2.1. Ligands $\mathbf{𝓛}$

The synthesis of the ligands has been adapted from literature [31, 32, 33, 34, 35, 36] and is described in the Supporting Information.

2.2. Characterizations of Cu($\mathbf{𝓛}$) complexes

The Cu(II) complexes, formed in situ by the mixture of quasi-stoichiometric ratio between $\mathbf{𝓛}$ and Cu(II) stock solutions (see Figure S1 for the determination of concentration of $\mathbf{𝓛}$) were characterized by UV–vis and EPR spectroscopies. Their signatures are shown in Figure 1, panels A and B, while the corresponding parameters are listed in Table 2. All the Cu($\mathbf{𝓛}$) complexes show similar features (note that the complexes are neutral with L, L′ and L_Me, ${\text{L}}_{\text{Me}}^{\prime }$ and di-anionic for ^sL, ^sL′ and ^sL_Me, but that charges will be omitted for clarity). In UV–vis spectra (Figure 1, panel A), d–d bands and phenolato-to-Cu(II) CT (Charge Transfer) transitions are observed near 600 nm and 380 nm, respectively. Some weak differences are observed according to $\mathbf{𝓛}$. In contrast to the d–d band, the LMCT band is not affected by the presence of sulfonato groups on the ligand phenol moiety, whereas the N-methyl group affects it. It should also be noted that the presence of cyclohexyl instead of an ethylene bridge did not affect the LMCT band characteristics. In EPR spectroscopy (Figure 1, panel B), typical spectra of square-planar Cu(II) complexes are obtained, with hyperfine and g-value parameters in line with a 2N2O equatorial site, according to the Peisach an Blumberg correlation [37]. Again, some slight differences in the EPR parameters are observed between the various ligands, whereas the presence of two sulfonato groups on the ligands has virtually no impact on the Cu($\mathbf{𝓛}$) EPR signatures. Again, as for UV–vis spectra, methylation of the secondary amine leads to the most significant difference with a decrease in the hyperfine coupling (A_∕∕) values and an increase in the g_∕∕ values, while grafting of the cyclohexyl on the ethylene diamine bridge has the opposite effect.

In brief, the spectroscopic characterizations indicate that the first coordination sphere of the complexes are identical in buffered solution, but that some minor second sphere structural changes occur.

2.3. Cu(II) removal from Cu(Aβ)

Next, we performed competition experiments, monitored by UV–vis and EPR spectroscopies (Figure 1, panels C and D in case of ligand L and Figures S2 and S3 for the other ligands), to check that the ligands $\mathbf{𝓛}$ were able to remove Cu(II) from Cu(Aβ) including in the presence of Zn. In other words, we check that $\mathbf{𝓛}$ have higher Cu(II) affinity and Cu(II) over Zn(II) selectivity than Aβ. The spectroscopic signatures of Cu($\mathbf{𝓛}$) were recovered during the competition experiments (Cu(Aβ) + $\mathbf{𝓛}$ or Cu,Zn(Aβ) + $\mathbf{𝓛}$). Thus all the ligands $\mathbf{𝓛}$ are efficient in retrieving Cu from Cu(Aβ) regardless of the presence of one equiv. of Zn(II). Hence, the thermodynamic parameters of $\mathbf{𝓛}$ are suitable for performing Cu(II) removal from Aβ including in the presence of one equiv. of Zn(II). However, during the course of the competition experiment, we noticed that the thermodynamic equilibrium is reached more slowly in the case of the ligand ${\text{L}}_{\text{Me}}^{\prime }$ (about 30 min versus less than 5 min for the others ligands, Figure S4).

