Glycoconjugated Metallohelices have Improved Nuclear Delivery and Suppress Tumour Growth In Vivo

Abstract Monosaccharides are added to the hydrophilic face of a self‐assembled asymmetric FeII metallohelix, using CuAAC chemistry. The sixteen resulting architectures are water‐stable and optically pure, and exhibit improved antiproliferative selectivity against colon cancer cells (HCT116 p53+/+) with respect to the non‐cancerous ARPE‐19 cell line. While the most selective compound is a glucose‐appended enantiomer, its cellular entry is not mainly glucose transporter‐mediated. Glucose conjugation nevertheless increases nuclear delivery ca 2.5‐fold, and a non‐destructive interaction with DNA is indicated. Addition of the glucose units affects the binding orientation of the metallohelix to naked DNA, but does not substantially alter the overall affinity. In a mouse model, the glucose conjugated compound was far better tolerated, and tumour growth delays for the parent compound (2.6 d) were improved to 4.3 d; performance as good as cisplatin but with the advantage of no weight loss in the subjects.


Introduction
We have developed several structurally distinct ranges of metallohelices comprising three organic ligands that encapsulate two metal ions, [1] such as that shown in Scheme 1a. Unlike conventional helicates, [2] these water-stable Fe II compounds self-assemble as optically pure architectures,p rincipally aresult of inter-ligand steric and secondary interactions including hydrophobic p-stacks. [3] There is mounting evidence that as ar esult of their charge,s hape,s ize and amphipathic structures,t hese compounds emulate some of the functional properties of short cationic a-helical peptides.O riented binding to various nucleic acid structures is observed. [1a, 4] One class [1b] inhibits ice recrystallization apparently as aresult of the facially amphipathic architecture that is also present in natural antifreeze peptides. [5] Asimilar structure binds to the central hydrophobic a-helical region of an amyloid b protein and attenuates toxicity. [6] Perhaps most convincingly,w e showed recently that ac lass of antimicrobial metallohelix in our library [1e] rapidly penetrates the formidable cell envelope of ac linically-relevant Gram negative microbe and causes apeptide-like genomic and transcriptomic response.
Cell-penetrating peptides (CPPs) are usually relatively short (5-50 residues) [7] and contain an excess of cationic amino acids (lysine and arginine). [8] It is proposed that they pass through the plasma membrane via an ion exchange mechanism [9] using negatively charged species such as anionic lipids and glycosaminoglycans.Since these components are in excess in cancer cell and microbial outer-leaflets, [10] ag eneralized source of selectivity over other cells is provided. Nevertheless,such polycationic molecules may also have nonspecific affinity for an umber of biomolecular structures [7,11] and the modification of CPPs with biocompatible fragments has been used in an attempt to modulate the attendant toxicity. [11b,12] In particular, glycoconjugation has been used extensively for the modification of potential therapeutics of an umber of kinds. [13] In nature,g lycosylation is one of the most common post-translational modifications [14] and glycopeptides are involved in cell signalling, [15] providing cell surface markers for recognition, and immune response. [16] From ad rug-design perspective,m onosaccharide-conjugated analogues have been reported in the literature since the early 1990s, [17] improving the water solubility and serum stability of their cargo, [17b] as well as altering drug metabolism and pharmacokinetics (DMPK) [18] including some literature precedent for exploiting the Warburg effect in cancer therapy. [19] Several groups have also shown that glycosylation of apeptide increases membrane penetration, including through the blood-brain barrier. [20] In recent work, Montenegro and co-workers developed astrategy for the glycosylation of short peptides,a nd have systematically characterized the uptake efficiency and distribution in various cell lines. [21] Our recent success in CuAACd erivatization of metallohelices using relatively simple functionality, [1d] and an in cellulo click staining protocol, [1e] gave us confidence to attempt the rather more ambitious glycoconjugations.W e report here that this chemistry,giving rise to some of the most complex functionalized metallosupramolecular structures known, proceeds smoothly and efficiently,l eading to improved cancer-cell targeting in vitro,and improved efficacyin vivo.

