α-Hydroxylactams as Efficient Entries to Diversely Functionalized Ferrociphenols: Synthesis and Antiproliferative Activity Studies

The [ferrocene-ene-phenol] motif has been identified as the pharmacophore responsible for the anticancer activity of the family of ferrocene-based molecules coined ferrocifens, owing to its unique redox properties. The addition of imide entities to the historical ferrociphenol scaffold tremendously enhanced the cytotoxic activity of a large panel of cancer cell cultures and preliminary studies showed that the reduction of one of the carbonyl groups of the imide groups to the corresponding α-hydroxylactams only slightly affected the antiproliferative activity. As a continuation to these studies, we took advantage of the facile conversion of α-hydroxylactams to highly electrophilic N-acyliminium ions to graft various substituents to the imide motif of phthalimido ferrocidiphenol. Cell viability studies showed that the newly synthesized compounds showed diverse cytotoxic activities on two breast cancer cell lines, while only one compound was significantly less active on the non-tumorigenic cell line hTERT-RPE1.


Introduction
Breast cancer is a burden for women health, and about 2.1 million were diagnosed in 2018 worldwide. Even if around 75% of early-stage and non-metastatic conditions are now curable, current treatments fail in the case of advanced and metastatic breast cancers [1]. This situation prompts chemists to propose alternatives to current chemotherapeutic agents, among which multifunctional hybrids of ferrocene currently focus particular attention on [2][3][4][5][6][7]. Our group designed a family of ferrocene-based organometallic compounds coined ferrocifens that were initially built up from hydroxytamoxifen, i.e., the main metabolite of tamoxifen, a molecule given for the adjuvant therapy of hormone-dependent breast cancer. Ferrocifens display potent anticancer activity on cell cultures and In vivo on animal models [8] via an original mechanism of action related to the redox properties of the ferrocene entity. Indeed, the [ferrocene-ene-phenol] motif common to ferrocifens has been identified as the pharmacophore owing to its facile oxidation in cellular medium, yielding electrophilic quinone methides [9] that are in turn able to react with nucleophiles in the biological context [10].
More recently, we found out that grafting of imide motifs (phthalimide, succinimide, glutarimide) at the end of a propyl chain carried by the central double bond of ferrociphenol ( Figure 1) markedly amplified the antiproliferative activity on cell cultures with half maximal inhibitory concentrations (IC 50 ) down to 18 nM for compound A on the glioblastoma cell model U87 [11] and 145 nM on the triple negative breast cancer cell line MDA-MB-231 [12]. When phthalimide was switched to succinimide or glutarimide (compounds B and C), the IC 50 on MDA-MB-231 cells dropped down to respectively 35 and with half maximal inhibitory concentrations (IC50) down to 18 nM for compound A on the glioblastoma cell model U87 [11] and 145 nM on the triple negative breast cancer cell line MDA-MB-231 [12]. When phthalimide was switched to succinimide or glutarimide (compounds B and C), the IC50 on MDA-MB-231 cells dropped down to respectively 35 and 70 nM, measured after 5 days of incubation [12]. Compound B was also tested on the NCI-60 panel of human cancer cells and its mean GI50 (half maximal growth inhibition) value was ca. 0.1 µM [13]. Interestingly, an extremely large heterogeneity of response to compound B was observed on a panel of 15 glioblastoma patient-derived cell lines, with IC50 ranging from 10 nM to ca. 30 µM [14]. In vivo, compound B formulated in lipid nanocapsules improved the survival rate and decreased the tumor growth in an orthotopic mouse model of B16F10 melanoma [15]. Mechanistic studies have highlighted that the enhanced cytotoxic activity of imidoferrocidiphenols was linked to the remarkable stability of the corresponding quinone methides via an unusual lone pair-π interaction between one of the carbonyl groups of the imide and the quinonic cycle [13]. On the whole, the new series of molecules appears as highly promising as anticancer drugs.
Herein we report the synthesis of diversely functionalized ferrocidiphenols derived from compound A, using the imide motif as an entry to the grafting of substituents carrying various functional groups via the well-known N-acyliminium ions chemistry. Indeed, N-acyliminium ions can be generated in situ from α-hydroxyamides using Lewis or Bronsted acids. As powerful electrophiles, these species can be trapped by various nucleophiles inter- [16] or intramolecularly [17,18], yielding α-substituted amides. Standard cell viability assays were performed on the two breast cancer cell lines MCF-7 and MDA-MB-231 and the results were compared with the ones obtained on the noncancerous cell line hTERT-RPE1.

Synthesis of Ferrocenyl Alkylimides 3
We carried out a preliminary investigation of the reactivity of N-acyliminium ions starting from simple models of the imidopropyl ferrocidiphenols A-C. Ferrocenyl alkylimides 3 carrying a succinimide, phthalimide, 2,3-naphthalenecarboximide, glutarimide or 1,8-naphthalimide substituent linked to the ferrocene by an alkyl chain of one to four methylene units were prepared with the purpose of investigating the reactivity of the imide groups and the effect of the ferrocenyl group ( Figure 2). Mechanistic studies have highlighted that the enhanced cytotoxic activity of imidoferrocidiphenols was linked to the remarkable stability of the corresponding quinone methides via an unusual lone pair-π interaction between one of the carbonyl groups of the imide and the quinonic cycle [13]. On the whole, the new series of molecules appears as highly promising as anticancer drugs.
Herein we report the synthesis of diversely functionalized ferrocidiphenols derived from compound A, using the imide motif as an entry to the grafting of substituents carrying various functional groups via the well-known N-acyliminium ions chemistry. Indeed, N-acyliminium ions can be generated in situ from α-hydroxyamides using Lewis or Bronsted acids. As powerful electrophiles, these species can be trapped by various nucleophiles inter- [16] or intramolecularly [17,18], yielding α-substituted amides. Standard cell viability assays were performed on the two breast cancer cell lines MCF-7 and MDA-MB-231 and the results were compared with the ones obtained on the non-cancerous cell line hTERT-RPE1.

