Synthesis, Microtubule-Binding Affinity, and Antiproliferative Activity of New Epothilone Analogs and of an EGFR-Targeted Epothilone-Peptide Conjugate

A new simplified, epoxide-free epothilone analog was prepared incorporating an N-(2-hydroxyethyl)-benzimidazole side chain, which binds to microtubules with high affinity and inhibits cancer cell growth in vitro with nM potency. Building on this scaffold, a disulfide-linked conjugate with the purported EGFR-binding (EGFR, epidermal growth factor receptor) peptide GE11 was then prepared. The conjugate retained significant microtubule-binding affinity, in spite of the size of the peptide attached to the benzimidazole side chain. The antiproliferative activity of the conjugate was significantly lower than for the parent scaffold and, surprisingly, was independent of the EGFR expression status of cells. Our data indicate that the disulfide-based conjugation with the GE11 peptide is not a viable approach for effective tumor-targeting of highly potent epothilones and probably not for other cytotoxics.


Introduction
Tubulin modulators represent an important class of anticancer drugs that includes microtubule-stabilizing agents (MSA) such as paclitaxel, docetaxel, and ixabepilone as well as microtubule disrupters such as vinblastine and vincristine [1]. However, in spite of their clinical relevance and therapeutic success, drugs acting on the tubulin/microtubule system are still associated with significant side effects, due to the ubiquitous nature of their target protein tubulin [2]. Thus, the specific targeting of tubulin modulators, and cytotoxic agents in general, to tumors, has emerged as an important strategy in anticancer drug discovery and development. The most advanced approach towards the tumor-targeting of cytotoxic agents is their attachment to tumor-specific antibodies, with three antibody-drug conjugates (ADCs) being currently approved for cancer treatment in humans [3,4]; for two of these ADCs, the cytotoxic payload is a microtubule disrupter, i.e., monomethyl auristatin for brentuximab vedotin and the maytansine derivative DM-1 for trastuzumab emtansine. Alternative targeting moieties are peptides [5] or small molecules that interact specifically with proteins on tumor cell surfaces (e.g., folic acid [6] or synthetic organic molecules [7,8]). Epothilones have been the subject of extensive SAR studies and at least eight epothilone-type agents have been advanced to clinical trials in humans in cancer [17,18], including the FDA-approved anticancer drug ixabepilone (the lactam analog of Epo B) [19]; in addition, 12,13-deoxy-Epo B (Epo D) has been investigated in Phase I clinical trials for Alzheimer's disease [20]. While development of most of these compounds has been terminated (including the development of Epo D for Alzheimer's disease), an analog termed utidelone (or UTD1) is currently undergoing Phase III clinical studies for breast cancer treatment in combination with capecitabine [21]. (Very surprisingly (and, in fact, irritatingly), several publications on this compound have appeared in the peer-reviewed literature, without its structure being revealed in any of these papers) [22]. Known attempts at the development of tumor-targeted derivatives of epothilones so far have been limited to the highly elaborate folate conjugate BMS-753493 (epofolate) (Figure 1) [23], where linker cleavage involves intracellular disulfide reduction followed by intramolecular ester cleavage to produce an N-(2-hydroxyethyl) derivative of aziridine-Epo A as the cytotoxic effector molecule.
In this paper, we report a different approach towards the design of epothilone-based tumortargeted conjugates that is based on the benzimidazole-containing, structurally simplified cyclopropyl-trans-epothilone A analog 1 (Figure 2). We have previously shown that the related epothilone analog 3 ( Figure 2) shows similar antiproliferative activity as Epo B [24]; at the same time, Epothilones have been the subject of extensive SAR studies and at least eight epothilone-type agents have been advanced to clinical trials in humans in cancer [17,18], including the FDA-approved anticancer drug ixabepilone (the lactam analog of Epo B) [19]; in addition, 12,13-deoxy-Epo B (Epo D) has been investigated in Phase I clinical trials for Alzheimer's disease [20]. While development of most of these compounds has been terminated (including the development of Epo D for Alzheimer's disease), an analog termed utidelone (or UTD1) is currently undergoing Phase III clinical studies for breast cancer treatment in combination with capecitabine [21]. (Very surprisingly (and, in fact, irritatingly), several publications on this compound have appeared in the peer-reviewed literature, without its structure being revealed in any of these papers) [22]. Known attempts at the development of tumor-targeted derivatives of epothilones so far have been limited to the highly elaborate folate conjugate BMS-753493 (epofolate) (Figure 1) [23], where linker cleavage involves intracellular disulfide reduction followed by intramolecular ester cleavage to produce an N-(2-hydroxyethyl) derivative of aziridine-Epo A as the cytotoxic effector molecule.
