To What Extent Do Fluorophores Bias the Biological Activity of Peptides? A Practical Approach Using Membrane-Active Peptides as Models

The characterization of biologically active peptides relies heavily on the study of their efficacy, toxicity, mechanism of action, cellular uptake, or intracellular location, using both in vitro and in vivo studies. These studies frequently depend on the use of fluorescence-based techniques. Since most peptides are not intrinsically fluorescent, they are conjugated to a fluorophore. The conjugation may interfere with peptide properties, thus biasing the results. The selection of the most suitable fluorophore is highly relevant. Here, a comprehensive study with blood–brain barrier (BBB) peptide shuttles (PepH3 and PepNeg) and antimicrobial peptides (AMPs) (vCPP2319 and Ctn[15-34]), tested as anticancer peptides (ACPs), having different fluorophores, namely 5(6)-carboxyfluorescein (CF), rhodamine B (RhB), quasar 570 (Q570), or tide fluor 3 (TF3) attached is presented. The goal is the evaluation of the impact of the selected fluorophores on peptide performance, applying routinely used techniques to assess cytotoxicity/toxicity, secondary structure, BBB translocation, and cellular internalization. Our results show that some fluorophores significantly modulate peptide activity when compared with unlabeled peptides, being more noticeable in hydrophobic and charged fluorophores. This study highlights the need for a careful experimental design for fluorescently labeled molecules, such as peptides.


INTRODUCTION
The development of molecules for biological or biomedical applications relies on their accurate and precise biophysical/biological characterization (La Gatta et al., 2016;Van Norman, 2016). The necessary data concerning the efficacy, toxicity, mechanism of action, cellular uptake, or intracellular location of such molecules can be gathered using in vitro or in vivo approaches (D'Addio et al., 2016). The collection of this information requires the use of highly sensitive techniques, which usually depend on the use of fluorescent probes (Gonzalez-Vera, 2012;Xu et al., 2018;Gao and Wu, 2019). Since most molecules are not intrinsically fluorescent in the visible spectrum range, conjugation to a fluorophore is needed. This way, fluorescencebased techniques, such as confocal laser scanning microscopy (CLSM), flow cytometry, and fluorimetry are possible options (Gautam et al., 2015;Radicioni et al., 2015). Despite their extensive use and value, most fluorophores are bulky, rigid, and hydrophobic molecules. Hence, their conjugation to other molecules may alter their physicochemical/biological properties, mainly when dealing with low molecular weight molecules, which may ultimately bias the results obtained by a given technique (Toseland, 2013;Sánchez-Rico et al., 2017;Hedegaard et al., 2018).
The selection of the fluorophore is usually based on chemical intuition, cost, and photophysical properties. The major parameters considered in the selection of a fluorophore are the excitation/emission wavelength, brightness, photobleaching, photostability, chemical reactions required, or conjugation yields.
The assumption that unlabeled-and labeled-molecules have the same properties often relies on wish much more than evidence . The advent of studies reporting differences between fluorophores raised awareness to the importance of their selection (Fischer et al., 2002;Toseland, 2013;Knutson et al., 2016;Zhao et al., 2016;Birch et al., 2017;Hedegaard et al., 2018). In addition, some studies also report the impact of the molecule on the properties of the fluorophore (Toseland, 2013;Szabó et al., 2018). The choice of the most suitable fluorophore to label molecules of interest is thus of utmost importance, not compatible with unworthy presumptions.
In the last decades, the interest in peptide-based systems has increased owing to improvements in knowledge and manipulation of peptide physicochemical properties (Fosgerau and Hoffmann, 2015;Boone et al., 2018;Ghasemy et al., 2018). As a result, peptides are now part of different therapeutic/diagnostic protocols. The main applications in therapeutic medicine are in cancer [anticancer peptides (ACPs)] and infectious diseases [antimicrobial peptides (AMPs)] (Gaspar et al., 2013;Shoombuatong et al., 2018). Radiolabeled-peptides for targeted nuclear molecular imaging and/or systemic radiotherapy play unique roles in nuclear medicine and oncology (Correia et al., 2011;Oliveira and Correia, 2019). Additionally, peptides have also been employed in drug-delivery systems for their receptor specificity and cargo translocation capacity across either epithelial or endothelial cellular barriers (Zou et al., 2013;Oller-Salvia et al., 2016;Gallo et al., 2019). Like many other molecules, peptides are not intrinsically fluorescent in the visible range of the electromagnetic spectrum. Thus, they are a good example of a molecule of therapeutic and biological interest that must be conjugated to a fluorophore. Recently, unnatural fluorescent amino acids have been tested as an alternative to the use of fluorophores (Mendive-Tapia et al., 2016;Cheng et al., 2020;Subiros-Funosas et al., 2020), presenting several advantages over fluorophores. Nevertheless, the use of fluorophores is still broadly used to investigate peptide biological activity.
To broaden the comprehensiveness of the study, we studied the peptide secondary structure, cellular uptake, intracellular location, and toxicity to red blood cells (RBCs). Also, in vitro BBB translocation assay for BBB peptide shuttles and cytotoxicity toward cancer cells of the last two peptides was carried out. Ultimately, we have concluded on the biasing each fluorophore cause on structural and functional data accounting to the methodologies applied in the experimental assays.
Analytical reversed-phase HPLC was performed on a Luna C18 column (4.6 mm × 50 mm, 3 µm; Phenomenex, United States). Linear gradients of solvent B (0.036% TFA in MeCN) into solvent A (0.045% TFA in H 2 O) were used at a flow rate of 1 mL/min and with UV detection at 220 nm. Preparative HPLC runs were performed on a Luna C18 column (21.2 mm × 250 mm, 10 µm; Phenomenex) using linear gradients of solvent B (0.1% TFA in MeCN) into solvent A (0.1% TFA in H 2 O) at a flow rate of 25 mL/min and with UV detection at 220 nm. Fractions of adequate HPLC homogeneity and with the expected mass were combined and lyophilized. LC-MS was performed in a LC-MS 2010EV instrument (Shimadzu, Kyoto, Japan) fitted with an XBridge C18 column (4.6 mm × 150 mm, 3.5 µm; Waters, Spain), eluting with linear gradients of HCOOH/MeCN (0.08% v/v) into HCOOH/H 2 O (0.1% v/v) over 15 min at 1 mL/min. Peptide stock solutions (1 mM) in filtered H 2 O were stored at −20 • C.