Synthesis of Ferrocenyl Alkylimides 3
We carried out a preliminary investigation of the reactivity of N-acyliminium ions starting from simple models of the imidopropyl ferrocidiphenols A-C. Ferrocenyl alkylimides 3 carrying a succinimide, phthalimide, 2,3-naphthalenecarboximide, glutarimide or 1,8naphthalimide substituent linked to the ferrocene by an alkyl chain of one to four methylene units were prepared with the purpose of investigating the reactivity of the imide groups and the effect of the ferrocenyl group ( Figure 2).
Ferrocenyl alkylimides of type 3 were synthesized according to four methods A, B, C, and D starting from 1B-D (methods A and B), 5A-D (method C) or (ferrocenylmethyl)trimethylammonium iodide (method D) as substrates (Scheme 1, Table 1). Ferrocenyl ketones 1B-D were obtained by Friedel-Craft reactions between ferrocene and ωchloroacyl chlorides [19][20][21]. For method A, ketones 1B-D (with n = 2-4) were first reduced by LiAlH 4 and AlCl 3 to afford the ω-chloroalkylferrocenes 2B-D (with n = 2-4) [19][20][21]. Then the imide substituents were introduced by nucleophilic substitution to afford the ferrocenyl alkylimides 3. On the opposite, for method B, the imide substituents were Ferrocenyl alkylimides of type 3 were synthesized according to four methods A, B, C, and D starting from 1B-D (methods A and B), 5A-D (method C) or (ferrocenylmethyl)trimethylammonium iodide (method D) as substrates (Scheme 1, Table  1). Ferrocenyl ketones 1B-D were obtained by Friedel-Craft reactions between ferrocene and ω-chloroacyl chlorides [19][20][21]. For method A, ketones 1B-D (with n = 2-4) were first reduced by LiAlH4 and AlCl3 to afford the ω-chloroalkylferrocenes 2B-D (with n = 2-4) [19][20][21]. Then the imide substituents were introduced by nucleophilic substitution to afford the ferrocenyl alkylimides 3. On the opposite, for method B, the imide substituents were introduced on compounds 1B-D first by a nucleophilic substitution to yield compounds of type 4, prior to selective ketone reduction with triethylsilane and TFA.  Method A was the most efficient and general since the ω-chloroalkylferrocene intermediates 2B-D can be prepared in large quantities. On the contrary, yields in ferrocenyl alkylimides 3 synthesized via method B depended on the imide, as shown in Table 1. We previously used ketone reduction to methylene group by triethylsilane and TFA (reaction iv in Scheme 1) for purely organic compounds that were complete in less than 1 day with a good yield (ca. 75%) [22]. Conversely, even if the ketone reduction in compounds 2B-D was selective, since the carbonyl groups of imides were unaffected, the reaction was very slow and incomplete even after 2 weeks. One possible explanation could be the steric hindrance brought by both the ferrocenyl and the imide groups.
Method C, based on the Mitsunobu reaction between ferrocenyl alcohols 5A-D (n = 1-4 [23]), was only applied to prepare ferrocenyl alkylimides 3 carrying phthalimide or succinimide substituents. Ferrocenyl methylimides 3a and 3e were also prepared via method D adapted from the literature [24] from the common commercial precursor (ferrocenylmethyl)trimethylammonium iodide using a mixture of phthalimide and potassium carbonate instead of potassium phthalimide. Method A was the most efficient and general since the ω-chloroalkylferrocene intermediates 2B-D can be prepared in large quantities. On the contrary, yields in ferrocenyl alkylimides 3 synthesized via method B depended on the imide, as shown in Table 1. We previously used ketone reduction to methylene group by triethylsilane and TFA (reaction iv in Scheme 1) for purely organic compounds that were complete in less than 1 day with a good yield (ca. 75%) [22]. Conversely, even if the ketone reduction in compounds 2B-D was selective, since the carbonyl groups of imides were unaffected, the reaction was very slow and incomplete even after 2 weeks. One possible explanation could be the steric hindrance brought by both the ferrocenyl and the imide groups.
Method C, based on the Mitsunobu reaction between ferrocenyl alcohols 5A-D (n = 1-4 [23]), was only applied to prepare ferrocenyl alkylimides 3 carrying phthalimide or succinimide substituents. Ferrocenyl methylimides 3a and 3e were also prepared via method D adapted from the literature [24] from the common commercial precursor (ferrocenylmethyl)trimethylammonium iodide using a mixture of phthalimide and potassium carbonate instead of potassium phthalimide.

