expressed as the percentage of PEP residual activity with respect to a control without peptide (100%).
The IC 50 value of a peptide was defined as the concentration required to inhibit 50% of the PEP activity and the value for each experiment was estimated by non-linear regression of the experimental data to a four-parameter logistic curve using the software package SigmaPlot v 13.0 (SPSS Inc., Chicago, IL, USA).

C. elegans Strains and Maintenance
Wild-type Bristol strain N2 and the transgenic strain CL4176 (smg-1 ts [pAF29(myo-3/Aβ 1-42 /let UTR) + pRF4(rol-6(su10069))]) were obtained from the Caenorhabditis Genetics Center (College of Biological Sciences, University of Minnesota, Saint Paul, MN, USA) and were used for oxidative stress experiments and paralysis assays, respectively. C. elegans strains were routinely propagated on Nematode Growth Medium (NGM) plates with Escherichia coli strain OP50 as a food source at 20 • C (N2) or at 16 • C (CL4176). Worms were synchronized by isolating eggs from gravid adults at 20 • C (N2) or at 16 • C (CL4176), and eggs were hatched overnight in NGM plates.
In the oxidative and paralysis experiments, worms were fed with the different compounds from egg to adult stages, and transferred to new plates every two days.

C. elegans Paralysis Assay
Paralysis is induced in strain CL4176 by the expression of a muscle-specific Aβ 1-42 which depends on up-shifting temperature from 16 to 25 • C [7,8]. For paralysis experiments, strain CL4176 maintained at 16 • C was egg-synchronized in plates containing NGM (control medium) and NGM with peptides (0.02-5 µg/mL). Transgene expression was induced by up-shifting the temperature from 16 to 25 • C, starting 24 h after egg laying and maintained for 24 h. The worms were incubated at 20 • C until all the worms in the experiment became paralyzed. Paralysis was scored 24 h after induction. Paralysis in induced worms was compared with non-induced worms (maintained at 16 • C until the end of the paralysis assay). Experiments were carried out at least in duplicate.

C. elegans Oxidative Stress Assays
For oxidative stress experiments, strain N2 was egg-synchronized in NGM plates (control medium) and NGM supplemented with 0.1 µg/mL peptides or 0.1 mg/mL vitamin C (positive control). The viability of C. elegans was assessed after 2 mM hydrogen peroxide (H 2 O 2 )-induced oxidative stress for 5 h [56]. Experiments were carried out in triplicate.

Statistics
Statistical analyses were carried out using the GraphPad Prism 4 software package (GraphPad Software, La Jolla, CA, USA). PEP residual activity data are mean ± standard deviation (SD) and were subjected to either Student's t-test or one-way ANOVA followed by Dunnett post-test or Tukey's honestly significant difference procedure (HSD). The comparison of the C. elegans paralysis curves was performed using the log-rank survival test. The survival data of the C. elegans was assessed after oxidative stress and differences between nematodes cultured in the control and treatment conditions were evaluated by means of a one-way ANOVA followed by Tukey's HSD. p < 0.05 was considered significant.
Author Contributions: P.Man. D.R., S.Ge., J.F.M. and P.Mar. conceived and designed the experiments; P.Man. and S.Ge. performed PEP assays; R.M. and P.Mar. performed C. elegans assays; S.Ga. and J.F.M. performed and analyzed docking results; D.R. and J.F.M. run statistical analyses; P.Man., D.R., S.Ge., J.F.M. and P.Mar. analyzed the data; P.Man. and P.Mar. wrote the paper; all authors revised and approved the manuscript.
S.Ga was recipient of a predoctoral scholarship (FPU13/04584) within the FPU program from "Ministerio de Educación, Cultura y Deporte" (MECD, Spain). We acknowledge support of the publication fee by the CSIC Open Access Publication Support Initiative through its Unit of Information Resources for Research (URICI).

Conflicts of Interest:
The authors declare no conflict of interest.