Selection of metallohelix system
Thep osition of hydrophobic regions within ap eptide is conventionally assessed by as imple residue-based approach, but this is not applicable here.Instead, analysis [5] ( Figure S1 in the Supporting Information) of the position of counter-anions in the solid state molecular structure reveals af avorable charge distribution for one of our so-called triplex [1b] architectures (Scheme 1) in that the two p-stacked arene rings, colorized pink in Figure 1, shield the cationic charge,leading to the creation of ar elatively hydrophobic upper ridge.A third p-stack is hidden at the rear of this view.T he yellow colorized atoms correspond to the positions of groups Ri n Scheme 1; they will surround arelatively hydrophilic face and hence by adding sugar units at these latter positions we retain the amphipathic architecture.T his,w ec onsidered, was the approach most likely to allow retention of the kinds of biological activity we have seen from the core structure,while allowing us to test the idea that glycosylation may lead to improvements in delivery,selectivity and tolerance.

Synthesis and characterization
Thestarting materials for new synthesis were assembled: the previously-reported [1d] enantiomerically pure triplex metallohelix [Fe 2 L 2 3 ]Cl 4 with alkynyl groups at the positions colorized yellow in Figure 1w as prepared on am ulti-g scale via aone-pot highly diastereoselective self-assembly reaction; the range of monosaccharide azides of Scheme 1b,i ncluding acylateda nalogues,w ere synthesized by literature procedures. [22] Thes ubsequent CuAACg lycosylation was not initially straightforward. Thec onventional copper sulfate/sodium ascorbate catalyst led to difficulties in isolation in this rather polar system, while the heterogeneous catalyst copper-incharcoal [23] failed to complete the reaction. We considered the copper free click reaction [24] but the requirement for cyclooctyne groups would significantly increase the synthetic challenge and restrict versatility.E ventually we found to our surprise that that while copper(I) iodide catalyst required elevated temperatures,this was not deleterious,the reactions were complete,a nd the work-up was trivial. This gave us access to the glycoconjugated triplex metallohelices [Fe 2 L 3ag 3 ]Cl 4 as optically pure isolated compounds.
Thes uccess of this post-assembly CuAACi sa pparent from the 1 H-NMR spectra (Figure 2A and B; for all spectra see Figure S2-S9). Fore xample,t he singlets H j at ca 3ppm corresponding to the three inequivalent alkyne units in the starting material are cleanly replaced by three new singlets at 8.06, 8.17, and 8.28 ppm (H m )f or the triazole rings in the product. It is also noteworthy that the two bipyridine protons involved in inter-strand hydrogen bonds,a nd thus giving rather low field resonances (ca 9.2 ppm), are present in both starting material and product, confirming that the asymmetric triplex architecture is unperturbed by the presence of the sugars.H igh resolution electrospray mass spectra were readily obtained;F igure 2C shows the expected tetracationic molecular ion pattern for S c ,L Fe -HHT-[Fe 2 L 3a 3 ]Cl 4 .T he circular dichroism (CD) spectra of the diastereoisomers ( [1d] Theglycoconjugated compounds were found to be extraordinarily stable under aqueous conditions;nodecomposition was observed on monitoring the absorption at the MLCT band in aqueous solution over many months,a nd even when dissolved in KCl/HCl buffer at pH 1.5 (at 8mm)n od ecomposition was observed over one month ( Figure S18).

Antiproliferative activity and cell studies
Thew hole panel of Fe II compounds of Scheme 1w ere evaluated alongside cisplatin for potencya gainst the human colorectal cancer cells with wild-type p53 (HCT116 p53 +/+ ) and non-cancerous human epithelial retinal pigment cells (ARPE-19) ( Figure 3).
We observe that the sugar-appended triplex systems all inhibit HCT116 p53 +/+ cell proliferation in the 2-30 mm concentration range (96 hI C 50 ), and for all examples the Ldiastereoisomers are more potent than D.T he selectivity indices (SI, defined as IC 50 [ARPE19]/IC 50 [HCT116 p53 +/+ ]) vary from 1.4 to 17, with greater selectivity observed most often with the D-diastereoisomers.W ith SI of 17, D-[Fe 2 L 3a 3 ]Cl 4 is the most selective compound in the panel for this pair of cells.S ince this indicates ap otential therapeutic window,w ec hose to focus on this compound for more detailed study.
We compared the antiproliferative activity of D-[Fe 2 L 3a 3 ]Cl 4 in both glucose-rich and glucose-free media and observed no difference in IC 50 (Table S2). We further incubated the drug with GLUT-1 overexpressing MCF-7 breast cancer cells and compared the IC 50 with wild type MCF-7 cells and found that rather than being more sensitive to the glucose derivative,t he GLUT-1 overexpressing cells are actually ca three-fold more resistant (Table S3). Firstly, this suggests that the cellular entry of these compounds is not (or not mainly) GLUT-mediated;g iven the specificity of binding of this receptor this is perhaps unsurprising,b ut the addition of glucose units to large molecules has been nevertheless been described as acancer cell-targeting strategy. [17b, 25] Secondly,w en ote that the resistance we observed may be beneficial in that normal cells that have high GLUT-1e xpression (e.g.r ed blood cells) will be less adversely affected.
Thec onjugation of sugars with therapeutic peptides and other drug candidates can alter pharmacokinetic properties, and has been demonstrated to improve physiological properties and bioavailability, [26] such as enhancing biodistribution in tissues, [27] improving membrane penetration [28] and targeted delivery. [29] We therefore firstly compared the effects of D-