Synthesis of α-Hydroxylactams 6
Reduction of phthalimides 3e-h with sodium borohydride in methanol afforded the α-hydroxylactams 6e-h in almost quantitative yield (Scheme 1, Table 2). Conversely, yields were lower for the succinimide and glutarimide derivatives, owing to the known phenomenon of over-reduction. Indeed, in basic medium, α-hydroxylactams 6 are in equilibrium with their amide-aldehyde open form (Scheme 1), whose aldehyde group can be further reduced to the amide-alcohols 7 [27]. Nevertheless, this over-reduction provides a convenient way to prepare compounds of type HO-(CH 2 ) x -CO-NH-(CH 2 ) y -R, the imide serving as a protecting group for the final alcohol function. To avoid the over-reduction, Speckamp [27] stated that the reduction of succinimide should be done below 5 • C and glutarimide below −10 • C. In these conditions, reactions were rather slow and we chose to run the reduction reaction at room temperature for all imides for yield comparison (Table 2). For the same reason, we did not add hydrochloric acid to limit the formation of the amide-aldehyde open form and its over-reduction [27]. Of note, the α-methoxylactams 10 were sometimes formed as by-products. We did not attempt to further optimize the reaction conditions for the succinimide derivatives since the αhydroxylactams were obtained in acceptable yield (33-66%) by this "fast method" and the removal of by-products was easy. The 2,3-naphthalenecarboximide derivative 3i was difficult to reduce and the 1,8naphthalimide 3k did not react at all. This behavior appears to be a common feature of naphthalimides in these conditions. Furthermore, reduction of the glutarimide derivative 3j gave a complex mixture, i.e., the expected α-hydroxylactam 6j and the starting material, 3j as well as unexpected side products, among which were the enamide 8j resulting from the dehydration of the α-hydroxylactam, the amide-ester 9j resulting from glutarimide ring opening by reaction of methanol under basic conditions, and the amide-alcohol 7j (Scheme 1). Moreover, the α-hydroxylactam 6j was unstable and not fully characterized. Indeed, after isolation, 6j was completely decomposed in one night in CDCl 3 , with the appearance of a dark precipitate. To prevent the formation of this complex mixture and possibly stabilize the unstable 6j, we tried to convert all the reduced cyclic products into α−methoxylactam 10j (Scheme 2) by the addition of hydrochloric acid to the reaction medium, 15 min after the addition of sodium borohydride following a method adapted from reference [27]. Indeed, this should hamper basic hydrolysis of the glutarimide and over-reduction of the αhydroxylactam 6 as stated above. However, the mixture remained still quite complex and the products difficult to separate by chromatography. To prevent the formation of this complex mixture and possibly stabilize the unstable 6j, we tried to convert all the reduced cyclic products into α−methoxylactam 10j (Scheme 2) by the addition of hydrochloric acid to the reaction medium, 15 min after the addition of sodium borohydride following a method adapted from reference [27]. Indeed, this should hamper basic hydrolysis of the glutarimide and over-reduction of the α-hydroxylactam 6 as stated above. However, the mixture remained still quite complex and the products difficult to separate by chromatography.
2.3. Reactivity of N-Acyliminium Ions Derived from α-Hydroxylactams 6 α-Hydroxylactams are known precursors of highly electrophilic N-acyliminium ions that can be further trapped by various nucleophiles. Herein, the formation of N-acyliminium ions by the action of Bronsted acid to α-hydroxylactams 6 was proved via trapping using various nucleophiles. We began by a simple O-nucleophile, namely MeOH, to prepare α−methoxylactams 10 (Scheme 2). However, we noted that the α-hydroxylactams 6 were not very soluble in solvents as DCM or methanol, so they had to be dissolved in THF. Thus, α-methoxylactams 10 were synthesized in mild conditions with a catalytic amount of TsOH at room temperature ( Table 2). Most of the reactions went to completion in less than 1 h, if not less than 15 min, except for the α-hydroxylactams 6a and 6e (n = 1), where no reaction occurred in 1 day. Compound 10e was obtained only after heating under reflux for 5-7 h, while an attempt to generate 10a only gave degradation. The proximity of the ferrocenyl moiety is likely responsible for this lower reactivity.