Measurement of Spectral Properties
Spectroscopic data were recorded on an FS900 fluorometer (Edinburgh Instruments, United Kingdom). The different peptides were dissolved at concentrations around 10-50 µM in 1× PBS and spectra recorded at r.t. (Supplementary Figures S1-S4).

Circular Dichroism
Circular dichroism (CD) spectra of the different peptides were acquired in a J-815 spectropolarimeter (Jasco, Japan) at 25 • C in the 190-260 nm wavelength range, with a bandwidth of 1 nm and a scan speed of 50 nm/min, using a 0.1 cm quartz cell. 50 µM peptide solutions were prepared in 10 mM sodium phosphate (75.4 mM Na 2 HPO 4 , 24.6 mM NaH 2 PO 4 , pH 7.4). The final spectra for each peptide were the average of three consecutive scans per sample after subtraction of buffer baselines (Supplementary Figure S5). Results were expressed as mean residue ellipticity ([θ] MRW ) (deg × cm 2 × dmol −1 ), as follows: where, θ obs is the observed ellipticity in degrees, MRW is the mean residue weight, d is the cell path length and c is the peptide molar concentration.

Hemolytic Activity
Fresh human blood was collected in EDTA tubes and centrifuged at 1,000 × g for 10 min at 4 • C. The supernatant was discharged, and the pellet containing RBCs was washed three times with 1× PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , and 1.8 mM KH 2 PO 4 , pH 7.4) and resuspended in 1× PBS to obtain a 2.0% (v/v) suspension. Then, RBCs were added to centrifuge tubes containing two-fold serially diluted peptides to a final concentration ranging from 0.01 to 100 µM. The suspension was incubated for 24 h at 37 • C with gentle stirring. After that, samples were centrifuged for 2 min at 1,000 × g. Supernatants were transferred to 96-well plates, and the hemoglobin released measured by absorbance at 570 nm in an Infinite F200 TECAN plate reader. 1× PBS with no peptides and Triton X-100 at 1 and 4% (v/v) were used as negative and positive controls, respectively. Hemolytic activity (%) was determined using the following equation: Hemolytic activity (%) = Abs PT − Abs NC Abs PC − Abs NC (2) where, Abs PT is the absorbance of treated samples, Abs NC is the absorbance from negative control, and Abs PC absorbance from positive control. HC 50 values were determined using the GraphPad Prism 7.0 software using a log(inhibitor) vs. normalized response. Experiments were performed on different days using independent blood donors.