In this paper, we report a different approach towards the design of epothilone-based tumor-targeted conjugates that is based on the benzimidazole-containing, structurally simplified cyclopropyl-trans-epothilone A analog 1 (Figure 2). We have previously shown that the related epothilone analog 3 ( Figure 2) shows similar antiproliferative activity as Epo B [24]; at the same time, the presence of the larger 2-hydroxyethyl substituent on the benzimidazole moiety was found to be well tolerated for a corresponding cyclopropyl-Epo B derivative [25]. Lastly, the replacement of the epoxide moiety in natural epothilones by a cyclopropane ring is expected to lead to enhanced metabolic stability. the presence of the larger 2-hydroxyethyl substituent on the benzimidazole moiety was found to be well tolerated for a corresponding cyclopropyl-Epo B derivative [25]. Lastly, the replacement of the epoxide moiety in natural epothilones by a cyclopropane ring is expected to lead to enhanced metabolic stability. In the actual tumor-targeted conjugate 4, the thiol analog of 1 (i.e. 2) is connected to the side chain of a Cys residue added to the N-terminus of the epidermal growth factor receptor (EGFR)binding peptide GE11 [26]. Reductive disulfide cleavage of this conjugate, thus, would directly lead to the cytotoxic effector molecule without the need for a subsequent immolative step (as in BMS-753493).
The EGFR (HER1, ErbB1) is a 170 kDa transmembrane glycoprotein with an intracellular tyrosine kinase effector domain [27]. Enhanced EGFR signaling is associated with increased proliferation, differentiation, angiogenesis, invasion and metastasis and overexpression or mutation of EGFR is characteristic of a variety of human tumors [28]. Importantly in the context of this work, the EGFR is the targeted cell surface element for one of the marketed ADCs (trastuzumab emtansine). GE11 is a 12-amino acid residue peptide of the sequence YHWYGYTPQNVI, which has been reported to bind to EGFR with high affinity and specificity (Kd 22 nM), to be internalized preferentially into cells with high EGFR expression levels and to accumulate in EGFR-overexpressing tumor xenografts after i.v. administration in vivo [26]. GE11 has been investigated as a tumor-targeting device, e.g., in liposomes [29], nanoparticles [30], or polymeric prodrugs [31], but also for receptor imaging purposes [32].

Chemistry
The synthesis of both epothilone analog 1 as well as GE11-conjuagte 4 proceeded through the protected macrolactone 12 as a key precursor; as outlined in Scheme 1, the latter was obtained from the known olefin 5 [25] as an advanced intermediate. Cross metathesis of 5 (83% ee) with cis-buten-1,4-diol gave allylic alcohol 6 as an inseparable ca. 3:1 mixture of E/Z isomers, which was submitted as such to Charette cyclopropanation [33]. The cyclopropanation reaction furnished a mixture of products 7 that could not be separated. According to 1 H-NMR, two major isomers were present in a ca. 9:1 ratio, but it was not determined if these isomers were distinguished by the configuration of the hydroxyl-bearing stereocenter (as a consequence of the imperfect stereochemical purity of olefin 5) or by the stereochemistry of the cyclopropane moiety. Oxidation of 7 with Dess-Martin periodinane furnished aldehyde 8 as a mixture of three detectable isomers in a 1:0.14:0.03 ratio (based on the aldehyde signal in the 1 H-NMR spectrum), which again proved to be inseparable. Aldehyde 8 was obtained in a 58% overall yield for the three-step sequence from olefin 5.
