Click‐Functionalized Cyanine Fluorogenic Dimers for Improved Detection of GPCRs: Application to Imaging of ApelinR in Living Cells

Abstract

Fluorogenic dimers enable background‐free imaging of biological targets under wash‐free conditions owing to a strong fluorescence enhancement in the apolar cell microenvironment. However, it is crucial that the imaging probe interacts solely with the target receptor to avoid nonspecific interactions and ensure detection with a high signal‐to‐noise ratio. Herein, we describe a convenient and rapid approach for the synthesis of various functionalized cyanine dyes by click chemistry allowing the fine‐tuning of the physicochemical and fluorogenic properties of the dimers. A structure‐interaction relationship study was conducted for the fluorogenic dimers in the presence of bovine serum albumin (BSA) and liposomes as models of serum proteins and cell membranes. We identified d─Cy─E which combined the lowest nonspecific interactions with the optimal fluorescence turn‐on properties. By conjugating d─Cy─E to a peptide ligand of the apelin GPCR, we developed Ap─d─Cy─E, the first fluorescent turn‐on probe for the background‐free imaging of this receptor in living cells.

A click‐mediated functionalization of a cyanine 5.5 bearing a terminal alkyne gave a rapid access to a library of ten new cyanine dyes and their corresponding dimers. A structure‐interaction relationship study led to the identification of an optimal dimer which was conjugated to the apelin‐17 ligand and enabled imaging of ApelinR under no‐wash conditions.

1. Introduction

Fluorescence has become a key imaging technique to explore biological phenomena, such as ligand–receptor interactions, at the cellular level with high spatial and temporal resolution.^{[ 1} , ^{2 ]} However, it relies on the performance of fluorescent probes that can light up their fluorescence upon binding to target (bio)molecules with high specificity and affinity.^{[ 3} , ⁴ , ^{5 ]} G protein‐coupled receptors (GPCRs) are the most prominent class of membrane proteins which have been identified as the prime target of more than 30% of drugs currently on the market.^{[ 6} , ⁷ , ^{8 ]} Therefore, understanding their signaling and imaging their distribution in cells and tissues is crucial in the search for new therapeutic approaches.^{[ 9 ]} The use of GPCR ligands labeled with fluorescent dyes is a common strategy to assess expression levels of receptors on the cell surface.^{[ 10 ]} However, the main drawbacks of classical “always‐on” GPCR fluorescent probes are the poor signal‐to‐noise ratio caused by residual fluorescence of the unbound probe in the extracellular medium and the necessity to perform multiple washing steps before imaging. In this context, designing new fluorogenic probes capable of switching on their fluorescence only when interacting with their biological targets has become of major interest.^{[ 4} , ¹¹ , ¹² , ^{13 ]}

Among the strategies used to design such probes, an attractive approach is to use dimeric dyes (fluorogenic dimers) based on flat aromatic scaffolds with a tendency to form poorly fluorescent folded intramolecular H‐aggregates in aqueous media and unfold in less polar media to recover their fluorescence.

Over the years, our group has developed various fluorogenic dimeric probes based on squaraine and cyanine dyes to image GPCRs such as the oxytocin receptor at the surface of living cells and in vivo (Figure 1).^{[ 14} , ^{15 ]} This approach was also successfully applied for sensing biotin receptors,^{[ 16 ]} targeting nucleic acid aptamers,^{[ 17 ]} as well as imaging biomembranes with superior resolution.^{[ 18 ]}

Figure 1.

Schematic representation of fluorogenic dimers based on the previously reported cyanine Cy─A and the novel functionalized fluorophores (Cy─C─L) obtained by click reaction on the alkyne derivatives Cy─B.

