Abstract

1. Introduction

Immunoglobulins, particularly G1 isotype (IgG1), are the most abundant glycoproteins in circulation and represent one of the most important mechanisms of protection against pathogens (1). They can exert protection through direct interaction with the antigen (2–4) and by recruiting specific receptors, that further activate effector functions such as antibody dependent cell cytotoxicity (ADCC), antibody dependent cell phagocytosis (ADCP), and complement dependent cytotoxicity (CDC) (5–7). IgG1 are often commercialized as monoclonal antibodies (mAbs), potent biotherapeutics used in the treatment against several diseases. From a structural perspective, mAbs are composed of four chains, two light and two heavy (LC and HC, respectively), that are assembled to form a symmetric structure organized in two fragment antigen binding (Fab) domains and one fragment crystallizable (Fc) linked by a flexible hinge region. LCs and HCs are connected by disulphide bonds and contain both variable and constant regions of which the variable ones are involved in the antigen binding (8). The neutralizing effect of mAbs is due to the complementarity determining regions (CDRs), located in the Fab region, that constitute the heterogeneity and uniqueness of the antibody. Moreover, the interaction with immune receptors is determined by the Fc domain, that is very conserved among the different classes of antibodies. A key role in modulating the interaction with Fc receptors is played by the N-glycosylation, a post-translational modification that occurs at the conserved Asn297 in the Fc region (9, 10). It is well known that differences in the N-glycosylation pattern, such as the presence of fucose, can modulate the Fc receptors recognition (11, 12) and several published studies demonstrated a role of this modification not only in modulating the bioactivity of the mAbs (13–20), but also their conformational behavior (21, 22).

Although Fc glycosylation is always present on IgGs and represents an important aspect to be considered during product development and in the post-marketing surveillance, N-glycosylation in the Fab region also needs to be evaluated. Actually, Fab glycosylation has been observed in approximately 15-25% of all circulating antibodies and it is associated with many physiological and pathological conditions (23): e.g., in pregnancy, where Fab-glycosylated IgG antibodies from the mother are directed against paternal antigens (24); in auto-immune diseases like rheumatoid arthritis (25); in primary Sjogren’s syndrome (26); and in B cell malignancies (27–29).

A prerequisite to have a N-glycosylated site in both Fab and Fc is the presence of the N-X-S/T sequence motif, where X is not Pro (9). This condition is however not sufficient to observe the Fab glycosylation, that can happen in both LC and HC, in CDRs, and in framework regions, because it is due to a somatic hypermutation during Ag-specific immune responses (30) and for biotherapeutics it can depend on the chosen germline. For instance, mouse IGKV5-45 allele presents a germline-encoded glycosylation site that is not occupied in both infliximab (31) and cetuximab (32, 33), two therapeutic commercial mAbs. However, the cetuximab produced via mouse IGHV2-2 germline is glycosylated at the same site (32–34). Many factors can also influence Fab glycosylation patterns, including the position (35, 36) and the proximal amino acids (37–39). Even if a certain population of high mannose species has been detected, the most frequent population of Fab glycosylation is based on a biantennary species, which often has one or two terminal α2,6-linked sialic acids, with a pattern that is very similar to the one in the Fc region (40). Several studies report that Fab glycosylation can increase or decrease Ag binding, block binding between two proteins by steric hindrance, extend Ab half-life because of sialylation, and presumably affect Ab aggregation and immune complex formation (23). In the context of biotherapeutic mAbs and to the best of our knowledge, cetuximab, marketed as Erbitux^®, is the only antibody presenting a Fab glycosylation approved world-wide. Targeting the epidermal growth factor receptor (EGFR), it is used for the treatment of metastatic-colorectal cancer and head and neck cancer (41, 42). Cetuximab is a chimeric mouse/human recombinant antibody produced in SP2/0 mouse myeloma cell line (42), presenting a wide structural heterogeneity due to a second non-human N-glycosylation site located at Asn88 within the framework 3 of the variable domain of HC (Fab portion), in addition to the classical site at Fc (Asn299) (43). The functional role of Fab glycosylation in cetuximab is still not clear. It seems to not affect the binding to the antigen because it is on the opposite site with respect to CDRs (33). Moreover, Giddens et al. showed that glycoengineered cetuximab uniformly carrying single fully sialylated Fab N-glycans exhibits the same affinity as commercial cetuximab carrying multiple N-glycans (44). On the other hand, the pharmacokinetic profile of cetuximab can be influenced by Fab properties, i.e., charge variants and N-glycosylation, as demonstrated by Schlothauer and colleagues who observed an increase of affinity between Fc and neonatal Fc receptor (FcRn) after the removal of Fab domains (45). As aforementioned, cetuximab is produced in mouse cells and N-glycans show a high heterogeneity. This aspect is linked to severe hypersensitivity reactions in some patients, especially because the 30% of glycans in the Fab contain one or two α1,3-galactose and a significant content of N-glycolylneuraminic acid (Neu5Gc) that are both considered highly immunogenic (43, 46). Cetuximab is also known to activate the ADCC function, which is claimed as secondary mode of action of the molecule (47–49) and that may be influenced by the surface charge of the antibody, since different charge variants can influence the affinity to FcγRIIIa (50). As already reported in our previous works, N-glycosylation at Fc domain can have not only an impact on FcγRIIIa recognition and ADCC activation, but also on the antibody conformational behavior, influencing the preference of the antibody for certain conformational states (21, 22). Accordingly, the scope of this work, that was entirely developed in silico, is to investigate via computational tools the role of Fab glycosylation on cetuximab dynamics, trying to shed light on its mechanism of action. Accelerated molecular dynamics (aMD) simulations were performed to compare the commercial cetuximab glycosylated at Fab and its aglycosylated form to evaluate the structural descriptors that can modulate the conformational freedom of the antibody, potentially affecting the Fc effector functions. This potential role of Fab glycosylation has never been investigated before and can be a focus for the development of a new class of biotherapeutic mAbs, where the modulation of the effector functions is performed throughout the Fab glycosylation.