Research Articles
fractions:( i) cytoskeletal fraction (total cellular insoluble proteins) plus genomic DNA), (ii)n uclear fraction (nuclear soluble proteins,i ncluding nuclear membrane proteins), (iii) membrane fraction(organelles and organelle membrane proteins,b ut excluding nuclear membrane proteins), and (iv) cytosolic fraction (total cytoplasmic soluble proteins). The cells were grown and treated as above and Fe content was again determined by ICP-MS.A ss hown in Figure 5B,t he localization of D-[Fe 2 L 3a 3 ]Cl 4 in the nuclear fraction (13.6 %) was more pronounced in comparison with D-[Fe 2 L 1 3 ]Cl 4 (4.4 %), and was consistent with the data observed in Figure 5A.B oth D-[Fe 2 L 1 3 ]Cl 4 and D-[Fe 2 L 3a 3 ]Cl 4 distribute most predominantly in the membrane fraction at 16 h(62.0 % and 55.7 %r espectively), whereas the localization of D-[Fe 2 L 1 3 ]Cl 4 (25.3 %) in the cytoskeleton fraction is more significant than for D-[Fe 2 L 3a 3 ]Cl 4 (19.6 %). There are several reports of glycosylation-dependent nuclear import of proteins and plasmids, [30] which could be related to the cytosolonuclear lectins shuttling between the cytosol and the nucleus. [30c] Single-cell gel electrophoresis studies (Comet Assay) in HCT116 p53 +/+ cells treated with D-[Fe 2 L 3a 3 ]Cl 4 revealed an absence of single-or double-strand DNAb reaks due to the lack of a"comet" tail ( Figure 5C). In addition, D-[Fe 2 L 3a 3 ]Cl 4 does not retard the formation of the "comets" in cells treated with DNAd amaging peroxide,i ndicating that it does not form DNAc ross-links.T he parent compound D-[Fe 2 L 1 3 ]Cl 4 behaves similarly. [1b] Thus if these metallohelices interact with DNAi nt he nucleus,t hey do not cause irreversible changes leading to cell death, as does cisplatin. [31] Notwithstanding these findings,w ec ompared the antiproliferative activity of these complexes in the pair of Chinese Hamster Ovary Cell lines CHO-K1 and MMC-2 (Table 1); as ystem previously used to identify the DNAd amage involvement of cytotoxic agents.M MC-2 is aC HO-K1 mutant carrying the ERCC3/XPB mutation, which renders this cell line deficient in DNAn ucleotide excision repair (NER). [32] Thef actor F ( Table 1), which compares IC 50 for Chinese Hamster Ovary cells (wild type) and the NER deficient system, is rather lower for D-[Fe 2 L 1 3 ]Cl 4 and D-[Fe 2 L 3a 3 ]Cl 4 than it is for the DNAd amaging agent cisplatin, but there is athree or six-fold difference between the response of the two cell lines;this prompted us to study DNAinteractions in vitro (below). We further compared the antiproliferative activity of D-[Fe 2 L 1 3 ]Cl 4 and D-[Fe 2 L 3a 3 ]Cl 4 against A2780 ovarian cancer cells,a nd the cisplatin-resistant strain A2780cisR (Table 2). No cross-resistance with cisplatin was detected. We also compared the response of p53-deficient and wild type HCT116 cells.Whilst p53-deficient cells were less responsive to cisplatin, there was no significant difference between the response of HCT116 p53 +/+ and p53 À/À cells (p > 0. 05   effects on the cells via ad ifferent mechanism to cisplatin, whilst indicating an on-destructive interaction with DNA, more so for D-[Fe 2 L 3a 3 ]Cl 4 .