Aromatic dyes such as cyanine 5.5 (Cy─A) are of particular interest for bioimaging due to their near‐infrared fluorescence and high brightness. However, their high lipophilicity may lead to nonspecific interactions with hydrophobic domains of circulating plasma proteins and cell membranes resulting in background noise during imaging. To tackle this issue, we have recently reported a dimeric probe based on the Cy5.25 fluorophore (Figure 1). It was characterized by minimized nonspecific interactions with albumin and lipid bilayers and allowed for wash‐free imaging of the oxytocin receptor directly in cell growth media.^{[ 19 ]} Although the Cy5.25 dimer exhibited high brightness, its fluorogenicity was reduced, and its spectra were blue‐shifted compared to the Cy5.5 dimers, limiting its potential for translation to in vivo studies. Thus, in this study, we aimed at functionalizing Cy5.5 dyes to reduce nonspecific interactions of dimeric probes while preserving the near‐infrared fluorescence properties of the scaffold. Since the synthesis of functionalized Cy5.5 dyes requires multiple tedious steps, does not tolerate a wide range of functional groups, and can be time‐consuming, we developed Cy─B, a Cy5.5 dye bearing terminal alkyne, for straightforward introduction of various functional groups via a CuAAC click chemistry strategy for the late stage functionalization. Following this approach, ten new cyanines Cy─C─L and their corresponding dimers bearing anionic, cationic, zwitterionic, neutral, or peptidic substituents were synthesized. A structure‐interaction relationship study was conducted on dimers to evaluate the impact of these functional moieties on their fluorogenic properties and nonspecific interactions with serum albumin and liposomes as models of cell membranes. This study resulted in the selection of the functionalized dimer with optimal properties, which was then conjugated to apelin‐17, a peptide targeting APJ GPCR (ApelinR). ApelinR and its endogenous ligand apelin‐17 represent promising targets for the development of innovative therapeutic strategies for the treatment of cardiovascular and kidney diseases.^{[ 20} , ²¹ , ²² , ^{23 ]} It has also been demonstrated that the ApelinR pathway is involved in tumor growth, angiogenesis, and metastasis development.^{[ 24 ]} In this context, a clear understanding of the expression and regulation of the ApelinR system appears to be crucial for diagnosis and therapeutic purposes. To date, only few fluorescent apelin derivatives have been reported and there is an increasing need for the development of more efficient imaging tools to better understand the biological role and distribution of ApelinR in living cells and in vivo.^{[ 25} , ²⁶ , ²⁷ , ^{28 ]} Therefore, we decided to use the rationally designed Cy5.5 dimer to obtain the first dimeric turn‐on probe for the detection of ApelinR in living cells.

2. Results and Discussion

2.1. Synthesis of Functionalized Cyanines and Their Dimers

To evaluate the impact of the functionalization of the cyanine core on non‐specific interactions, we first designed and synthesized the key intermediate Cy─B bearing a terminal alkyne on one of the benzoindole moieties. Its structural framework is derived from that of the dyes designed and optimized from our previous studies, (Figure 1) incorporating a carboxylic acid group essential for reacting with the terminal amino group of the tweezers, along with a flexible PEG8 chain to ensure solubility of the dye.^{[ 14} , ¹⁵ , ^{19 ]} As the nitrogen heteroatoms of each benzoindolium units are already alkylated with these previously mentioned functional groups, we introduced the alkyne moiety at position 1 of one of the benzoindoles (Scheme 1). To this end, gamma‐acetylenic ketone 1 was condensed with 2‐naphthylhydrazine to give indole 2 in 64% yield. Next, indole 2 was alkylated with 6‐bromohexanoic acid to give the carboxy‐indolium salt 3. Finally, cyanine Cy─B was obtained by a stepwise condensation of functionalized indoliums 3 and 4 on the polyene‐chain with a yield of 62% over two subsequent steps. This key intermediate in hands, a series of ten azide derivatives X─N₃ (see Scheme S1 in Supporting Information for detailed structures) bearing anionic, cationic, zwiterrionic, neutral, or peptidic substituents were reacted with Cy─B under copper‐catalyzed alkyne–azide cycloaddition (CuAAC) conditions using CuSO₄ and sodium ascorbate in a mixture of DMF and water as a solvent. The progress of the reactions was monitored by analytical reverse‐phase HPLC. All click reactions were completed in 15 minutes at room temperature to give ten new click‐functionalized cyanines, Cy─C to Cy─L, with yields ranging from 31% to 93%.

Scheme 1.