2.4. Effect of ligands $\mathbf{𝓛}$ on Cu(Aβ)-induced ROS production

Finally, Cu(Aβ)-induced ROS formation was evaluated using a very suitable and straightforward experiment, namely the ascorbate (Asc) consumption assay. This is an appropriate method to monitor ROS formation. Briefly, it consists in measuring the absorption of Asc (at 265 nm, 𝜀 = 14,500 M⁻¹⋅cm⁻¹) that is the reductant fueling the incomplete reduction of O₂ to ${\text{O}}_2^{°-}$, H₂O₂ and HO^° [21], and that does not absorb once oxidized. It has been previously shown that Cu and Cu(Aβ)-induced Asc consumption mirrors the formation of H₂O₂ and HO^° [38, 39]. A ligand will be efficient in stopping Asc consumption if it can bind Cu(II) or remove it from Aβ, and then form a Cu(II) complex resistant to reduction by Asc, eventually interrupting the redox cycle of Cu and thus the production of ROS.

With Asc consumption assays, two distinct experiments can be performed, either by incubating the various chemical partners at play and then triggering the reaction by addition of Asc (named “experiments A”, Figure S5 and Figure 2, panel D), or by adding ligand $\mathbf{𝓛}$ during Cu(Aβ)-induced Asc consumption (named “experiments B”, Figure 2, panels A to C). In experiments A, if the incubation time is long enough, the results indicate the ability of $\mathbf{𝓛}$ to remove Cu from Cu(Aβ) including in the presence of increasing stoichiometry of Zn(II), thus documenting the thermodynamics of the reaction of Cu(II) extraction out of Aβ. This is a very straightforward way to evaluate whether the selectivity of a ligand L is appropriate [22, 23]. More specifically, such a method is more suited to screen between various ligands than the individual determination of Cu(L) and Zn(L) formation that would also release the selectivity of the L (see introduction). In experiments B, additional kinetic parameters are involved (Scheme 2A) [40, 41, 42, 43]. Indeed, during Asc consumption, Cu oscillates between the +I and +II redox states. To be efficient, the tested ligand thus has to be faster in removing Cu(II) from Aβ than the reduction of Cu(II)(Aβ) to Cu(I)(Aβ) by Asc. In the $\mathbf{𝓛}$ series, we wanted to decipher the importance of Zn stoichiometry in both the thermodynamics (experiments “A”) and kinetics (experiments “B”) of Cu(II) removal. In the following, all the experiments were performed with 10 μM of Cu(II) ions, 1.2 equiv. of Aβ and/or $\mathbf{𝓛}$, and 0, 1 or 10 equiv. of Zn(II). Note that for practical reasons, the experiments are run with a slight excess of Aβ and/or $\mathbf{𝓛}$ to avoid any possibility of unbound Cu(II) ions and/or Aβ-bound Cu(II) complexes, respectively. Indeed, this would dramatically change the rate of Asc. consumption.

In the absence of Zn and after short incubation (300 s) (Figure S5, panel A), all the ligands except ${\text{L}}_{\text{Me}}^{\prime }$ are able to stop Cu(Aβ) induced Asc consumption. After a longer incubation time (>18 h), ${\text{L}}_{\text{Me}}^{\prime }$ becomes efficient (Figure S6). This indicates that Cu(II) removal from Aβ by ${\text{L}}_{\text{Me}}^{\prime }$ is slower than that with the other $\mathbf{𝓛}$ ligands, but thermodynamically possible, in line with the competition experiments previously described. In addition to showing that all the ligands are able to extract Cu(II) from Aβ, these experiments confirm that all the Cu($\mathbf{𝓛}$) complexes formed are resistant to reduction by ascorbate. In line with these first results, all the ligands, except ${\text{L}}_{\text{Me}}^{\prime }$ and to a lesser extent L′ and L_Me, are able to stop Cu(Aβ)-induced Asc consumption when added during the course of the experiments (Figure 2, panel A).

2.5. Effect of ligands $\mathbf{𝓛}$ on the Cu(Aβ)-induced ROS production in the presence of Zn(II)

In the presence of Zn, the differences between the various ligands appear more clearly. This may be due to the additional competition reaction between Cu(II) or Zn(II) removal from Cu,Zn(Aβ) (Scheme 2, panel B). With a short incubation (300 s) (Figure S5, panel B) or when the ligand is added during Cu(Aβ)-induced Asc consumption (Figure 2, panel B), the trends are the same. The “efficiency trend” (i.e. the ability of the ligand to stop Cu(Aβ)-induced Asc consumption in the presence of Zn) is as follows: L ∼ ^sL > L_Me ∼^sL_Me ≳ ^sL′ ≳ L′  ≫ ${\text{L}}_{\text{Me}}^{\prime }$. With higher stoichiometry of Zn(II) (10 equiv.), the differences are more obvious and only the L and ^sL ligands maintain their ability to stop Cu(Aβ)-induced Asc consumption (Figure 2, panel C).