Let us note that α−methoxylactams 10 are also precursors of N-acyliminium ions with the advantage of being more soluble than α−hydroxylactams 6, which can be convenient for future reactions (Scheme 2).
The N-acyliminium ion derived from the α-hydroxylactam 6e was also successfully trapped by the S-nucleophile thioglycolic acid under reflux for 8 h to afford 11 in good yield (Scheme 2) following a method adapted from our previous work [28]. Treatment of α-hydroxylactam 6e with oxalyl chloride gave the moisture-sensitive α-chlorolactam 12 that was immediately converted to α-aminolactam 13 with an overall 59% yield by the addition of ammonia (Scheme 2). Oxalyl chloride was used instead of the more classical reagent thionyl chloride previously employed for the same purpose on a closely related molecule [28] because of the known incompatibility with ferrocenic compounds (S. Top, personal communication).
Electrophilic N-acyliminium ions can also be trapped intramolecularly by π-nucleophiles following S E Ar reaction to generate polycyclic products [18]. In the ferrocene series, Achari et al. reported the cyclisation of α−hydroxylactam 6f to the air-sensitive tetracyclic γ-lactam 14 in 79% yield after 2 days in the presence of BF 3 .Et 2 O [26]. In our case, heating of a mixture of α−hydroxylactam 6f and TsOH to 70 • C in dichloroethane (DCE) also afforded lactam 14 with an 84% yield after only 10 min (Scheme 2). Interestingly the corresponding α-methoxylactam 10f also afforded lactam 14 in a comparable yield (83%) under the same conditions, even though α−hydroxylactam 6f was not completely soluble in DCE at 70 • C. 1 H and 13 C NMR confirmed the formation of a single diastereoisomer in agreement with the literature [29]. As mentioned, lactam 14 was indeed unstable since its solution in CDCl 3 turned to green in one night. In the same line, α−methoxylactam 10b was converted to the tricyclic γ-lactam 15 as a single diastereoisomer (see crystal structure below). Finally, the crude mixture resulting from the reduction of the glutarimide derivative 3j afforded the tricyclic γ-lactam 16 in 74% yield. This yield was higher than that obtained from 6j alone, indicating that both α-hydroxylactam 6j and enamide 8j contributed to the formation of the N-acyliminium ion intermediate. The assumption was confirmed by the disappearance of 8j observed by TLC of the reaction medium. In the same conditions, and even when eating overnight, 6e did not give the five-membered ring product. We previously reported that the cyclisation of structurally related compounds (thiophene instead of ferrocene) [30] using TFA at room temperature readily occurred for six-membered ring products but hardly for the five-membered ring ones.
Single crystals of tricyclic γ-lactam 15 as a racemate were obtained by slow evaporation of a solution of complex in DCM/cyclohexane mixture at room temperature. The X-ray diffraction analysis of 15 confirmed the formation of a single diastereomer ( Figure 3). As mentioned above, α-hydroxylactams 6 are in equilibrium with their corresponding amide-aldehyde open forms. In this respect, the Wittig reaction between 6e and (carbethoxymethylene) triphenylphosphorane gave the α,β-ethylenic ester that underwent an intramolecular 1,4-Michael addition to form the corresponding cyclic ester, according to a mechanism previously observed for thiophene derivatives [31] that was eventually saponified into the acid 17 in 90% overall yield (Scheme 3). As mentioned above, α-hydroxylactams 6 are in equilibrium with their corresponding amide-aldehyde open forms. In this respect, the Wittig reaction between 6e and (carbethoxymethylene) triphenylphosphorane gave the α,β-ethylenic ester that underwent an intramolecular 1,4-Michael addition to form the corresponding cyclic ester, according to a mechanism previously observed for thiophene derivatives [31] that was eventually saponified into the acid 17 in 90% overall yield (Scheme 3). On the whole, the high electrophilicity of N-acyliminium ions derived from imides yielded original ferrocenyl polycyclic γ-lactams as well as various functionalized products, among which compounds 11, 13, and 17 than can provide interesting starting points for further conjugation of amino acids or peptides by their C-or N-terminus.