Synthesis of cyanine dyes and their corresponding dimers. (i) AcOH, 100 °C, 16 hours. (ii) NaI, ACN, 80 °C, 16 hours. (iii) Malonaldehydedianilide, 100 °C, 3 hours, then 4, pyridine, 60 °C, 16 hours. (iv) Cy─C─L were synthesized by CuAAC click reaction from the appropriate azide derivative X─N₃ and Cy─B, CuSO₄, sodium ascorbate, DMF/H₂O, 25 °C, 15 minutes. (v) d─Cy─A─L were synthesized from 5, 2 equivalents of corresponding Cy─A─L, PyBOP, DIPEA, DMF, 25 °C, 2 hours (except for d─Cy─C). (vi) d─Cy─C was synthesized by CuAAC click reaction from 2 equivalents of (3‐azidopropyl)phosphonic acid, d─Cy─B, CuSO₄, sodium ascorbate, DMF/H₂O, 25 °C, 15 minutes.

Next, click‐functionalized cyanines were dimerized using the bifunctional tweezers 5 composed of a lysine residue and a PEG4 linker (Scheme 1). With such a design, the indole moieties of two cyanines are separated by 19 atoms, which should offer the dyes enough flexibility to allow the dimer to open in the apolar environment and to efficiently close in an aqueous medium. On the other hand, the presence of the PEG4 linker should limit the impact of the bulky dyes on the pharmacological properties of the ApelinR ligand. Dimers d─Cy─A to d─Cy─L were obtained by amide coupling between 2 equivalents of the appropriate cyanine dye and the two free amino groups of the tweezers 5 under PyBOP in situ activation with 29%–79% isolated yields. It is worth mentioning that the direct coupling of the phosphonate cyanine Cy─C to 5 was not compatible with these experimental conditions and led to the formation of several unidentified by‐products. However, our versatile strategy enabled the cyanine functionalization after the formation of the dimer d─Cy─B bearing two alkyne groups. Therefore, d─Cy─C was obtained by CuAAC between (3‐azidopropyl)phosphonic acid and d─Cy─B with 87% yield.

2.2. Photophysical Properties of Functionalized Cyanines and Their Dimers

First, the photophysical properties of new cyanines 5.5 were evaluated in a series of organic and aqueous solvents of different polarities (Table S1). The molar absorption coefficients measured at the absorption maximum in MeOH were very close to that of the parent dye Cy─A, ranging from 207, 000 to 224, 000 M⁻¹cm⁻¹. The functionalization did not influence the absorption and emission maxima of the cyanine dyes, which were situated around 686 and 710 nm, respectively, in MeOH. All the cyanines were emissive in organic solvents with fluorescence quantum yields ranging from 28% to 36% in DMF and with slightly lower quantum yields in water (ranging from 9% to 13%). Thus, the new functionalized cyanines preserved the near‐infrared character and the brightness of the parent Cy5.5 dye, and, similarly to the parent dye, displayed a non‐negligible fluorescence in aqueous media (Figure 2A).

Figure 2.

Examples of absorption (solid lines) and fluorescence (dashed lines) spectra of a fluorescent monomer Cy─E (A) and its corresponding dimer d─Cy─E (B) in DMF and water.

Then, the photophysical properties of the dimers were evaluated in the same solvents in order to analyze the impact of ionic or bulky substituents on the formation of H‐aggregates (Table 1 and Table S2). While the absorption spectra of the monomeric cyanine dyes exhibit absorption maxima around 690 nm in water and DMF (Figure 2A), the absorption spectra of the dimeric dyes in water support the hypothesis of the H‐aggregate formation with a distinct blue‐shifted absorption maximum around 635 nm (Figures 2B and S1–S11). The polar‐sensitive fluorogenic properties of the new dimeric probes are also evident in the fluorescence spectra, where the H‐aggregate shows weak emission in water, which is restored upon unfolding in a less polar solvent. Despite the presence of substituents on their aromatic cores, all dimers were able to form intramolecular H‐aggregates in aqueous media resulting in a dramatic decrease of their fluorescence quantum yields (below or close to 1%). Among all the substituents, the introduction of amine (d─Cy─G), triol (d─Cy─I) or peptide chain (d─Cy─L) appeared to have a negative impact on the formation of H‐aggregates, which resulted in slightly higher quantum yield in aqueous media. The possible explanation for the higher quantum yields in water is an under‐optimal π‐stacking of the mentioned dyes likely due to ionic or steric repulsions. On the other hand, in less polar solvents, such as DMF, ethanol or methanol, the H‐aggregates of all the dimers were distorted and the dimers were completely opened, resulting in high fluorescence quantum yields similar to those measured for the monomers (Table S2).

Table 1.

Fluorogenic properties of the dimers.