In vitro Translocation Studies
HBEC-5i cells were allowed to grow until confluence in a gelatin-coated T-flask. Then, cells were carefully harvested with trypsin-EDTA (Gibco/Thermo Fisher, United States) and seeded 8,000 cells/well to 0.1% gelatin solution coated tissue culture inserts [transparent polyester (PET) membrane with 1.0 µm pores] for 24-well plates (BD Falcon, United States). During 8 days, the medium was changed every 2 days. After 8 days, cells were washed two times with 1× PBS and once with DMEM:F12 medium without phenol red (Gibco/Thermo Fisher, United States). Then, peptides diluted in DMEM:F12 medium without phenol red to a final concentration of 10 µM were added to the apical side of the in vitro BBB model (Figure 1). Experiments were performed on different days using independently grown cell cultures.

Evaluation by Fluorescence Emission
The translocation of peptides labeled with different fluorophores was determined by fluorescence emission. After 24 h incubation, samples from the apical and basolateral side were collected and analyzed. Fluorescence was measured using the infinite F200 TECAN plate reader. The percentage (%) of translocation was calculated using the following equation: F i is the fluorescence intensity recovered, F cells is the fluorescence intensity recovered from cells without peptide incubation, Frontiers in Bioengineering and Biotechnology | www.frontiersin.org F peptide is the fluorescence intensity of total peptide initially added to the transwell apical side, and F Medium is the fluorescence intensity of the medium.

Evaluation by HPLC
AUC i is the AUC of peptide recovered and AUC peptide is the AUC of total peptide initially added to the transwell apical side.

In vitro Integrity Assay
After the 24 h incubation period with the peptides, an in vitro integrity assay was performed. Herein, cells were washed two times with 1× PBS and once with DMEM:F12 medium without phenol. Then, previously diluted fluorescein isothiocyanatedextran with an MW of 4 kDa (FD4) (Sigma-Aldrich, Spain) was added to the apical side and incubated for 2 h. FD4 was diluted in DMEM:F12 medium without phenol to an absorbance of 0.1. Samples were collected from the apical and basolateral side, and fluorescence intensity was measured at λ with an excitation of 493 nm and maximum emission at 560 nm using the infinite F200 TECAN plate reader. The percentage of FD4 recovered was determined using the following equation: F i is the fluorescence intensity recovered, F cells is the fluorescence intensity recovered from cells without FD4 incubation, F FD4 is the fluorescence intensity of total FD4 initially added to the transwell apical side, and F Medium is the fluorescence intensity of DMEM:F12 medium without phenol red. The integrity of the barrier is indirectly proportional to the percentage of FD4 recovered and was determined using the following equation:

Cell Viability Measurements
Peptides cytotoxicity against both HBEC-5i and MDA-MB-231 cells was determined using CellTiter-Blue R Cell Viability Assay, according to the manufacturer's instructions. The assay is based on the ability of viable cells to reduce resazurin into resorufin, a highly fluorescent metabolite. On the other hand, non-viable cells rapidly lose metabolic capacity and, consequently, do not generate resorufin. Therefore, using this methodology is possible to distinguish between metabolic and non-metabolic cells and indirectly determine the cytotoxicity of peptides. Briefly, HBEC-5i and MDA-MB-231 cells were carefully detached from T-flasks, as described previously, and seeded 15,000/100 and 10,000/100 µL, respectively, in 96-well plates (Corning, United States) and incubated for 24 h. After medium removal, cells were washed two times with 1× PBS and 100 µL of previously diluted peptides (range between 0.01 and 100 µM) in either DMEM or DMEM:F12 medium were added to MDA-MB-231 or HBEC-5i cells, respectively. After 24 h incubation, cells were washed two times with 1× PBS, and 20 µL of CellTiter-Blue R Reagent (diluted in 100 µL of medium) was added to each well and incubated for 3 h in a humidified atmosphere of 5% CO2 at 37 • C. The fluorescence intensity was measured at λ with an excitation of 560 nm and maximum emission at 590 nm using the infinite F200 TECAN plate reader. Medium and 1% Triton X-100-containing medium were used as positive controls (100% cell viability) and negative controls (0% cell viability), respectively. Cell viability (%) was determined using the following equation: Frontiers in Bioengineering and Biotechnology | www.frontiersin.org F p is the fluorescence intensity of peptide-treated cells, F NC is the fluorescence intensity for negative controls, and F PC is the fluorescence intensity for positive controls. IC 50 values were determined using the GraphPad Prism 7.0 software using a log(inhibitor) vs. normalized response. Experiments were performed on different days using independently grown cell cultures.