Hence, it appears here that Zn(II) induces a change in the ability of the ligands $\mathbf{𝓛}$ to stop Cu(Aβ)-induced ROS production and that this effect is dependent on the ligands $\mathbf{𝓛}$, with a dependence that is more clearly revealed in the presence of 10 equiv. of Zn(II). Two explanations can be proposed (i): when increasing the ratio of Zn(II), the thermodynamic equilibria at play change and the ligand becomes thermodynamically unable to remove Cu(II) from Aβ, or (ii) Zn(II) has a kinetic impact on Cu extraction from Aβ. This kinetic effect would be due to the faster formation of the Zn($\mathbf{𝓛}$) complexes that have further to swap their metallic center with Cu(Aβ) (Scheme 2, panel B). To discriminate between these two hypotheses, we also performed experiments A with a longer pre-incubation time of Cu(Aβ) + $\mathbf{𝓛}$ in the presence of 10 of equiv. Zn(II). Several incubation times were tested: 1, 3, 18, 24, and 48 h (Figure 2, panel D and Figure S7). The results obtained show that with a sufficient incubation time, all the ligands prevent Cu(Aβ)-induced ROS production and thus that the overall effect of Zn(II) is kinetic. For all ligands, except ${\text{L}}_{\text{Me}}^{\prime }$ which requires at least 24 h of incubation, (Figure 2D), 1 h of incubation is sufficient to prevent Asc consumption (Figure S7).

3.1. Effect of p-sulfonation on phenol rings

The sulfonated version of three ligands were tested. Sulfonation has no significant effect on the properties of the parent ligand regarding Cu(Aβ) and Cu,Zn(Aβ)-induced ROS production. This makes possible the use of such water-soluble counterparts for in vitro investigations, while the parent ligands could be engaged in in vivo experiments.

3.2. Structure–activity relationship

3.2.1. Ligands other than ${\text{L}}_{\text{Me}}^{\prime }$

Ligands other than ${\text{L}}_{\text{Me}}^{\prime }$ are all able to stop Cu(Aβ)-induced ROS production in the absence of Zn(II) when added in the course of Cu(Aβ)-induced Asc. consumption (Experiments “B”, Figure 3, first set of columns). Besides, in the presence of increasing ratios of Zn(II) and with long enough incubation times, they all maintain this ability (Figure 3, fourth set of columns, Figures S5 and S7). Finally, when added during the course of Asc consumption, some of them become less and less efficient as the stoichiometry of Zn(II) is increased (Figure 3, second and third sets of columns).

To explain such observations, the chemical reactions shown in Scheme 3 are proposed. According to Scheme 3, the overall Zn(II)-induced slowdown of Cu(II) removal out of Aβ (and consequent effect on Cu(Aβ)-induced Asc consumption) could be due to both thermodynamic and kinetic factors for each individual reaction at play, the determination of which is beyond the scope of the present study.

Reaction (1) corresponds to Cu(II) removal from Aβ by $\mathbf{𝓛}$ in the absence of Zn(II). Then, in the presence of Zn(II), the addition of $\mathbf{𝓛}$ to Cu,Zn(Aβ) leads to the two possible reactions (2) and (3) and the formation of Cu($\mathbf{𝓛}$) or Zn($\mathbf{𝓛}$), respectively. From a thermodynamic point of view, reaction (2) is much more favored than reaction (3) as probed by the experiments with long incubation times (Experiments “A”). However, reactions (2) and (3) are in kinetic competition, with reaction (3) being faster than reaction (2) because otherwise the presence of Zn(II) would have no strong effect on experiments “B”. The progressive loss of the ability to stop Cu(Aβ)-induced Asc consumption with increasing levels of Zn(II) (Figure 3) thus mirrors the impact of Zn(II) on the overall rate of Cu(II) removal from Aβ. Such an effect may be due (i) to the kinetic competition of Zn(II) with Cu(II) for binding to $\mathbf{𝓛}$ (reaction (2) versus reaction (3)) and/or (ii) to a modification in the rate of reaction (6) (Scheme 3).