Probe	QY DMF (%)[ a]	QY Water (%)[ a]	Turn‐on Ratio[ b]
d‐Cy‐A	30.9	0.35	88
d‐Cy‐C	24.8	0.46	54
d‐Cy‐D	26.5	0.52	51
d‐Cy‐E	32.9	0.48	69
d‐Cy‐F	33.6	0.58	58
d‐Cy‐G	33.3	1.0	33
d‐Cy‐H	31.8	0.55	58
d‐Cy‐I	23.8	0.99	24
d‐Cy‐J	33.4	0.75	45
d‐Cy‐K	33.3	0.58	57
d‐Cy‐L	34.8	1.2	29
Ap‐d‐Cy‐A	33.4	0.44	77
Ap‐d‐Cy‐E	30.4	0.53	57
Ap‐m‐Cy‐E	39.7	12.4	3

^a) Relative quantum yields (QY) were measured by using rhodamine 800 in EtOH as a standard.^{[ 29 ]}

^b) Turn‐on was calculated as a ratio of quantum yields in DMF and water. Concentration of dimers was 200 nM.

To quantify the fluorogenicity of the dimers, their relative turn‐on values were calculated as the ratio between the quantum yield of the highly fluorescent form in DMF and the quantum yield of the poorly fluorescent H‐aggregate in water (Table 1). The non‐functionalized dimer d─Cy─A was characterized by a turn‐on ratio of 88. With the exception of the previously mentioned dimers d─Cy─G, d─Cy─I and d─Cy─L exhibiting less efficient H‐aggregation, the functionalized dimers exhibited excellent turn‐on values ranging from 45 to 69, which highlighted their potential for bioimaging under nowash conditions.

2.3. Evaluation of Nonspecific Interactions of Fluorogenic Dimers

After this insight into the fluorogenic behavior of the dimers, we addressed their potential nonspecific interactions with plasma proteins and cell membranes. Such interactions must be minimized as they may cause the disaggregation of the dimer and generate background noise in wash‐free cell imaging experiments. In the present study, we measured the increase of fluorescence of the dimers in the presence of bovine serum albumin (BSA), as a mimic of circulating proteins in the bloodstream, as well as with liposomes–lipid bilayers mimicking the cell plasma membranes. Nonspecific interactions were evaluated by calculating the ratio of the integral fluorescence measured in HBS (HEPES buffered saline) containing 1% BSA or 200 µM DOPC/cholesterol 2:1 liposomes (I) to the integral fluorescence measured in HEPES buffer alone (I₀) under the same experimental conditions. The results shown in Figure 3 demonstrated a strong increase of fluorescence intensity (I/I₀) for the nonfunctionalized dimer d─Cy─A in the presence of both BSA and liposomes (respectively 10‐ and 38‐fold increase). All click‐functionalized dimers d─Cy─C to d─Cy─L were characterized by lower fluorescence increase in the presence of BSA and liposomes, probably due to the introduction of rather polar triazole ring on the dye scaffold, which hampered interactions with hydrophobic binding pockets of BSA and lipid membranes. The presence of ammonium (d─Cy─E), zwitterion (d─Cy─F), tertiary amine (d─Cy─G), PEG (d─Cy─J and K) or peptide (d─Cy─L) substituents on the cyanine core further limited the interactions with BSA and kept the increase of fluorescence I/I₀ below a 3‐fold ratio. The dimers functionalized with ammonium (d─Cy─E), zwitterion (d─Cy─F), PEG8 (d─Cy─K) or peptide (d─Cy─L) groups exhibited the lowest increase in nonspecific fluorescence I/I₀ in the presence of the liposomes, suggesting that they were less prone to open‐up inside lipid cell membranes. After careful examination of these data, d─Cy─E was identified as the optimal dimer for the subsequent imaging studies, as it combined the highest turn‐on fluorescence ratio among all functionalized dimers (i.e., high fluorogenicity) and limited non‐specific interactions with albumin and liposomes.

Figure 3.

Fluorescence integral ratios (I/I₀) for evaluation of nonspecific interactions of the dimers with 1% BSA in HBS (A) and 200 µM liposomes (B) in HEPES buffer. Mean ± SEM from three independent experiments.