The elaboration of aldehyde 8 into the epothilone macrocyclic framework in a first step entailed Julia-Kocienski olefination with sulfone 9 (Scheme 1) [34]. The reaction was best carried out under Barbier conditions in the presence of two equivalents of LiHMDS, which furnished the desired olefin in a 72% yield with ca. 2/1 selectivity (based on 1 H-NMR). The low selectivity of the olefination In the actual tumor-targeted conjugate 4, the thiol analog of 1 (i.e., 2) is connected to the side chain of a Cys residue added to the N-terminus of the epidermal growth factor receptor (EGFR)-binding peptide GE11 [26]. Reductive disulfide cleavage of this conjugate, thus, would directly lead to the cytotoxic effector molecule without the need for a subsequent immolative step (as in BMS-753493).
The EGFR (HER1, ErbB1) is a 170 kDa transmembrane glycoprotein with an intracellular tyrosine kinase effector domain [27]. Enhanced EGFR signaling is associated with increased proliferation, differentiation, angiogenesis, invasion and metastasis and overexpression or mutation of EGFR is characteristic of a variety of human tumors [28]. Importantly in the context of this work, the EGFR is the targeted cell surface element for one of the marketed ADCs (trastuzumab emtansine). GE11 is a 12-amino acid residue peptide of the sequence YHWYGYTPQNVI, which has been reported to bind to EGFR with high affinity and specificity (K d 22 nM), to be internalized preferentially into cells with high EGFR expression levels and to accumulate in EGFR-overexpressing tumor xenografts after i.v. administration in vivo [26]. GE11 has been investigated as a tumor-targeting device, e.g., in liposomes [29], nanoparticles [30], or polymeric prodrugs [31], but also for receptor imaging purposes [32].

Chemistry
The synthesis of both epothilone analog 1 as well as GE11-conjuagte 4 proceeded through the protected macrolactone 12 as a key precursor; as outlined in Scheme 1, the latter was obtained from the known olefin 5 [25] as an advanced intermediate. Cross metathesis of 5 (83% ee) with cis-buten-1,4-diol gave allylic alcohol 6 as an inseparable ca. 3:1 mixture of E/Z isomers, which was submitted as such to Charette cyclopropanation [33]. The cyclopropanation reaction furnished a mixture of products 7 that could not be separated. According to 1 H-NMR, two major isomers were present in a ca. 9:1 ratio, but it was not determined if these isomers were distinguished by the configuration of the hydroxyl-bearing stereocenter (as a consequence of the imperfect stereochemical purity of olefin 5) or by the stereochemistry of the cyclopropane moiety. Oxidation of 7 with Dess-Martin periodinane furnished aldehyde 8 as a mixture of three detectable isomers in a 1:0.14:0.03 ratio (based on the aldehyde signal in the 1 H-NMR spectrum), which again proved to be inseparable. Aldehyde 8 was obtained in a 58% overall yield for the three-step sequence from olefin 5.