2. Materials and methods

3. Results

3.2. Analysis of Fab behavior

3.2.1. Free energy surface along θ angles

Both models, cetuximab and cetuximab without Fab glycosylation, were submitted to aMD simulations 1µs long to find an energy minimum corresponding to a stable conformation. The FES of the two antibodies was plotted along θ angles, Theta1 and Theta2, that describe the position of Fab1 and Fab2 domains with respect to Fc. This descriptor allows the classification of antibody conformations in two categories: Y-shaped and T-shaped. Y-shaped forms present θ values < 90°for both Fab domains, while T-shaped conformations show θ ≥ 90° in at least one Fab (22). Accordingly, in Figure 1 the FES profile of both cetuximab forms is reported, showing not only a different exploration in terms of potential energy surface, but also two different conformations, isolated by cluster analysis in the corresponding energy minimum region ( Supplementary Figure 2 ). Specifically, cetuximab presents a Y-shaped conformation while cetuximab without Fab glycosylation presents a form that can be approximately described as T-shaped. The minimum energy region corresponds to the value of potential of mean force (PMF) < 0.5 kcal/mol. Herein, cetuximab presents both Theta angles ≈ 70°, while the aglycosylated form shows Theta1 ≈ 95° and Theta2 ≈ 70°. Considering this result, a role of Fab glycosylation in driving the conformational behavior of cetuximab can be hypothesized and the following analyses were performed in the minimum energy frames thus identified.

Open in a separate window

Figure 1

Free energy surface of cetuximab and cetuximab without glycosylated Fab along θ angles. The reweighted free energy profile of cetuximab (A) and cetuximab without Fab glycosylation (B) along Theta1 and Theta2 angles. The molecular surface of medoid structures isolated from the minimum energy region (in dark pink) by cluster analysis is also shown. The color bar represents the PMF value in kcal/mol.