With the series of ligands $\mathbf{𝓛}$, Zn(II) induces an overall slowdown of Cu(II) removal. This contrasts with a previous study on peptide-based ligands [44], where an acceleration effect due to Zn-induced Cu expelling from Cu,Zn(Aβ) was reported (corresponding to the path shown by reactions (4) and (4′)). This indicates that reaction (4) is not predominant in the case of $\mathbf{𝓛}$. Hence, it will not be considered in the rest of the discussion.

The best ligands able to resist up to 10 equiv. of Zn are L and ^sL, whereas ^sL′, L_Me, and ^sL_Me work correctly only in the presence of one equiv. of Zn(II). Hence, the cyclohexyl-grafted ethylene bridge and, to a lesser extent the N-methyl substituents, increase the level of Zn($\mathbf{𝓛}$) versus Cu($\mathbf{𝓛}$) at a given time (before the thermodynamic equilibrium is reached). The weaker effect of L′ on the arrest of Cu(Aβ)-induced Asc consumption may be in line with the reported cis-β arrangement (i.e. with the two phenolato moiety being in cis) of ligand L′ in [Cu₂(HL′)₂]²⁺ species [45] unobserved with other ligands, for which trans arrangements (i.e. all the endogenous atoms from $\mathbf{𝓛}$ are in the equatorial plane) were reported [25, 46, 47, 48]. We can indeed anticipate that such cis-β arrangement is more appropriate for Zn(II) versus Cu(II) binding in contrast to the trans arrangement which is more appropriate for Cu(II) versus Zn(II).

3.3. Kinetic impact of Zn(II)

In brief, the study of the series of ligands $\mathbf{𝓛}$ reveals an impact of Zn(II) on Cu(II) extraction from Aβ that is apparent during the Asc consumption experiments and which is beyond only thermodynamic considerations. Indeed, if Cu,Zn(Aβ) and the ligands are incubated long enough, all the ligands can prevent Cu(Aβ)-induced ROS production. However, in the course of Asc. consumption (no incubation) or when the incubation is short (300 s), ligands $\mathbf{𝓛}$ can be differentiated by their efficiency to stop Cu(Aβ)-induced ROS production in the presence of various Zn(II) levels. Their ability to stop Cu(Aβ)-induced ROS production strongly depends on subtle variations in the ligand scaffold because all the ligands studied here were built on the very same first coordination sphere. Thermodynamic selectivity is now a parameter considered in the design of Cu-targeting ligands in the context of AD [21, 22]. In the present study, we also show that it is important to consider the effect of Zn(II) which can modify the rate of Cu(II) extraction from Aβ by a ligand and can thus prevent the formation of the corresponding Cu(II) complex resistant to Asc. reduction and the associated arrest of Cu(Aβ)-induced ROS production.

3.4. Perspectives

Among the tested ligands, the simplest L appears to be the ideal candidate with respect to its ability to stop Cu(Aβ)-induced ROS formation in the presence of a biologically relevant ratio of the competing Zn(II) ion. Further studies on this ligand series will aim to relate the in vitro data obtained here with their ability to relieve the cellular toxicity induced by ROS produced by Cu(Aβ) with and without Zn, as reported for other ligands [50, 51, 52, 53, 54]. Since L is expected to be partly neutral at pH 7 based on potentiometric data reported for ^G∕SL_Me [24, 25], and thus to fulfill Lipinski’s rules, it appears as a good candidate for blood–brain barrier penetration by passive diffusion and thus for further in vivo studies on AD animal models.