suggested the presence of at least one minor isomer in both cases, whose content was difficult to quantify, however). Saponification of methyl ester 11 gave a crude acid that was directly submitted to Yamaguchi macrolactonization [37], to give the fully protected macrolactone 12 in 51% yield after purification by preparative RP-HPLC. Ca. 500 mg of this key intermediate were prepared in a purity that was sufficient for subsequent manipulations, thus highlighting the practicability of the process developed. As illustrated in Scheme 2, global deprotection of 12 with HF•pyridine gave trans-cyclopropyl epothilone 1 in 33% yield after purification by preparative RP-HPLC. Alternatively, 12 could be selectively deprotected at the primary hydroxyl group with TASF (tris(dimethylamino)sulfonium difluoromethylsilicate) [38] to give the partially protected macrolactone 13 in quantitative yield. Mitsunobu reaction with thioacetic acid as the nucleophile then gave a thioester derivative from which the O-TIPS (triisopropylsilylether) protecting group was removed with HF•pyridine, to provide thioacetate 14. The latter could be directly converted into the activated mixed disulfide 15 by The elaboration of aldehyde 8 into the epothilone macrocyclic framework in a first step entailed Julia-Kocienski olefination with sulfone 9 (Scheme 1) [34]. The reaction was best carried out under Barbier conditions in the presence of two equivalents of LiHMDS, which furnished the desired olefin in a 72% yield with ca. 2/1 selectivity (based on 1 H-NMR). The low selectivity of the olefination reaction was inconsequential, as the double bond was reduced in the next step with diimide to provide the fully protected seco ester 10 in quantitative yield. Either 2,4,6-triisopropylbenzenesulfonylhydrazide (TPSH) [35] or o-nitrobenzenesulfonylhydrazide (NBSH, as a cheaper alternative) [36] could be employed as a diimide source, with both methods delivering 10 in excellent yields. Selective cleavage of the benzylic silyl-ether with CSA in DCM/MeOH 1:1 then furnished alcohol 11 in quantitative yield. (Careful inspection of the NMR spectra of 10 and 11 suggested the presence of at least one minor isomer in both cases, whose content was difficult to quantify, however). Saponification of methyl ester 11 gave a crude acid that was directly submitted to Yamaguchi macrolactonization [37], to give the fully protected macrolactone 12 in 51% yield after purification by preparative RP-HPLC. Ca. 500 mg of this key intermediate were prepared in a purity that was sufficient for subsequent manipulations, thus highlighting the practicability of the process developed.
As illustrated in Scheme 2, global deprotection of 12 with HF·pyridine gave trans-cyclopropyl epothilone 1 in 33% yield after purification by preparative RP-HPLC. Alternatively, 12 could be selectively deprotected at the primary hydroxyl group with TASF (tris(dimethylamino)sulfonium difluoromethylsilicate) [38] to give the partially protected macrolactone 13 in quantitative yield. Mitsunobu reaction with thioacetic acid as the nucleophile then gave a thioester derivative from which the O-TIPS (triisopropylsilylether) protecting group was removed with HF·pyridine, to provide thioacetate 14. The latter could be directly converted into the activated mixed disulfide 15 by reaction with 2,2 -dipyridyl disulfide under slightly basic conditions. Reaction of 15 with CysGE11 then provided the desired epothilone-GE11 conjugate 4. While the thioester moiety of 14 could be readily cleaved with K2CO3/MeOH, the resulting free thiol 2 could not be isolated as a pure material. Disulfide formation was already observed between elution of the material from the HPLC column and lyophilization of the sample. Only disulfide 16 could be isolated and characterized.

Biological Assessment
In order to provide a rational framework for the biological assessment of the epothilone peptide conjugate 4, EGFR was quantified in cell lines that were to be used for those experiments, i.e., A431 epidermoid squamous cell carcinoma cells, SW480 colorectal adenocarcinoma cells, and HEK293 embryonic kidney cells. A431 cells have been reported to highly overexpress EGFR [39]; consistent overexpression of EGFR has also been described for SW480 cells [40]. In contrast, HEK293 cells were While the thioester moiety of 14 could be readily cleaved with K 2 CO 3 /MeOH, the resulting free thiol 2 could not be isolated as a pure material. Disulfide formation was already observed between elution of the material from the HPLC column and lyophilization of the sample. Only disulfide 16 could be isolated and characterized.

Biological Assessment
In order to provide a rational framework for the biological assessment of the epothilone peptide conjugate 4, EGFR was quantified in cell lines that were to be used for those experiments, i.e., A431 epidermoid squamous cell carcinoma cells, SW480 colorectal adenocarcinoma cells, and HEK293 embryonic kidney cells. A431 cells have been reported to highly overexpress EGFR [39]; consistent overexpression of EGFR has also been described for SW480 cells [40]. In contrast, HEK293 cells were previously found not to express EGFR [39]. Expression levels of EGFR were assessed by the treatment of cells with cetuximab followed by FACS-based quantification of cetuximab with an Alexa Fluor 647 goat anti-human antibody. In line with the existing literature, EGFR expression levels were found to be highest in A431 cells, followed by SW480 cells. HEK293 cells gave fluorescence intensities at the level of the isotype control ( Figure S1). Therefore, 4 would have been expected to display strongest cytotoxicity against A431 and SW480 cells, while HEK293 cells should have been largely insensitive to the action of the conjugate. As will be discussed below, the experimental results did not conform to these predictions.