2.4. Synthesis and Evaluation of the Fluorogenic Dimer Targeting ApelinR for Live‐Cell Imaging Under No‐Wash Conditions

Next, we aimed to evaluate the potential of the rationally designed dimer d─Cy─E for confocal microscopy imaging of GPCR in living cells under nowash conditions. For the proof of concept, it was decided to conjugate d─Cy─E to the peptide ApelinR agonist apelin‐17 and to compare the properties of the conjugate Ap─d─Cy─E with those derived from the nonfunctionalized dimer d─Cy─A (Ap─d─Cy─A).

First, the tert‐butyl esters of the dimers d─Cy─A and d─Cy─E were cleaved under acidic conditions in dichloromethane/TFA mixture to give the carboxylic acid derivatives 6‐A and 6‐E in quantitative yields (Scheme 2). Next, coupling of these carboxylic acids to propargylamine under in situ activation with PyBOP in the presence of DIPEA in DMF provided the terminal‐alkyne dimers 7‐A and 7‐E. In parallel, an apelin‐17 derivative Lys(N₃)‐Ap17 in which lysine in position 1 was replaced by azido‐lysine was synthesized on solid phase using a standard Fmoc/tBu approach. The N‐terminal position of the peptide was selected for the introduction of the dimer as it was reported to be positioned outside the receptor binding pocket reducing the risk of interfering with its binding properties.^{[ 30 ]} Finally, a CuAAC reaction between the dimers 7‐A and 7‐E and Lys(N₃)‐Ap17 in the presence of CuSO₄ and sodium ascorbate in a DMF/water mixture gave the desired conjugates Ap─d─Cy─A and Ap─d─Cy─E in 65% and 24% yields, respectively.

Scheme 2.

Synthesis of apelin‐17 conjugates Ap─d─Cy─A and Ap─d─Cy─E. (i) TFA, DCM, 25 °C, 1 hour. (ii) Propargylamine, PyBOP, DIPEA, DMF, 25 °C, 1 hour. (iii) Lys(N₃)‐Ap17, CuSO₄, sodium ascorbate, DMF/water, 37 °C, 1 hour.

We then confirmed that the fluorogenic properties of Ap─d─Cy─A and Ap─d─Cy─E were retained upon conjugation to apelin‐17 by measuring the fluorescence quantum yields in DMF and water and calculating the relative turn‐on ratios, which were found to be 77 and 57, respectively (Table 1). In parallel, Ap─m─Cy─E— a monomeric analogue of Ap─d─Cy─E bearing a single fluorophore—was synthesized following a similar synthetic strategy (Scheme 2). Due to its monomeric nature and the inability to form nonfluorescent intramolecular H‐aggregates in water, its turn‐on ratio was determined to be only 3 (Table 1). In order to verify whether the conjugate of Cy─E was still characterized by lower nonspecific fluorescence in the presence of model albumin and cell membranes compared to the Cy─A derivative, we measured their quantum yields in the presence of BSA and DOPC/cholesterol liposomes (Figure 2 and Table S6). In both cases, Ap‐d‐Cy─E displayed almost negligible non‐specific fluorescence turn‐on, whereas the quantum yields for Ap─d─Cy‐A increased up to 7% in the presence of BSA and up to 16% in the presence of liposomes.

Finally, to assess the ability of the conjugates to specifically label ApelinR in living cells, microscopy studies were performed on living HEK293 cells overexpressing ApelinR fused to enhanced green fluorescent protein (EGFP). Both conjugates Ap─d─Cy─A and Ap─d─Cy─E enabled visualization of ApelinR under nowash conditions with a high signal‐to‐noise ratio after 15 minutes of incubation at 37 °C (Figure 4A). As previously observed for apelin‐17,^{[ 22 ]} the ability of both conjugates to trigger ApelinR internalization was confirmed by the observation of intracellular dots in the cyanine and in the EGFP channels. Once internalized the receptor can be either recycled or degraded, thus these dots might likely correspond to endosomes or lysosomes organelles.^{[ 31 ]} This result demonstrated that the incorporation of fluorogenic dimers on the apelin‐17 peptide had no impact on the binding of the conjugates and their agonist character regarding ApelinR internalization. We then evaluated the specificity of ApelinR labeling by performing the same experiment on wild‐type HEK293 cells which did not express ApelinR. Despite the absence of ApelinR, the nonfunctionalized conjugate Ap─d─Cy─A accumulated inside the wild‐type cell line forming bright fluorescent aggregates. Their intensity was comparable to that of the signal recorded using the ApelinR‐expressing cell, making it impossible to distinguish fluorescence coming from the receptor‐bound and the nonspecifically bound fractions of Ap─d─Cy─A. On the other hand, almost no nonspecific staining of wild‐type HEK cells was observed for the functionalized conjugate Ap─d─Cy─E highlighting its high specificity toward ApelinR. This result, visible to the naked eye, was further quantified by calculating the inhomogeneity of the images of wild‐type HEK cells in the cyanine channel expressed as normalized variance σ ²/μ, where μ stands for the mean grey value of the entire image and σ for its standard deviation. The result obtained in Figure 4B confirmed the higher specificity of Ap‐d─Cy─E to its receptor versus Ap─d─Cy─A.