Confocal Microscopy
HBEC-5i and MDA-MB-231 cells were seeded 50,000/200 µL on an ibiTreat-coated 8-well µ-slide (Ibidi, Germany) for 24 h. Then, cells were washed carefully two times with 1× PBS and once with medium and incubated for 2 h with labeled peptides at a final concentration of 10 µM. Nucleus was stained with Hoescht 33342 (Thermo Fisher, United States). After cell washing, nucleus dye was added to cells at a final concentration of 5 µg/mL and for 10 min at 37 • C. Finally, cells were washed twice with 1× PBS and imaged.
The acquisition was made on a confocal point-scanning Zeiss LSM 880 microscope (Carl Zeiss, Germany) equipped with an alpha Plan-Apochromat X 63 oil immersion objective (1.40 numerical aperture). Diode 405-30 laser was used to excite Hoechst 33342 (Sigma-Aldrich, Spain). The 488 nm line from an argon laser was used to excite peptide labeled with CF and NeNe594 laser was used to excite peptides labeled with RhB, Q570, and TF3. In the normal confocal mode, X 0.6 zoom images were recorded at 2048 × 2048 resolution. ZEN software as used for image acquisition. Fiji software was used for image processing. At least 12 total images were acquired in three independent replicates.
To compare the fluorescence intensities of the different fluorophores within different cells, we calculated the corrected total cell fluorescence (CTCF) using the following equation:

Statistical Analysis
Quantitative data were processed using Excel 2013 (Microsoft, United States) and the GraphPad Prism 7.0 software package. Medians, means and standard deviations are shown in the figures and tables. Pairwise significances were calculated using oneway ANOVA followed by Tukey's multiple comparison test, and non-parametric Mann-Whitney, Kruskal-Wallis.

Toxicity to Human Erythrocytes
The characterization of new biomolecules, such as peptides, includes the evaluation of their toxicity. The information collected helps avoiding rejection in preclinical stages, for instance (Hughes et al., 2011;Lau and Dunn, 2018). A standard evaluation of safety is to determine toxicity toward human RBCs since hemolytic assays are easy to perform, robust, cheap, and highly informative (Colonna et al., 2017).
In the present work, we evaluated the tendency of both labeled-and unlabeled-peptides to induce hemolysis in freshly collected RBCs. The results reveal differential hemolytic activity concerning the fluorophores and peptides employed (Figure 2). The peptides showing the lowest hemolytic activity are PepH3 and PepNeg. Unlabeled-PepH3 is non-hemolytic up to 100 µM (Figure 2A and Supplementary Table S4). Upon derivatization, only RhB-PepH3 shows a significantly increased toxicity (HC 10 = 9.208 µM, and HC 50 = 128.3 µM) (Figure 2A and Supplementary Table S4), while PepNeg has non-hemolytic activity up to 100 µM (Figure 2B and Supplementary Since the most common mechanism of action of AMPs corresponds to membrane disruption, these findings are in agreement with the high affinity of these peptides toward lipid membranes (Hoskin and Ramamoorthy, 2008). Nevertheless, the (B2) Fluorescence intensity of FD4 measured after translocation assay with PepNeg. The values were obtained from triplicates of three independent experiments. # represents the quantification using an HPLC. Statistical significance analysis was evaluated with a one-way ANOVA followed by Tukey's multiple comparison test and no statistical significance difference was observed between samples. toxicity increases with their conjugation to highly hydrophobic fluorophores (RhB and Q570).
The hemolysis rate in RBCs exposed to labeled-peptides show that CF-peptides do not exert damaging effects in comparison to other derivatives. Overall, all other labeledpeptides exhibit low to moderate hemolysis (TF3-peptide) or pronounced hemolysis (RhB-peptide and Q570-peptide). The correlation of this finding to the structure of fluorophores shows that negatively charged fluorophores contribute the least to membrane damaging (CF-peptide). By contrast, we observe more severe hemolysis with the conjugation to fluorophores that increase net charge compared with unlabeled-peptides, i.e., RhB-peptide, Q570-peptide, and TF3-peptide. Moreover, the combination of increased positive net charge and high hydrophobicity observed with RhB-peptides and Q570-peptides expand the hemolytic properties of unlabeled-peptides. This observation supports the idea that cationic and hydrophobic derivatives can disrupt membranes in a more efficient way than peptides conjugated to neutral or anionic fluorophores, which is in line with previous works (Birch et al., 2017;Avci et al., 2018). Thus, the outcome of a hemolytic assessment is highly dependent on the fluorophore-conjugated.