Application to Phthalimido Ferrocidiphenol A
Some of the transformations reported above were applied to ferrocidiphenol A. This compound was selected for two reasons. First, reduction of ferrocidiphenol A to the corresponding α−hydroxylactam 18 is more efficient as compared to ferrocidiphenol B, i.e., 92% vs. 58% yield [25]. Second, we also showed that the IC50 of compound 18 on MDA-MB-231 breast cancer cell line was comparable to that of the phthalimide ferrocidiphenol A [25].
The corresponding N-acyliminium ion generated by Brønsted acid treatment of the α−hydroxylactam 18 was readily trapped by methanol to form the α-methoxylactam 19, as already reported (89% yield) [25]. However, the α−hydroxylactam 18 was recovered by stirring 19 in acetone, with water and hydrochloric acid overnight at room temperature, Scheme 3. α-hydroxylactam 6a can also undergo a Wittig reaction. Reagents are i: Ph 3 P = CHCOOEt; ii: K 2 CO 3 .
On the whole, the high electrophilicity of N-acyliminium ions derived from imides yielded original ferrocenyl polycyclic γ-lactams as well as various functionalized products, among which compounds 11, 13, and 17 than can provide interesting starting points for further conjugation of amino acids or peptides by their C-or N-terminus.

Application to Phthalimido Ferrocidiphenol A
Some of the transformations reported above were applied to ferrocidiphenol A. This compound was selected for two reasons. First, reduction of ferrocidiphenol A to the corresponding α−hydroxylactam 18 is more efficient as compared to ferrocidiphenol B, i.e., 92% vs. 58% yield [25]. Second, we also showed that the IC 50 of compound 18 on MDA-MB-231 breast cancer cell line was comparable to that of the phthalimide ferrocidiphenol A [25]. The corresponding N-acyliminium ion generated by Brønsted acid treatment of the α−hydroxylactam 18 was readily trapped by methanol to form the α-methoxylactam 19, as already reported (89% yield) [25]. However, the α−hydroxylactam 18 was recovered by stirring 19 in acetone, with water and hydrochloric acid overnight at room temperature, indicating the reaction was reversible. We extended the range of O-nucleophiles to graft various substituents, for instance linkers carrying useful chemical functions at the other end for future reactions or oligoethylene glycol chains to improve the solubility (Scheme 4). The reaction of 18 with ethylene glycol was performed with 10 eq. diol in the presence of TsOH to prevent double addition to the two hydroxyl groups and displace the equilibrium towards the α−alkyloxylactam. Because of its poor solubility in the THF/DCM mixture, compound 20 precipitated from the reaction medium and was obtained in 75% yield. In the same conditions as for 20, reaction of 18 with 1,6-hexanediol yielded 21 with a 65% yield. Grafting of the hydroxypentyl chain improved the solubility in organic solvents including DCM. Due to water insolubility being a known issue for the ferrocidiphenol family, we exploited this strategy to attach substituents that could increase their solubility in a biologically relevant context. Thus, attaching short and monodisperse PEG chains seemed interesting. The tetraethylene glycol chain was readily grafted to A to afford 22 in 79% yield at room temperature in less than 3 h. Unfortunately, solubility tests performed by shaking suspensions of compounds 18, 20, 22, 24, and 27 in water or in PBS overnight at room temperature, followed by RP-HPLC analysis of the supernatants, indicated that none of these compounds were soluble in aqueous medium (see Supplementary Materials). Unexpectedly, compound 20 was poorly soluble in organic solvents, except for DMSO.
For future development in chemical proteomics, we also grafted terminal alkyne and azide functions to ferrocidiphenol A. A pentynyl chain was grafted to give compound 23 with a 91% yield. An azide function attached to a short oligoethylene glycol linker was also grafted to the ferrocidiphenol scaffold to increase the water solubility for future bioorthogonal chemistry studies in biologically relevant conditions. The starting azido PEG being expensive, only 1.5 equivalents were added to generate 24, instead of 10 equivalents for other nucleophiles. Compound 24 was obtained with only a 28% yield as the probable consequence. However, attempts to displace the equilibrium toward the formation of the alkoxylactam by removing water with magnesium sulfate and then adding calcium chloride to the reaction mixture did not improve the conversion. Thus, the relatively low yield is not only due to the equilibrium between the hydroxylactam and the alkoxylactam. An alternative would be to add the α−hydroxylactam 18 in excess to completely consume the expensive azido-PEG-OH, since 18 is easy to recover.
To prevent the reverse hydrolysis reaction, another strategy could consist in replacing the O-nucleophile by an S-nucleophile. Indeed, the formed α-alkylthiolactam should be less likely to hydrolyze. To prove this assumption, we first attempted the reaction with methyl mercaptopropionate or mercaptopropionic acid to form products 25 and 26 in 83% and 88% yield, respectively. In the same line, we grafted a methoxyterminated short PEG (7 glycol units) to afford compound 27 in 76% yield, using only 1.2 eq. thiol reagent.
Single crystals of compounds of 23 and 25 were obtained by slow evaporation of Due to water insolubility being a known issue for the ferrocidiphenol family, we exploited this strategy to attach substituents that could increase their solubility in a biologically relevant context. Thus, attaching short and monodisperse PEG chains seemed interesting. The tetraethylene glycol chain was readily grafted to A to afford 22 in 79% yield at room temperature in less than 3 h. Unfortunately, solubility tests performed by shaking suspensions of compounds 18, 20, 22, 24, and 27 in water or in PBS overnight at room temperature, followed by RP-HPLC analysis of the supernatants, indicated that none of these compounds were soluble in aqueous medium (see Supplementary Materials). Unexpectedly, compound 20 was poorly soluble in organic solvents, except for DMSO.
For future development in chemical proteomics, we also grafted terminal alkyne and azide functions to ferrocidiphenol A. A pentynyl chain was grafted to give compound 23 with a 91% yield. An azide function attached to a short oligoethylene glycol linker was also grafted to the ferrocidiphenol scaffold to increase the water solubility for future bioorthogonal chemistry studies in biologically relevant conditions. The starting azido PEG being expensive, only 1.5 equivalents were added to generate 24, instead of 10 equivalents for other nucleophiles. Compound 24 was obtained with only a 28% yield as the probable consequence. However, attempts to displace the equilibrium toward the formation of the alkoxylactam by removing water with magnesium sulfate and then adding calcium chloride to the reaction mixture did not improve the conversion. Thus, the relatively low yield is not only due to the equilibrium between the hydroxylactam and the alkoxylactam. An alternative would be to add the α−hydroxylactam 18 in excess to completely consume the expensive azido-PEG-OH, since 18 is easy to recover.
To prevent the reverse hydrolysis reaction, another strategy could consist in replacing the O-nucleophile by an S-nucleophile. Indeed, the formed α-alkylthiolactam should be less likely to hydrolyze. To prove this assumption, we first attempted the reaction with methyl mercaptopropionate or mercaptopropionic acid to form products 25 and 26 in 83% and 88% yield, respectively. In the same line, we grafted a methoxy-terminated short PEG (7 glycol units) to afford compound 27 in 76% yield, using only 1.2 eq. thiol reagent.
Single crystals of compounds of 23 and 25 were obtained by slow evaporation of concentrated solutions in acetone-d 6 at room temperature. X-ray diffraction analysis confirmed the molecular structure of the complexes (Figure 4). A comparison with the previously reported molecular structure of the monophenol analog of A [13] showed that the presence of a substituent on the lactam ring did not markedly alter the overall 3D arrangement of the ferrocidiphenol scaffold.