3.2.2. Description of Fab rotation

Another descriptor useful to investigate the behavior of Fab domains is the ϕ angle. This angle is representative of Fab rotation and can define the flexibility of this domain considering movements in a different direction than θ angle. To observe the shift of Fab position with respect to the starting frame, Δϕ distribution (Eq. 1) was plotted ( Figure 2 ). Accordingly, as already observed in our previous work (21), Fab domains move asymmetrically, with Fab2 rotating more than Fab1 in both cetuximab with and without Fab glycosylation. Between the two models, cetuximab with glycosylated Fab domains shows the highest difference in terms of their rotational propensity, with Δϕ values around -70° for Fab1 and 300°for Fab2. In Figure 2A , Δϕ distribution for each Fab domain in each cetuximab model is reported, together with a graphical representation of the movement that Fab domains perform with respect to their starting position ( Figure 2B ). The graph shows that the angle between Fabs depends on the species, with aglycosylated Fab domains having a median angle of 93.2° and the glycosylated one of 107.1°, compared to the starting angle of 88.1°, suggesting that this parameter can influence the final conformation of the antibody, too. In Supplementary Figure 3 , the boxplot showing the distribution of Fab angles is reported. To explain the different rotational propensity of glycosylated Fab domains, an investigation of the interaction network between Fab glycans and the antibody was carried out. The number of contacts between glycans and Fab was computed, showing that Fab1 is involved in much more interactions than Fab2 ( Figure 3A ). This result suggests both the role of N-glycans linked to Asn88 in stabilizing the movements of this domain and the asymmetric behavior typical of mAbs ( Figure 3B ). The analysis of H-bonds network between Fab and sugars shows that there is a specific peptide in LC that is involved in the interaction with glycans ( Figure 4 ). This peptide comprises the region from Tyr140 to Asp170 and its interaction network with sugars is less stable for LC2 than for LC1, as suggested by the lower frequency values of the H-bonds ( Supplementary Table 2 ). This asymmetric behavior can explain the different rotational propensity of the two Fab domains, supporting the analysis of contacts number and suggesting a direct structural impact of glycosylation on LC behavior. Accordingly, the secondary structure content of this LC peptide was evaluated in both cetuximab and in the cetuximab without Fab glycosylation showing not only a difference between LC1 and LC2 in the glycosylated form, but also between the two simulated models. As reported in Figure 5A , the presence of N-glycosylation in Fab can induce a rearrangement of the secondary structure of this peptide that, with respect to the aglycosylated form, loses its typical α-helix motif in the region between res. 151-159 in favor of unstructured amino acids (i.e., bend, turn, or no structure). In LC2, there is a complete change of the structure in this region, that from α-helix is converted in extended β-sheets, bend or turn. In Figure 5B the structural representation of Fab domains is reported highlighting also the differences in the LC peptide between the two cetuximab forms. Since the structural proximity of this region to res. 161-191 in HC, also the secondary structure content of this region was evaluated ( Supplementary Figure 4 ). As a result, no big impact was observed, suggesting a stronger effect of Fab glycosylation on LC than on HC. The statistical significance of the observed differences in Fab-sugars contacts and the angle between Fab domains is reported in Supplementary Material (“Statistical significance results” section) and in Supplementary Table 3 .

Open in a separate window

Figure 2

Δϕ distribution and schematic representation of Fab rotation. (A) Δϕ distribution of cetuximab and cetuximab without Fab glycosylation. (B) Schematic representation of the change of ϕ angles from the starting position (continuous lines) to the minimum energy conformation (dashed lines). Lines corresponding to Fab1 are shown in purple, while those corresponding to Fab2 in pink. Black arrows indicate the angle between two Fab domains.

Open in a separate window

Figure 3

Number of contacts between N-glycans and Fab domains and structural representation of the interaction surface. (A) The box plot showing the distribution of contacts number between Asn88 glycans and Fab domains. (B) The structural representation of Fab domains as ribbons and the interaction surface (in blue) between the protein and sugars. N-glycans are shown as light gray sticks.

Open in a separate window

Figure 4

H-bonds between N-glycans and Fab domains. (A) A matrix showing the H-bonds between sugars and mAb colored by chain. The size of the markers represents the frequency of the H-bond in the minimum energy frames: low size corresponds to low frequency, high size to high frequency. (B) Structural representation of cetuximab in the minimum energy conformation with the LC residues involved in the interaction with sugars. The antibody is represented as ribbons colored by chain (in blue and light blue, LC1 and LC2, respectively; in dark magenta and pink, HC1 and HC2, respectively). N-glycans are shown as spheres, while LC residues as sticks.

Open in a separate window

Figure 5

Secondary structure content of LC peptide in cetuximab forms. (A) Bar plots showing the per residue secondary structure content in the LC peptide (res. 140-170) of cetuximab (top) and cetuximab without Fab glycosylation (bottom). The frequency of each secondary structure type according to DSSP classification is reported in percent on y-axis. (B) Structural superposition of glycosylated (in blue) and aglycosylated (in gray) Fab domains showing the differences in the secondary structure of LC peptides.