Translocation and Toxicity Across an in vitro BBB Model
The BBB is a physiological barrier responsible for the maintenance of the brain homeostasis (Weiss et al., 2009;Serlin et al., 2015). Therefore, strategies to overcome the BBB are an unmet clinical need (Hersh et al., 2016;Wang et al., 2019). Among others, the conjugation of therapeutics to BBB peptide shuttles has been one of the most promising (Oller-Salvia et al., 2016;Sánchez-Navarro et al., 2017). In addition, a good in vitro BBB model is also fundamental for a proper characterization of the translocation properties of the conjugates. Although different models have been described, (Helms et al., 2016) our lab optimized the use of monoculture BBB models. Thus, in a quick and robust way, we can easily access the translocation capabilities  CF, 5(6)-carboxyfluorescein; HC 10 , HC 50 , HC 90 , concentration that induces hemolysis in 10, 50, and 90% of red blood cells, respectively; IC 10 , IC 50 , and IC 90 , concentration that causes cell death in 10, 50, and 90% of human cells, respectively; N.A., Not applicable; RhB, rhodamine B; TF3, tide fluor 3; Q570, quasar 570.
Frontiers in Bioengineering and Biotechnology | www.frontiersin.org of many BBB peptide shuttles. Among the different cell lines that can be used within the model, (Helms et al., 2016) in the present study, we have successfully used HBEC-5i cells (Figure 1).
To assess the effect of the fluorophore on the translocation properties of BBB peptide shuttles, we selected two different peptides. PepH3 is a seven amino acid peptide derived from the α3-helical domain of the Dengue virus capsid protein (DEN2C) . It is a cationic and hydrophobic peptide able to translocate endothelial membranes in vitro and in vivo (Côrte-Real et al., 2016;. PepNeg is a new anionic peptide designed based on the sequence of PepH3 (Neves-Coelho et al., 2017). Our in vitro data also demonstrates that PepNeg efficiently transports cargo through the in vitro BBB model without disrupting the barrier (Neves-Coelho et al., 2017).
Within this study, results show that both PepH3 and PepNeg translocate the HBEC-5i in vitro BBB model in an efficient way (Figures 3A1,B1). For the labeled-peptides, we performed the quantification of the translocation percentage by measuring the fluorescence intensity in the basolateral side. Concerning unlabeled-peptides, we performed the quantification by HPLC. In all cases, the translocation (%) was above 50%, which is in line with previous results Neves-Coelho et al., 2017). Thus, this data suggests that none of the fluorophores has an impact on the translocation capacity of both BBB peptide shuttles. Next, we assessed the endothelial barrier integrity of the in vitro BBB model by the fluorescence intensity of FD4 recovered (Figures 3A2,B2). None of the unlabeled-or labeled-peptides influence the barrier permeability (HBEC-5i integrity > 95%). This finding also suggests that none of the derivatives is toxic toward HBEC-5i cell line.
In addition, we also performed a cell viability assay to confirm the absence of toxicity. For all the unlabeled-and labeled-peptides, the IC 50 is always > 200 µM. Nevertheless, the labeling to more hydrophobic fluorophores, such as TF3, Q570, or RhB, seems to increase cell death. Even so, in both assays, all peptides show no significant toxicity toward HBEC-5i cells (Figures 4A,B and Supplementary Table S5).
The cytotoxicity results are shown in Figures 4C,D and Supplementary Table S5. Unlabeled-and labeled-vCPP2319 FIGURE 5 | Representative confocal microscopy images of peptide internalization. Confocal microscopy analysis of HBEC-5i and MDA-MB-231 cells was conducted after incubating the cells at 37 • C for 1 h with BBB peptide shuttles and ACP at a final concentration of 10 µM, respectively. Blue is Hoechst 33342 (nucleus) and green is the derivatives. Scale bar = 20 µm. The fluorescence intensity of different images was determined to compare the fluorescence intensity of different fluorophores. This analysis was performed by calculating the corrected total cell fluorescence (CTCF).