Oxidation with HRP and H2O2
Enzymatic oxidation of 19, 20, 22, 23, and 25 were performed with a fourfold molar excess of H2O2 in the presence of HRP at pH 8.1. All compounds were rapidly converted to bright pink adducts (λmax = 571-572 nm, Figure S51). This behavior is reminiscent of that previously observed for a structurally related diphenol complex [32], which lets us conclude that enzymatic oxidation of 19, 20, 22, 23, and 25 affords the corresponding quinone methide in the anionic phenolate form, regardless of the substituent grafted from the phthalimide ring.

Antiproliferative Activity
A prerequisite to further bioorthogonal modifications is to preserve the biological properties of ferrociphenol compounds. Then, the impact of some of the grafted substituents on the antiproliferative activity was examined by a classical MTT cell viability assay on the two breast cancer cell lines MDA-MB-231 and MCF-7 as well as on the nontumorigenic cell line hTERT-RPE1 (Table 3). The two cancer cell lines were chosen since most of the previously reported data on ferrocifens were carried out on these cellular models. The experimental protocol is described in Section 3.14., and it is important to emphasize here that the EC50 values reported in Table 3 are not to be compared to the IC50 values reported in the literature, since the assay conditions are different. For this reason, we included as a reference the ferrocenyl derivative of hydroxytamoxifen Fc-OH-TAM-3 [8] that is known to be cytotoxic to numerous cell lines. Furthermore, careful analysis of the dose-response curves showed that some of them displayed biphasic features with two inflection points, corresponding to two inhibitory phases at low (<1 µM) and high (>1 µM) concentration ranges. This feature was taken into account when performing data fitting.
Predictably, compound 3h did not inhibit cell growth for any of the cell models. On the whole, the newly synthesized compounds showed lower EC50 values for MDA-MB-231 cells than for MCF-7, except for compound 25 (see below). With respect to Fc-OH-TAM-3, grafting different substituents from one of the carbonyl groups of the phthalimide impaired the antiproliferative activity of both cancer cells, especially the hormonedependent MCF-7 cells. The EC50 values of compound 22 were similar for the hTERT-   Figure S51). This behavior is reminiscent of that previously observed for a structurally related diphenol complex [32], which lets us conclude that enzymatic oxidation of 19, 20, 22, 23, and 25 affords the corresponding quinone methide in the anionic phenolate form, regardless of the substituent grafted from the phthalimide ring.

Antiproliferative Activity
A prerequisite to further bioorthogonal modifications is to preserve the biological properties of ferrociphenol compounds. Then, the impact of some of the grafted substituents on the antiproliferative activity was examined by a classical MTT cell viability assay on the two breast cancer cell lines MDA-MB-231 and MCF-7 as well as on the non-tumorigenic cell line hTERT-RPE1 (Table 3). The two cancer cell lines were chosen since most of the previously reported data on ferrocifens were carried out on these cellular models. The experimental protocol is described in Section 3.14., and it is important to emphasize here that the EC 50 values reported in Table 3 are not to be compared to the IC 50 values reported in the literature, since the assay conditions are different. For this reason, we included as a reference the ferrocenyl derivative of hydroxytamoxifen Fc-OH-TAM-3 [8] that is known to be cytotoxic to numerous cell lines. Furthermore, careful analysis of the dose-response curves showed that some of them displayed biphasic features with two inflection points, corresponding to two inhibitory phases at low (<1 µM) and high (>1 µM) concentration ranges. This feature was taken into account when performing data fitting. Predictably, compound 3h did not inhibit cell growth for any of the cell models. On the whole, the newly synthesized compounds showed lower EC 50 values for MDA-MB-231 cells than for MCF-7, except for compound 25 (see below). With respect to Fc-OH-TAM-3, grafting different substituents from one of the carbonyl groups of the phthalimide impaired the antiproliferative activity of both cancer cells, especially the hormone-dependent MCF-7 cells. The EC 50 values of compound 22 were similar for the hTERT-RPE1 and MCF-7 cells but compounds 20 and 25 were twice more toxic than the non-tumorigenic cells. Let us note that the ester group in 25 can be easily hydrolyzed in the biological medium, which might have consequences on its cytotoxicity. Interestingly, compound 23 appears as the most active compound of the series and seems to be selective of cancer cells (6-or 3-fold less toxic on hTERT-RPE1 with respect to MDA-MB-231 and MCF-7, respectively). These results are very promising and provide a good omen for future bioorthogonal reactions and chemical proteomics studies.

General
All 1 H and 13 C-NMR spectra were acquired on Bruker 300 and 400 MHz spectrometers. Elemental analysis was performed at the "Service de Micro-analyse ICSN" (Gif sur Yvette, France) or "Service d'Analyses-Chromato-Masse BioCIS-UMR 8076" (Châtenay-Malabry, France). High-resolution mass spectra (HRMS) were performed at the MS 3 platform of Sorbonne Université. Thin-layer chromatography was performed on silica gel 60 GF254. Purification by column chromatography was performed on the Puriflash 430 system (Interchim) using pre-packed silica gel cartridges (Grace). Ketones 1B-D were synthesized according to [19][20][21]. Ferrocenyl alcohols 5A-D were prepared according to reference [23]. The synthesis of compounds 3d, 4d, 18, and 19 was previously reported in reference [25] and the synthesis of compounds 3f, 6f and 14 in reference [26]. Other reagents were obtained from commercial suppliers and used as received.