The stability of the epothilone-GE11 conjugate 4 in complete cell culture medium was evaluated under conditions resembling those encountered in the cytotoxicity assays. Thus, an 8.6 µM solution of 4 in RPMI with 10% FBS was incubated for 3 days at 37 • C and subsequently analyzed by RP-HPLC. Only minor degradation of 4 (<10%) was observed under these conditions (data not shown).
The reductive cleavage of conjugate 4 was analyzed under conditions mimicking those found in the endosome and the cytoplasm of cancer cells [41]. Endosome-like conditions were assumed to be reproduced by a 10 mM solution of glutathione (GSH) in acetate and phosphate buffer pH 4.9; cytoplasm-like conditions entailed 10 mM GSH in phosphate buffer pH 7.4. Under both conditions, rapid disappearance of 4 was observed with apparent first order kinetics ( Figure 3). goat anti-human antibody. In line with the existing literature, EGFR expression levels were found to be highest in A431 cells, followed by SW480 cells. HEK293 cells gave fluorescence intensities at the level of the isotype control ( Figure S1). Therefore, 4 would have been expected to display strongest cytotoxicity against A431 and SW480 cells, while HEK293 cells should have been largely insensitive to the action of the conjugate. As will be discussed below, the experimental results did not conform to these predictions. The stability of the epothilone-GE11 conjugate 4 in complete cell culture medium was evaluated under conditions resembling those encountered in the cytotoxicity assays. Thus, an 8.6 μM solution of 4 in RPMI with 10% FBS was incubated for 3 days at 37 °C and subsequently analyzed by RP-HPLC. Only minor degradation of 4 (<10%) was observed under these conditions (data not shown).
The reductive cleavage of conjugate 4 was analyzed under conditions mimicking those found in the endosome and the cytoplasm of cancer cells [41]. Endosome-like conditions were assumed to be reproduced by a 10 mM solution of glutathione (GSH) in acetate and phosphate buffer pH 4.9; cytoplasm-like conditions entailed 10 mM GSH in phosphate buffer pH 7.4. Under both conditions, rapid disappearance of 4 was observed with apparent first order kinetics ( Figure 3). The half-life of 4 was 1.1 min under endosome-like conditions, while no residual conjugate was detectable even 1 min after addition to the glutathione solution under cytoplasm-like conditions (estimated half-life of ca. 0.2 min). LC/MS analysis of cleavage solutions showed that the products obtained at both pH values included the mono-and dimeric CysGE11 peptide, thiol-containing epothilone analog 2, the mixed disulfide of 2 with glutathione, and dimer 16. The mixed disulfide of 2 with glutathione appeared very early in the reaction ( Figure S2); it was later transformed into 2 and 16. The shorter half-life of 4 under cytoplasm-like conditions is in line with the fact that disulfide reduction is pH-dependent and has been shown to be significantly slower at lower pH [41].
The binding affinity of epothilone analogs 1, 14, and 16 and of targeted conjugate 4 for microtubules was determined as previously described by Díaz and co-workers, using partly crosslinked microtubules and determining the changes in fluorescence anisotropy upon displacement of the fluorescent taxol derivative Flutax-2 [42,43] (Figure S3). Not unexpectedly, the microtubule- The half-life of 4 was 1.1 min under endosome-like conditions, while no residual conjugate was detectable even 1 min after addition to the glutathione solution under cytoplasm-like conditions (estimated half-life of ca. 0.2 min). LC/MS analysis of cleavage solutions showed that the products obtained at both pH values included the mono-and dimeric CysGE11 peptide, thiol-containing epothilone analog 2, the mixed disulfide of 2 with glutathione, and dimer 16. The mixed disulfide of 2 with glutathione appeared very early in the reaction ( Figure S2); it was later transformed into 2 and 16. The shorter half-life of 4 under cytoplasm-like conditions is in line with the fact that disulfide reduction is pH-dependent and has been shown to be significantly slower at lower pH [41].