Figure 4.

Confocal microscopy imaging with fluorescent conjugates. (A) HEK293 cells overexpressing the ApelinR fused to EGFP (left) or wild‐type HEK293 (right) cells were stained with dimeric Ap─d─Cy─A,E (200 nM) or with monomeric Ap─m─Cy─E (400 nM) conjugates in HBS supplemented with 0.1% BSA under nowash conditions for 15 minutes at 37 °C. Fluorescence of EGFP is shown in green, fluorescence of the cyanines is shown in red, nuclei stained with Hoechst are shown in blue. Scale bars, 20 µm. (B) Quantification of inhomogeneity of the images of wild‐type HEK293 cells in the cyanine channel by calculating normalized variance (σ ²/μ). Data presented as mean with SD, n = 6. (C) Evaluation of signal‐to‐noise ratio for the images of HEK293 cells overexpressing the ApelinR. Data presented as mean with SD, n = 6.

Next, to demonstrate the importance of the fluorogenic character of the new ApelinR probes for nowash imaging, a comparison study was performed with the monomeric analogue Ap─m─Cy─E under the same labelling and imaging conditions. In both HEK cells expressing ApelinR and wild‐type HEK cells, the excess of the monomeric probe in solution generated a significant background signal visible to the naked eye. Such a strong background signal was the result of the poor fluorogenic character of the monomeric conjugate Ap─m─Cy─E, whose turn‐on ratio is 20‐fold lower than for the dimeric analogue Ap─d─Cy─E (Table 1). Moreover, a significant plasma membrane labeling was observed in the presence of Ap─m─Cy─E without co‐localization with the EGFP signal, suggesting non‐specific binding to the plasma membrane. Quantification of the signal‐to‐noise ratio for the images of ApelinR expressing cells confirmed the visual observation and thus the importance of the fluorogenic character of the probes intended for live‐cell imaging under no‐wash conditions (Figure 4C).

3. Conclusion

In summary, a CuAAC‐mediated functionalization of a cyanine Cy─B bearing a terminal alkyne at position 1 of its indolenine unit gave a straightforward access to a library of ten new Cy5.5 dyes. A series of fluorogenic dimers was synthesized from these click‐functionalized cyanines and a structure‐interaction relationship study enabled the identification of an optimal dimer with a high potential for bioimaging applications. This probe, bearing the quaternary ammonium functional group, combines a strong fluorogenic character and a low nonspecific fluorescence resulting from interactions with model albumin and lipid membranes. To develop the first turn‐on probe for ApelinR imaging, we conjugated the selected dimer to its peptide agonist apelin‐17. The improved performance of the functionalized conjugate Ap─d─Cy─E was demonstrated by confocal microscopy on living cells expressing ApelinR. The conjugate allowed the visualization of ApelinR under nowash conditions with a better specificity than the reference conjugate Ap─d─Cy─A and a higher signal‐to‐noise ratio than the monomeric analogue Ap─m─Cy─E. These results provide new synthetic strategies for the rapid development of innovative fluorescent probes for wash‐free detection of ApelinR. It could be applied to other GPCRs of therapeutic interest for a deeper understanding of their biological roles in both healthy and pathological pathways.

Supporting Information

The authors have cited additional references within the Supporting Information.^{[ 32} , ³³ , ³⁴ , ³⁵ , ³⁶ , ³⁷ , ³⁸ , ³⁹ , ^{40 ]}

Conflict of Interests

The authors declare no conflict of interest

Supporting information

Supporting Information

Data Availability Statement

The data that support the findings of this study are available in the Supporting Information of this article.