General Procedure for the Synthesis of Imides 3 (Method A) or 4 (Method B)
In a flask, a mixture of ferrocenyl compounds 1B-D or 2B-D, imide (3 eq.), potassium carbonate (3 eq.), and DMF (10 mL/mol of substrate) was heated to 80 • C under stirring for 2 days. After cooling to r.t., the mixture was poured into an aqueous solution of sodium hydroxide (3 eq. in 80-100 mL of water/mol substrate) and rapidly extracted (to prevent opening of the imide ring) three times with diethyl ether. Organic layers were combined and washed with water and dried with magnesium sulfate. The solution was concentrated under reduced pressure and the residue was chromatographed on silica gel with DCM/petroleum ether 9:1 mixture for 3b-c,f-k or dichloromethane for 4b-d,f-h, affording products as orange-red solids.

General Procedure for the Synthesis of Imides 3 by Reduction of 4 (Method B)
Imides 4b-d,f-h were dissolved into dry dichloromethane. Triethylsilane (3.5 eq.) and trifluoroacetic acid (30 eq.) were added and the solution was stirred for 2 weeks at r.t. The solution was slowly poured into an aqueous solution of hydrogen carbonate under stirring, and solid hydrogen carbonate was added until the gas ceased to form. The mixture was extracted three times with dichloromethane and the combined organic layer was washed with water and dried with magnesium sulfate. After concentration under reduced pressure, the crude mixture was chromatographed on silica gel with DCM/petroleum ether 9:1 mixture to afford pure imides 3b,d,f-h as yellow-orange solids (3c failed).

General Procedure for the Synthesis of Imides 3 by Mitsunobu Reaction (Method C)
To a mixture of phthalimide or succinimide, triphenylphosphine (1 eq.) and hydroxyalkylferrocene 5A-D (1 eq.) in dry THF (1.5 mL/mmol of imide) were slowly added to diethyl azodicarboxylate (DEAD, 1 eq.) in dry THF (1.5 mL/mmol of imide). The reaction mixture was stirred overnight. The solvent was evaporated under reduced pressure and the residue suspended in Et 2 O. The precipitate was filtered, the solvent was evaporated, and the residue was purified by column chromatography on silica gel with DCM/petroleum ether 9:1 mixture to afford pure imides 3b-d,f-h as yellow-orange solids.

General Procedure for the Synthesis of Imides 3 by Method D
This method is adapted from a procedure described in reference [24], replacing potassium phthalimide by a mixture of phthalimide and potassium carbonate (for 3e). To (ferrocenylmethyl) trimethylammonium iodide, potassium carbonate and phthalimide or succinimide, DMF was added (10 mL/mmol of iodide), and the mixture was heated to 80 • C overnight. After cooling to room temperature, the mixture was poured into water and was extracted three times with Et 2 O. The combined organic layer was washed with water then dried with magnesium sulfate. After concentration under reduced pressure, the crude mixture was chromatographed on silica gel with a DCM/petroleum ether 9:1 mixture to afford pure imides. The spectroscopic data were in agreement with reference [24]. This compound has been reported in the literature using another pathway but NMR signals were not attributed [26].

General Procedure for the Reduction of Imides 3 into α-Hydroxylactams 6
In a flask, imides 3 were dissolved in a minimum quantity of THF, then methanol (approx. 50 mL/mmol of imide, the flask should not be filled more than 2/3 of its capacity because of gas emission) was added and stirring was started. Solid sodium borohydride was added portionwise (10-15 eq. each 10 min for 40 min, with control of the amount of emitted gas). The mixture was poured into a sodium hydrogen carbonate solution and extracted twice with DCM. The combined organic layer was washed with water and dried with magnesium sulfate. The solution was concentrated under reduced pressure and the residue was chromatographed on silica gel with dichloromethane as eluent, affording the α-hydroxylactams 6 as orange-yellow solids. 3.6.6. 2,3-Dihydro-3-hydroxy-2-(2-ferrocenylethyl)-1H-isoindol-1-one (6f) From 3f. This compound has been reported in the literature using another method, but NMR signals were not attributed [26]. Yield 99%. 1

General Procedure for the Synthesis of α-Methoxylactams 10
In a flask, α-hydroxylactams 6 were dissolved into a minimum of THF. Methanol (10 mL/mmol of 6) and a spatula tip of TsOH were added. The mixture was stirred and monitored by TLC until substrate disappearance, then a solution of sodium hydrogen carbonate was added. The mixture was poured into water and extracted twice with dichloromethane. The combined organic layer was washed with water and dried with magnesium sulfate. The solution was concentrated under reduced pressure and the residue was chromatographed on silica gel with DCM/petroleum ether mixture (4:1), affording the α-methoxylactams 10 as orange-yellow solids. 3.8. 2,3-Dihydro-3-carboxymethylthio-2-(ferrocenylmethyl)-1H-isoindol-1-one (11) In a flask, α-hydroxylactam 6e was dissolved into a minimum of THF. Dichloromethane (50 mL), thioglycolic acid (0.122 g, 1.3 mmol), and a spatula tip of TsOH were added. The solution was stirred and refluxed for 8 h and monitored by TLC. After cooling, the solution was poured into a solution of sodium hydrogen carbonate and extracted twice with dichloromethane. The combined organic layer was washed with water and dried with magnesium sulfate. The solvent was evaporated under reduced pressure and the residue was dissolved in acetone and concentrated to a minimum volume under reduced pressure. The solution was left overnight in a freezer and the crystals were filtered off, affording compound 11 as an orange-red solid. Yield 76%. 1 13