3.3. Analysis of Fc behavior

The dynamical behavior of Fc fragment was investigated according to two descriptors previously identified (21, 22). The first one is the distance between CH2 domains that is an indicator of Fc opening. According to literature, larger CH2 distance corresponds to open Fc domains and to higher affinity for FcγRIIIa (70). As shown in Figure 6A the two antibodies present different distributions of this distance, with higher values in cetuximab without Fab glycosylation. This suggests that the presence of glycosylation in the Fab domains could potentially lead to close Fc conformations that are less prone to interact with the receptor ( Figure 6B ). The second descriptor is the distance between the center of mass of sugar chains inside the Fc cavity ( Figure 6C ). As previously observed, fucosylated sugars prefer to stay into the Fc hydrophobic pocket more than afucosylated ones (21, 22) suggesting that the lower is this distance the less is the affinity of the antibody for FcγRIIIa. In this case, even if this distance is very small for both systems, cetuximab without Fab glycosylation presents slightly higher values, suggesting that the abrogation of Fab glycosylation can also influence the behavior of sugars at the Fc level, in principle favoring conformations more prone to interact with the receptor ( Figure 6D ).

Open in a separate window

Figure 6

CH2 distance and minimum distance between Fc sugar chains. (A) The density plot of the distance between CH2 domains of Fc. (B) The structural superposition of Fc domains of cetuximab (in steelblue) and cetuximab without Fab glycosylation (in gray). The CH2 distance in the minimum energy structure is also reported. (C) The density plot of the minimum distance between the center of mass of Fc sugar chains. (D) The structural superposition of Fc domains of cetuximab (in light steelblue) and cetuximab without Fab glycosylation (in light gray) showing the position of sugars in the Fc cavity in the minimum energy structure.

3.4. Covariance analysis

An analysis of covariance based on the fluctuation of Cα atoms was carried out to identify the independent correlations among the movements of antibody residues. According to the results, several positive correlations were identified between structured domains both in cetuximab and in cetuximab without Fab glycosylation. Considering the symmetric structure of the mAb, the correlations were classified in inter-chain and inter-halves. Inter-chain correlations refer to the positive correlations between LC and HC belonging to the same half of the antibody, while the inter-halves correlations are those occurring between chains included in different halves ( Figure 7A ). Regarding the inter-chain correlations, as summarized in Figures 7B, C , in both cases they involve regions that are in structural proximity, i.e., VH-VL and CL-CH1. However long-distance correlations were also observed between CL-VH, VL-CH2, CL-CH2 and CL-CH3 in cetuximab, and between CL-Hinge and CL-CH2 in cetuximab without Fab glycosylation. Considering the inter-halves correlations, they are more complex and evident in commercial cetuximab ( Figure 8A ) than in the one without Fab glycosylation ( Figure 8B ). Specifically, looking at LC correlations, they are quite conserved between VL1-VL2 and CL1-CL2, while the correlation between CL1-VL2 is present only in the commercial mAb. Regarding HC, much more correlations among domains are present in cetuximab than in the form aglycosylated at Fab domains, suggesting a direct influence of VH domains behavior on Fc and hinge. VH domains include N-glycosylation at Asn88 site, and this let to hypothesize an allosteric effect of this PTM on the general behavior of the mAb and on its propensity to recognize FcγRIIIa. Globally, the correlation network of the commercial mAb is more complicated than that observed in the aglycosylated antibody, suggesting that Fab glycosylation can generate structural constraints in the whole molecule, inducing a specific conformation.

Open in a separate window

Figure 7

Inter-chain correlations in cetuximab and cetuximab without Fab glycosylation. (A) Schematic representation of a classical IgG1 architecture indicating single domains of each chain. Correlation matrices between LC and HC for each half of cetuximab (B) and cetuximab without Fab glycosylation (C). Black squares highlight strong positive correlations (in yellow) between domains.

Open in a separate window

Figure 8

Inter-halves correlations in cetuximab and cetuximab without Fab glycosylation. Correlation matrices between corresponding LC and HC belonging to the different antibody halves of cetuximab (A) and cetuximab without Fab glycosylation (B). Black squares highlight strong positive correlations (in yellow) between domains. A schematic representation of inter-chain and inter-halves correlations in the two antibodies is reported alongside the correlation matrices.