The binding affinity of epothilone analogs 1, 14, and 16 and of targeted conjugate 4 for microtubules was determined as previously described by Díaz and co-workers, using partly cross-linked microtubules and determining the changes in fluorescence anisotropy upon displacement of the fluorescent taxol derivative Flutax-2 [42,43] (Figure S3). Not unexpectedly, the microtubule-binding affinity of the simplified epothilone analog 1 even exceeds that of natural Epo A (K b of 7.85 × 10 7 M −1 vs. 3.17 × 10 7 M −1 for Epo A; Table 1); this observation is in line with existing SAR data for other epothilone analogs that are derived from trans-Epo A and/or incorporate a benzimidazole side chain [24,25,44], although analog 1 has not been investigated previously. Compared to 1, the K b for thioacetate 14 is considerably lower (>10-fold), but its microtubule-binding affinity is still comparable to that of other epothilone analogs that have shown nanomolar antiproliferative activity against different cancer cell lines [45]. Intriguingly, analog 16 and conjugate 4 exhibit similar microtubule-binding affinity as thioacetate 14, in spite of the significant enlargement in size of the substituent group on the benzimidazole side chain. Importantly, the targeting peptide CysCE11 showed no relevant microtubule binding. The antiproliferative activity of the targeted epothilone conjugate 4 was assessed for A431, SW480, and HEK293 cells, in comparison with epothilone analogs 1 and 16. Of the three compounds investigated, conjugate 4 proved to be the least potent by one order of magnitude (Table 2). While the lower activity of 4 compared to 1 against A431 and SW480 cells could potentially be rationalized by a lower rate of receptor-mediated uptake vs. uptake by passive diffusion (which is likely to be operative for 1) and/or the efficiency of formation of 2 from 4 in, and its release from, the endosomal compartment, the lack of selectivity between EGFR-expressing A431 and SW480 cells and essentially EGFR-free HEK293 cells is difficult to explain. This finding may point to cleavage of the disulfide bond in the extracellular space, and the passive diffusion of 2 into the cells (assuming that passive diffusion of conjugate 4 through the cell membrane followed by intracellular disulfide reduction is essentially excluded). Should this indeed be the case, the same mechanism may also underlie the activity of 4 against A431 and SW480 cells, without the involvement of EGFR-mediated endocytosis. In this context, it should be remembered that 10% disulfide cleavage or less would already account for the observed IC 50 values, if one assumes the (elusive) epothilone 2 to exhibit similar activity as 1. Extracellular disulfide cleavage as the source of the antiproliferative activity of 4 would thus still be compatible with the stability data obtained for the compound in cell culture medium (vide supra). Preliminary experiments aiming at the quantification of intracellular concentrations of 4, 2, or 16 did not produce any interpretable results.
In a broader context, we note that recycling endosomes, late endosomes, and lysosomes, in contrast to our own basic assumption, had been suggested to provide a reducing, rather than oxidizing environment, even prior to the start of our work on 4 [46]. At the same time, careful inspection of the literature suggests that, while the GE11 peptide does seem to allow selective targeting of EGFR-overexpressing cells by nanoparticles [30] or polymeric prodrugs [31], the selectivity that has been achieved is rather moderate. In line with these observations, the reported K d of 22 nM of GE11 for the EGFR could not be reproduced in subsequent studies by other laboratories. Thus, Levitzki and co-workers determined the IC 50 for the displacement of [ 125 I]-EGF from the EGFR by GE11 to be >1 mM (!) (vs. 5.1 nM for EGF) [47]; more recently, Lin and co-workers determined the K d of GE11 for the EGFR as 459 µM by means of surface plasmon resonance [48]. Similar findings have been reported by Sihver and co-workers in the context of studies on potential GE11-derived PET imaging agents [49]. Thus, in retrospect, neither the choice of the GE11 peptide as an EGFR-targeting moiety nor its combination with a disulfide linker may have been optimal.