2,3-Dihydro-3-amino-2-(ferrocenylmethyl)-1H-isoindol-1-one (13)
A solution of ammonia in DCM was prepared by the extraction of 400 mL of concentrated aqueous ammonia solution with 400 mL DCM. The aqueous layer was kept in the funnel for later use and the solution of ammonia in DCM was dried over calcium chloride and filtered. Hydroxylactam 6e (4.58 g, 13.19 mmol) and oxalyl chloride (1.727 g, 1.19 mL, 13.6 mmol) were stirred in dry DCM until the solids disappeared, then the reaction was continued for 30 min to form compound 12. This solution was poured into the previous solution of ammonia in DCM and the mixture was stirred for 10 min. The solution was transferred into the funnel containing aqueous ammonia, then the mixture was made more basic with the addition of sodium hydroxide solution. The organic layer was decanted then conc. HCl was added until the yellow precipitate stopped to appear. The precipitate was poured into a solution of NaOH 10 wt% and the compound was extracted twice with DCM. The combined organic layer was concentrated under reduced pressure to afford pure 13 in 59% yield. Mp: 218 • C. 1  In a flask containing dichloroethane (DCE, 10 mL/mmol of α-hydroxylactam 6 or α-methoxylactam 10), TsOH (spatula tip) was added and the solution was heated at 70 • C. α-hydroxylactam 6 or α-methoxylactam 10 was rapidly added and the reaction was monitored by TLC until disappearance of substrate. After 15 min, the solution was cooled and was poured into a sodium hydrogen carbonate solution and extracted twice with dichloromethane. The combined organic layer was washed with water and dried with magnesium sulfate. The solution was concentrated under reduced pressure and the residue was chromatographed on silica gel with DCM/petroleum ether 2:1 mixture, affording the fused lactams 14-16 as orange-yellow solids.

General Procedure for the Synthesis of Ferrocidiphenols 20-27
In a flask, α-hydroxylactam 18 was dissolved into a minimum of THF, and DCM was added. The alcohol or thiol and a spatula tip of TsOH were added. The solution was stirred at room temperature and the reaction was monitored by TLC. When the reaction was complete, the solution was poured into a solution of sodium hydrogen carbonate and extracted twice with DCM. The combined organic layer was washed with water and dried with magnesium sulfate. The solution was concentrated under reduced pressure and the residue was chromatographed on silica gel or precipitated, affording the pure compounds 20-27 as orange-yellow solids.

Cell Culture and Viability Assay
MDA-MB-231, MCF-7, and hTERT-RPE1 cells were obtained from ATCC. Cells were maintained in a monolayer culture in DMEM with phenol red with Glutamax I (MDA-MB-231) or 2 mM L-glutamine (MCF-7) or 15 mM HEPES (hTERT-RPE1) supplemented with 10% FBS at 37 • C in a humidified atmosphere and 5% CO 2 . Stock solutions of the complexes were prepared in DMSO. Cells were seeded in 96-well plates at a density of 4000 cells/well. After overnight attachment, a dilution series of the compounds were added in the medium, and cells were incubated for a further 72 h. The percentage of DMSO in the culture medium did not exceed 1%. Cell viability was evaluated by using a colorimetric method based on the tetrazolium salt MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide], which is reduced by viable cells to yield purple formazan crystals. After 72 h, the medium was removed and the cells were incubated with MTT solution in PBS (10 µL of a 5 mg/mL) for 2-3 h of incubation. The formed purple formazan crystals were dissolved in 100 µL DMSO by thorough shaking, and the absorbance at 560 nm was read using a microplate reader (FLUOstar OPTIMA, BMG Labtech, Ortenberg, Germany). Each test was performed with at least three replicates and repeated at least three times. EC 50 were determined using Dr Fit software [33].

Conclusions
Our aim was to design a versatile strategy to graft various functionalities to the scaffold of imido-ferrocidiphenols and evaluate their impact on their biological activity. The presence of the imide group provided us with an efficient entry to achieve this aim. Ferrocenyl alkylimides were initially used as simple models to investigate the conversion of imides to α-hydroxylactams by selective reduction of one of their carbonyl groups. These α-hydroxylactams are precursors of highly electrophilic N-acyliminium ions that were efficiently trapped by O-, S-, and π-nucleophiles to afford lactams in good to excellent yield. Translation of this strategy to phthalimido ferrocidiphenol enabled the grafting of various substituents carrying useful chemical functions such as carboxylic acid, alcohol, azide, or terminal alkyne. Cell viability assays on breast cancer cells showed that grafting of various substituents on the phthalimide moiety had diverse effects on the overall antiproliferative activity with respect to the reference Fc-OH-TAM-3. The chemistry of α-hydroxylactams/N-acyliminium ions then appears as an efficient synthetic route to connect functional groups of interest to ferrocidiphenol, such as azide or terminal alkyne. As a short-term perspective, these groups will provide starting points to conjugate bioligands for targeting purposes or for click-based chemical proteomics with the final aim to uncover the protein targets of ferrociphenols.