3.5. Evaluation of FcγRIIIa and FcRn binding sites behavior

An analysis of the structural behavior of hinge and those residues involved in the interaction with FcγRIIIa was performed to evaluate possible effects of Fab glycosylation on the ability of the antibody to bind the receptor and consequently activate ADCC response. Hinge residues were selected according to IMGT convention (71), while residues involved in the receptor binding were selected according to the work by Shields and colleagues (72). The SASA of these regions was computed showing that both in the case of the hinge and in that of FcγRIIIa binding site, cetuximab without Fab glycosylation presents a higher solvent exposure, suggesting a higher propensity to interact with the receptor than the commercial form ( Figures 9A–C ). Also, in this case the asymmetric behavior typical of mAbs was observed since the main differences can be seen for Hinge2 and HC2. Furthermore, charged and hydrophobic patches on the binding site were evaluated in the minimum energy conformations, showing that the type of exposed surface can change between the two forms. In Figure 9D , the total amount of patches area is reported for both cetuximab forms, showing that - in the absence of Fab glycosylation - the total area of negative and hydrophobic patches increases, whereas that of positive patches decreases. Figure 9E shows how in particular the hydrophobic patches determined by Leu234 and Leu235 residues are less exposed in the commercial cetuximab. It is well known that these two Leu residues are critical for receptor recognition since mutagenesis analysis showed a reduction of receptor binding when they are mutated in Ala (“LALA” mutation) (73). Overall, these data suggest that Fab glycosylation can have a negative structural impact on the interaction with the receptor and points again out the key role of this modification in modulating the antibody behavior. In addition, an evaluation of the effect of Fab glycosylation on the pharmacokinetics of the antibody, considering the potential binding to FcRn, was performed. Specifically, the SASA of residues involved in the binding to FcRn, selected based on literature data (74–76), was computed and it is reported in Supplementary Figure 5 together with a structural representation of the binding site in the two conditions. According to these data, the negative impact of Fab glycosylation on the behavior of Fc region is confirmed again. Also in this case, in fact, a higher solvent exposure of the FcRn binding residues is observed in cetuximab without Fab glycosylation (particularly on HC2) than the commercial mAb. This corresponds also to a different structural orientation of the A-B turn in CH2 domain that is more exposed and probably more prone to interact with the receptor. The statistical significance of the observed differences in these data is reported in Supplementary Material (“Statistical significance results” section) and in Supplementary Table 3 .

Open in a separate window

Figure 9

SASA of hinge and FcγRIIIA binding site residues and exposed patches analysis. (A) Boxplot showing the distribution of SASA values for hinge. (B) Structural representation of hinge orientation in cetuximab and cetuximab w/o Fab glycosylation after Fc superposition. (C) Boxplot showing the distribution of SASA values for FcγRIIIa binding residues in the two forms. (D) Analysis of protein patches in the minimum energy structures with a bar plot showing the total area of exposed patches by type. (E) The structural representation of protein patches surface with the position of Leu234 and Leu235 highlighted in the two models. For clarity, only hinge and Fc are shown as ribbons.