The antiproliferative activity of disulfide 16 is similar to that of analog 1, in spite of its >10-fold lower microtubule-binding affinity. While microtubule-binding affinity is not necessarily linearly correlated with cell growth inhibition, this finding may point to thiol-containing epothilone analog 2 as the species mainly responsible for the antiproliferative effects of 16. It is easily conceivable for disulfide 16 to readily cross the cellular membrane by passive diffusion (the molecular weight of 16 is 1054 vs., e.g., 856 for taxol, which easily enters cells); immediate reduction in the cytoplasmic compartment would then produce 2. However, we do not have any experimental data at this point to support this (plausible) mechanistic hypothesis.

Chemistry
All non-aqueous reactions were carried out under anhydrous conditions and under an argon atmosphere unless otherwise noted. CH 2 Cl 2 was distilled from CaH; THF, Et 2 O, benzene, and toluene were distilled from Na/benzophenone. All other absolute solvents were purchased from Fluka (absolute over molecular sieves). Commercial chemicals were used without further purification, unless otherwise noted.
Solvents for extractions, column chromatography, and thin layer chromatography (TLC) were purchased as commercial grade and distilled prior to use. TLC was performed on Merck TLC aluminum sheets (silica gel 60, F254). Spots were visualized with UV light (254 nm) or through staining with an aqueous solution of phosphomolybdic acid, cerium sulfate, and sulfuric acid (CPS). Chromatographic purification of products was performed by flash chromatography (FC) using Fluka silica gel 60 for preparative column chromatography (particle size 40-63 µm). Organic solutions were concentrated by rotary evaporation at 40 • C and approximately 20 mbar. The compounds were further dried under high vacuum (0.01-0.001 mbar). Yields refer to compounds isolated after FC, unless otherwise specified. NMR spectra were recorded on a Bruker AV-400 400 MHz or a Bruker AV-500 500 MHz NMR spectrometer (Bruker Biospin AG, Fällanden, Switzerland) at 298 K. NMR spectra are referenced relative to the residual hydrogen signal of the deuterated solvent ( 1 H 7.26 ppm, 13 C 77.0 ppm for CDCl 3 ). All 13 C spectra were measured with complete proton decoupling. For inseparable diastereomeric mixtures, only signals of the major diastereomer are reported. Spin multiplicities are reported as follows: s = singlet, d = doublet, t = triplet, q = quartet, quint = quintet, sext = sextet, m = multiplet, mc = centered multiplet, br = broad signal; J = coupling constant in Hz. Infrared spectra (IR) were recorded on a Jasco FT/IR-6200 instrument. Positions of absorption bands are given in wavenumbers [cm −1 ]. Optical rotations were measured on a Jasco P-1020 polarimeter and are reported as follows: [α] 24 D : concentration (g/100 mL) and solvent. High resolution mass spectra (HRMS) were recorded by the ETH Zürich MS service on a Varian IonSpec Ultima (ESI) or a Waters Micromass Autospec Ultima spectrometer (EI). RP-HPLC was carried out on a Merck Hitachi device (column oven L-2350, diode array detector L-2450, autosampler L-2200, pump L-2130) using a Waters Symmetry C18 column, 3.5 µm, 4.6 × 100 mm, at a flow rate of 1 mL/min for analytical purposes; a Waters Symmetry C18 column, 5 µm, 7.8 × 100 mm, at a flow rate of 2 mL/min was employed for semi-preparative applications. The column temperature for analytical and semi-preparative applications was 40 • C. Preparative RP-HPLC was carried out on a Gilson device (Gilson (Schweiz) AG, Mettmenstetten, Switzerland) with a Waters Symmetry C18 column (Waters AG, Baden-Dättwil, Switzerland), 5 µm, 19 × 100 mm, at a flow rate of 25 mL/min at room temperature. Peaks were detected at 280, 254, or 220 nm. The Cys-GE11 peptide was custom-made by Biosyntan GmbH (Berlin, Germany).