Biophysical studies in vitro
Given the above observations,weinvestigated the in vitro DNA-binding of D-[Fe 2 L 1 3 ]Cl 4 and D-[Fe 2 L 3a 3 ]Cl 4 via af luorescence competition assay. [33] Theb ehavior was very similar for both compounds ( Figure S19 Thus DNA-binding affinity is not responsible for the higher accumulation of D-[Fe 2 L 3a 3 ]Cl 4 in the nucleus. Further,l inear dichroism (LD) studies indicate that the complexes bind to naked calf thymus DNAi nas pecific orientation, probably the major groove. [1a] These results, alongside the negative comet assays suggest that the DNA interactions are non-covalent, and probably reversible,a kin to those of peptide a-helices and zinc fingers. [34,35] In vivo studies Based on their potencya nd selectivity, L-[Fe 2 L 1 3 ]Cl 4 and D-[Fe 2 L 3a 3 ]Cl 4 were selected for initial in vivo evaluation. They were administered as asingle intravenous (IV) injection in HCT116 p53 À/À bearing athymic nude mice.P rior to these studies,the maximum tolerated dose (MTD) was determined for both compounds;t he glucose-appended metallohelix D-   clinical drug agent cisplatin (Table 3, Figure 6). Importantly, no weight loss effects were observed following treatment with L-[Fe 2 L 1 3 ]Cl 4 or of D-[Fe 2 L 3a 3 ]Cl 4 ,w hereas cisplatin induced as howed 6% loss of body weight in the first day following injection.

Conclusion
We have developed av ery efficient method for the conjugation of triplex metallohelices with sugar units.T he highly complex products have amphipathic structures,a re optically pure,w ater-soluble,a nd extremely stable in water and biological media.
Theaddition of the carbohydrate units leads to substantial changes in the antiproliferative activity.M ost strikingly,f or the D-configured (right-handed helix) compounds,the apparent selectivity for cancer cells is greatly increased. In amouse model, the drug tolerance and effect, as measured by MTD and tumour growth delay,a re substantially improved versus the parent system. Encouragingly,n ow eight loss was recorded in the subjects following the dose.  ]Cl 4 . Mice were administrated with asingle dose on day 0b yintravenous injection. Mean relative tumour volumes (A) and Mean relative bodyweight B) were measured at different time points, plotted, and expressed with AE standard error;the significance p value < 0.01 was considered to be statistically significant (n = 8).
Thetriplex metallohelix system is also shown to be arare example of aclass of DNA-binding/aligning metallohelix. The parent and glycosylated compounds bind and align with DNA with very similar strength, thus validating our structural strategy of appending these polar units to the hydrophilic face of the helix, leaving the relatively hydrophobic ridge unperturbed.
In mechanistic terms,t he addition of the glucose units leads to drug-like dose-dependent cell cycle effects,a nd the response observed in the cell cycle differs significantly between diastereoisomers of the metallohelices.F urther, while the glucose derivative was found to be the most selective for the chosen cancer cell system, we conclude that this is not due to GLUT receptor targeting. Indeed, the cellular uptake is actually attenuated by addition of the sugars.Interestingly however, intranuclear transport is overall increased, perhaps by as ugar-mediated process. [30c] Notably, the intranuclear transport, and the presumed DNAb inding events in cellulo,d onot lead to DNAdamage.
Overall it would appear that the modification of triplex metallohelices in this way is worthy of investigation as as trategy for improvement of targeting and efficacyi nt his system, just as it is for the natural a-helical systems.Also,we can add this behaviour to agrowing list of evidences that this class of molecule,w ith its many variants,s hare features with cationic antimicrobial and anticancer peptides.
In vivo evaluation was performed under contract at the Institute of Cancer Therapeutics UK under Home Office licence PPL 40/3670. Local ethical approval was obtained on 07 April 2016 by the Animal Welfare and Ethical Review Body (AWERB) of the University of Warwick (reference AW ERB.26/15-16).