4. Discussion

The scope of this study was to evaluate the impact of Fab glycosylation on the conformational behavior of cetuximab (Erbitux^®) that is used against metastatic-colorectal and head and neck cancer (41, 42). The potential structural role of Fab glycosylation in modulating immune effector functions was never investigated before, and cetuximab was chosen as a case study because it is the only mAb glycosylated at Fab currently on the market. The study was entirely developed in silico, applying aMD to the commercial cetuximab and to its form without Fab glycosylation to find out minimum energy conformations of the two species. The results clearly show a role of Fab glycans in modulating the conformation of the antibody because, two different conformations were found: the commercial antibody presents a Y-shaped conformation, while the cetuximab without Fab glycosylation a T-shaped like form. The T-shaped conformation has been already observed for classical IgG1 presenting fucosylated N-glycans at Fc level (22) and, for this reason, this result was expected. On the other hand, the observation of a Y-shaped conformation even in presence of fucosylated Fc glycans suggests that Fab glycosylation can modulate the dynamics of the mAb. In addition, glycosylated Fab domains are influenced by the glycans in terms of rotational propensity and secondary structure content, especially at the LC (e.g., res. 140-170). Considering Fc, a more open Fc conformation in cetuximab without Fab glycosylation with respect to commercial mAb was observed. According to literature data, opened Fc structures are more prone to interact with the receptor and to activate ADCC response (70). Therefore our results suggest a negative impact of this Fab glycosylation on the ability of Fc to interact with FcγRIIIa. Considering all these aspects, a correlation between the presence of Fab glycosylation and the ability of cetuximab to bind the Fc receptor, activating ADCC cascade, was hypothesized. An analysis of long-distance correlations among domains was performed showing that in cetuximab without Fab glycosylation there is a prevalence of linear correlations, likely corresponding to a more flexible behavior of the protein. On the other hand, in the commercial mAb, the correlations are transversal and even more complicated, occurring between variable domains, especially the glycosylated VH ones, and the hinge and Fc portion. This result, together with the observation of a blocked Y-shaped conformation, may suggest the presence of structural constraints induced by Fab glycans. Finally, a higher solvent exposure of hinge residues and of those residues involved in the FcγRIIIa recognition was observed in the case of cetuximab without Fab glycosylation, suggesting a higher propensity of this form to interact with the receptor. Looking at the structure, the hinge also presents a different orientation between the two species. Moreover, the evaluation of charged and hydrophobic patches in the minimum energy structure highlighted an increased hydrophobic exposed area in cetuximab without Fab glycosylation. The hydrophobic patches include Leu234 and Leu235, residues known to be critical for receptor recognition, suggesting in another way the negative impact of Fab glycosylation on ADCC activation. In summary, this study suggests that: i) Fab glycosylation has an impact on the antibody conformation and on the dynamics of the whole cetuximab; ii) the impact may be negative because of the lower values of CH2 distance, hinge mobility and exposure, and SASA of the residues involved in receptor recognition. These results let hypothesize that also glycosylation pattern could differently modulate the ADCC function, as already mentioned in literature (77), an aspect that will be further investigated in the next future. Our results are supported by the work published by Lippold and colleagues, who applying affinity chromatography and mass spectrometry observed a higher affinity of cetuximab Fc to FcγRIIIa than the intact antibody (78). A similar investigation was performed for FcRn binding residues, showing a decrease of the solvent exposure of these residues, and suggesting a negative impact of this modification also on the FcRn recognition.

Moreover, even if no data are available specifically for cetuximab, recent studies have shown that Fab glycosylation can affect the binding of IgG to human FcRn. This is in line with the data published by Schlothauer et al. who, using IdeS or Plasmin digests that were processed via affinity chromatography, clearly observed an effect of Fab domains on the interaction between Fc and FcRn (45) and also with the study performed by Brinkhaus and colleagues who, via cell membrane - based assays and crystallographic data, showed a negative effect of Fab arms on the FcRn recognition, hypothesizing a steric clash between Fab and the membrane (79). Furthermore, a study carried out by Volkov and colleagues on different Fab glycosylated recombinant antibodies demonstrated a negative impact of Fab glycosylation on the IgG-FcRn interaction in a cellular context. In detail, comparing glycosylated Fab antibodies to non-glycosylated Fab ones an increase in the binding to human FcRn was observed in absence of Fab glycosylation (80). For what concerns the antigen recognition, even if it was not directly investigated herein, our data support the idea that Fab glycans do not influence this aspect, also because their localization in the structure is far away and on the opposite site with respect to CDRs. Accordingly, at least steric hindrance effects are not expected. In conclusion, it can be hypothesized that Fab glycosylation, representing the highest heterogeneity of commercial cetuximab, is a critical aspect to consider during the development of biosimilars in relationship with the potential impact in modulating the ADCC, pharmacokinetics, and, due to conformational constraints, all Fc effector functions. Of note, since this study was entirely developed in silico, because of the impossibility to mimic all the in vivo physiological conditions, further experimental studies would be required, especially regarding the pharmacokinetics profile and the ADCC activity of cetuximab. On the other hand, the structural complexity due to Fab glycosylation makes the generation of Erbitux^® biosimilars very difficult (42) and the results presented in this study can not only provide a rationale to explain this complexity but also pave the way to novel strategies of biotherapeutics development.

Data availability statement

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Author contributions

SS: Writing – original draft, Visualization, Investigation, Formal analysis. DB: Writing – review & editing, Visualization, Formal analysis. OB: Writing – review & editing, Visualization, Formal analysis. MR: Writing – review & editing, Resources, Project administration. UG: Writing – review & editing, Software, Methodology. IE: Writing – review & editing, Supervision, Funding acquisition. FC: Writing – review & editing, Supervision, Conceptualization.

Funding Statement

The author(s) declare financial support was received for the research, authorship, and/or publication of this article. DB, OB, UG, and IE were financially supported by grants from MIUR — “Progetto Eccellenza 2023–2027”.

References