Glutathione Cleavage Assay
For cytoplasm-like conditions, 100 µL of a 173 µM solution of conjugate 4 in degassed 100 mM phosphate buffer pH 7.4 were diluted with 400 µL of a degassed phosphate buffer and 500 µL of 20 mM glutathione in 100 mM phosphate buffer pH 7.4. Aliquots of 80 µL were removed at predefined time points, and the reaction was quenched with one volume of 10% metaphosphoric acid, giving a pH of 1 approximately. Fifty microliters of this solution were analyzed by analytical RP-HPLC. The same protocol was used for cleavage under endosome-like conditions, except that the glutathione solution was prepared in a 100 mM acetate buffer at pH 4.8, and the same buffer was also used to dilute the conjugate solution in a 100 mM phosphate buffer at pH 7.4. The final pH of the cleavage solution was 4.9.
Analytical RP-HPLC: H 2 O with 0.1% TFA (A)/acetonitrile/water 8/2 with 0.05% TFA (B). Linear gradient from 5% B to 80% B over 30 min. The percentage of remaining conjugate in the solutions was plotted against time ( Figure 3). Analysis of the cleavage solutions by LC/MS provided information of the fractions of the various components in the redox mixture. As an example, Figure S2 shows the product distribution after 5 min of reaction time under endosome-like conditions.

Determination of Microtubule Binding Constants
Purified calf-brain tubulin and chemicals were obtained as previously described [50]. Stabilized, moderately crosslinked microtubules were prepared as reported in [42]. Binding constants of azathilones to stabilized microtubules were measured as previously described by Buey et al. [41]. For details of the experimental procedures and Flutax-2 displacement curves cf. Section SI.3 of the Supplementary Materials.
The final DMSO (dimethylsulfoxide) concentration was 0.1%. After 72 h, a WST-1 reagent [43] was added and the absorption was measured at 450 nm with a plate reader after 30 min to 4 h. The percentage of viable cells was calculated based on vehicle-treated cells and was plotted in GraphPadPrism. At least three independent experiments were performed. A431 cells were obtained from CLS. The other cell lines were available in the laboratory. WST-1 was purchased from Roche.
For inhibition curves cf. Section SI.5 of the Supplementary Materials.

Conclusions
We have successfully prepared a disulfide-linked conjugate between a novel epothilone analog and the (purported) EGFR-binding peptide GE11, and we have determined its microtubule-binding affinity and its in vitro antiproliferative activity against EGFR-overexpressing cells and cells being devoid of EGFR. While the conjugate had been designed to specifically target EGFR-overexpressing cells, its in vitro antiproliferative activity was found to be independent of EGFR expression status, at least for the limited number of cell lines evaluated. While our study has been limited in scope, in combination with other data in the literature, our findings could suggest that GE11 does not represent an effective targeting moiety for EGFR-overexpressing tumor cells, and even less so in combination with a disulfide linker (contrary to our original working hypothesis).
Independent of the uncertainties surrounding the activity of the specific epothilone conjugate 4 that we have investigated in this study, our work does demonstrate that the new epothilone analog 1 can serve as highly active template for the construction of targeted prodrugs. The compound exhibits nM antiproliferative activity against a variety of cell lines and offers a readily accessible, sterically unencumbered primary hydroxyl group for chemical manipulation; the requisite partially protected precursor for these manipulations (i.e., intermediate 13) can be prepared on a scale that allows the production of reasonable amounts of prodrugs for subsequent biochemical, cell biological, and pharmacological studies. The preparation of such prodrugs will be the subject of future studies in our laboratory.

Funding:
We are indebted to the Swiss National Science Foundation (F.G.Z.; project number 205320-117594) and the ETH Zürich for generous financial support. This work was also supported by Ministerio de Economia y Competitividad grant BFU2016-75319-R to FDP (AEI/FEDER, UE). We thank Ganadería Fernando Díaz for the supply of calf brains. The authors acknowledge networking contributions by the COST Action CM1407 "Challenging organic syntheses inspired by nature -from natural products chemistry to drug discovery" and the COST action CM1470.