Cellular Imaging of Labeled Peptides
The use of microscopy techniques in the characterization of peptides is a common practice. The results obtained allowed us to evaluate the capacity of the peptide to be internalized, the specificity of peptides toward some cell lines, the internalization mechanism, and the possible mechanism of actions, for instance (Shen et al., 2007;Matsumoto et al., 2015;Rezgui et al., 2016;Kumar et al., 2018). Herein, we assessed the uptake of both BBB peptide shuttles on HBEC-5i cells and both AMPs on MDA-MB-231 cells.
The incubation of both BBB peptide shuttles with HBEC-5i shows that all derivatives are internalized ( Figure 5). Thus, it indicates that there are no interference in cellular uptake. The fluorophore showing the highest background was Q570. Considering both AMPs incubated with MDA-MB-231 cells, the results are similar. All derivatives show high internalization capacity (Figure 5). Similarly to both BBB peptide shuttles, Q570 was the fluorophore showing the highest background.
The optimization of both the sensitivity of the detector and the laser intensity are also parameters to consider for each fluorophore. The derivatives that require the use of a more sensitive detector and a higher laser intensity were the CFpeptides. In addition, the exposition time was short, owing to the low photostability of the fluorophore. The use of RhB-peptides also requires some fine adjustments. Although it has a higher signal intensity, which allows the use of less sensitive detectors, the detection of RhB peptides requires the use of medium laser intensities. The use of CF and RhB is widely applied in research, mostly owing to their price. The use of highly advanced microscopes allows the detection of CF-or RhB-compounds. The use of either Q570-peptides or TF3-peptides overcome the previous limitations reported. Both fluorophores possess high signal intensities even at low laser intensities or using a less sensitive detector. In the absence of a high sensitive microspore, the use of these fluorophores might be highly advantageous.
In addition, all the derivatives demonstrate stability at long exposition times.

CONCLUSION
The study of peptides and the characterization of their potential biomedical application is limited to the capacity to visualize/quantify these peptides in cells. To do so, researchers rely on the use of fluorescent-labeled peptides to perform high sensitive techniques. However, until recently, very little was known about the influence of the fluorophore on the outcome of a given technique. The choice of the fluorescent probe was empirical and mainly based on spectral properties. In the present work, we selected four commercially available and highly used fluorophores with different physicochemical properties. Then, we conjugated them to four different peptides comprising two of the most important peptide applications, namely, BBB peptide shuttles and AMPs, which were tested as ACPs in this study owing to the high fraction of AMPs with anticancer activity.
Our results show that, indeed, fluorophores have an impact on peptide activity/toxicity depending on the peptide ( Table 1). In general, the main characteristics of fluorophore groups are: • CF has medium hydrophobicity, no toxicity, does not interfere with peptides' activity, but has low fluorescence intensity signal; • RhB has high hydrophobicity, high toxicity (conjugated to PepH3, vCPP2319, and Ctn[15-34]), decreases IC 10 , IC 50 , and IC 90 in cancer cells, and medium fluorescence intensity signal; • Q570 has high hydrophobicity, high toxicity (conjugated to vCPP2319, and Ctn[15-34]), decreases IC 10 , IC 50 , and IC 90 in cancer cells, and high fluorescence intensity signal; • TF3 has high hydrophobicity, and medium toxicity (conjugated to vCPP2319, and Ctn[15-34]), decreases IC 10 , IC 50 , and IC 90 in cancer cells, and high fluorescence intensity signal.
Consequently, it is important to highlight that for low toxicity, the fluorescence intensity signal might be compromised. Overall, to select a fluorophore, it is necessary to consider the specific application/technique, since it can contribute to biased results.

DATA AVAILABILITY STATEMENT
All datasets presented in this study are included in the article/Supplementary Material.

AUTHOR CONTRIBUTIONS
VN, MAC, DA, and MC conceived and designed the experiments and wrote the manuscript with contribution from all other authors. MC, CP-P, JV, and RS performed the experiments. All authors contributed to the data analysis.