Multivariate Modulation of the Zr MOF UiO‐66 for Defect‐Controlled Combination Anticancer Drug Delivery

Abstract Metal–organic frameworks (MOFs) are emerging as leading candidates for nanoscale drug delivery, as a consequence of their high drug capacities, ease of functionality, and the ability to carefully engineer key physical properties. Despite many anticancer treatment regimens consisting of a cocktail of different drugs, examples of delivery of multiple drugs from one MOF are rare, potentially hampered by difficulties in postsynthetic loading of more than one cargo molecule. Herein, we report a new strategy, multivariate modulation, which allows incorporation of up to three drugs in the Zr MOF UiO‐66 by defect‐loading. The drugs are added to one‐pot solvothermal synthesis and are distributed throughout the MOF at defect sites by coordination to the metal clusters. This tight binding comes with retention of crystallinity and porosity, allowing a fourth drug to be postsynthetically loaded into the MOFs to yield nanoparticles loaded with cocktails of drugs that show enhancements in selective anticancer cytotoxicity against MCF‐7 breast cancer cells in vitro. We believe that multivariate modulation is a significant advance in the application of MOFs in biomedicine, and anticipate the protocol will also be adopted in other areas of MOF chemistry, to easily produce defective MOFs with arrays of highly functionalised pores for potential application in gas separations and catalysis.


Introduction
Metal-organic frameworks (MOFs) [1] are an ew generation of highly porous macromolecular structures composed of metal ions or clusters linked by multidentate organic bridging ligands,w hich, owing to their attractive properties,h ave notable potential for applications in different contexts, including gas capture,s torage and separation, [2] catalysis, [3] water treatment [4] and drug delivery. [5] MOFs have almost infinite tunability due to the effectively unlimited range of metal ions and ligands available to form their structures. [6] Additionally,aclass of mixed-linker materials-so-called multivariate (MTV) MOFs-have been synthesised, tuning physicochemical properties and their performance in different applications.B yi ntroducing multiple analogous linkers into one framework and effectively forming asolid solution of organic functionality within the pores (Figure 1a), cooperative effects for gas adsorption and heterogeneous catalysis have been found, [7] fore xample,Y aghi et al. synthesised MTV-MOF-5 type structures that contain up to eight distinct functionalities (by introducing eight functionalised terephthalate linkers during synthesis) in one phase,resulting,inthe case of aM OF linked by three different terephthalate derivatives,i nu pt o4 00 %b etter CO 2 selectivity (vs.C O) than their separate MOF-5 counterparts. [7c] MOFs have been proposed as an attractive alternative to mitigate drawbacks that other drug delivery systems (DDSs) face, [5,8] as they can exhibit low toxicity,good clearance,high drug loadings,a nd are easy to functionalise,y et examples of MTV-MOFs for drug delivery applications are still limited. [9] Thec oordination modulation (CM) protocol [10] -in which monotopic ligands (modulators) compete with the MOF linkers for metal cluster coordination sites during synthesishas been widely studied to control physical properties such as size, [11] crystallinity,c olloidal dispersion, [12] stability and porosity (through defect chemistry), [13] while we have recently shown that it can be used to control MOF surface chemistry and functionality. [14] Additionally,w eh ave introduced the concept of in situ defect drug loading in MOFs,showing that using dichloroacetic acid as am odulator during assembly of UiO-66-type MOFs of ideal formula [Zr 6 O 4 (OH) 4 (L) 6 ] n , results in highly porous,w ell-dispersed nanoparticles with ah igh incorporation of dichloroacetate (DCA)a sadefectcompensating ligand. [15] Thep rotocol is amenable to surface functionalisation, both during synthesis by incorporating as econd modulator and through postsynthetic coating, and also allows further postsynthetic drug loading of 5-fluorouracil (5-FU)i nto the MOF pores, [14c] in both cases without major DCA leakage.M any clinical anticancer treatments involve acocktail of drugs (e.g.FOLFORINOXcombines 5-FU,l eucovorin, irinotecan and oxaliplatin against metastatic pancreatic adenocarcinoma) [16] yet examples of MOFs capable of delivering multiple drugs are relatively scarce, [8d, 17] suggesting defect-loading could be an attractive strategy for preparing multimodal chemotherapeutic formulations.
Despite the fact that multiple drugs contain metal-binding units such as carboxylates,and the easier industrial manufacturing of producing drug-loaded MOFs in one-pot syntheses, examples of drugs used as modulators are still uncommon in the literature. [18] Based on our previous work [14c, 15] we sought to examine the potential for incorporating multiple drugs as modulators in one-pot syntheses in ap rocess we call MTV modulation (MTVM, Figure 1b). Herein we show it is possible to introduce up to three drugs into the Zr benzene-1,4-dicarboxylate (BDC) MOF UiO-66 with versatility based on the simultaneous introduction of carboxylate and phosphonate-containing moieties.O ur MTVM protocol provides particle size control (ca. 100 nm), while the resultant multidrug MTVM MOFs retain their porosity,w hich is used to postsynthetically store af ourth drug (Scheme 1a nd SI, Section S2). We believe that the MTVM protocols could be applied to aw ide range of modulators for alternative enhanced applications such as catalysis and gas adsorption/ separation.

Results and Discussion
Previously,w ea nd others have shown that UiO-66 is biocompatible, [8c,19] while the cytotoxicity of DCA-loaded MOFs is dependent on surface chemistry, [14c,d] and incorporation of as econd drug can dramatically enhance overall cytotoxicity. [15] As well as DCA,w eh ave chosen a-cyano-4hydroxycinnamic acid (a-CHC), am olecule recently proposed as an anticancer agent, due to the relatively low pK a of its carboxylate unit (2.2), comparable with the one of DCA (1.4). As ap roof-of-concept of the effect of the pK a in the drug-modulated synthesis,w eh ave also studied modulation with ibuprofen (IBU), due to its similar structure to a-CHC and the higher pK a of its carboxylate (4.9). Alendronate (AL) [20] is an anticancer drug which contains two phosphonates-which are expected to show higher affinity for Zr than carboxylates-with afirst pK a of 2.4. 5-FU has been chosen as ad rug to be loaded postsynthetically,a si td oes not contain ametal-binding unit.
At first, we postsynthetically loaded our previously reported DCA@UiO-66, [15] prepared by coordination modulation, separately with 5-FU, a-CHC, IBU,a nd AL (Scheme 1). While 5-FU loading resulted in ca. 1.2 %( w/w) incorporation with minimal DCA leakage,a sd etermined by ICP-OES,p ostsynthetic loading of drugs containing metalbinding units resulted in partial or total displacement of DCA (Table S1), as determined by 1 HNMR spectroscopic analysis of acid-digested samples,s uggesting their loading occurs through attachment to the Zr positions (subsequently detaching DCA)r ather than pore storage (see SI, Section S3.1 for full characterisation). AL loading resulted in ac omplete structural change,likely due to the affinity of its phosphonate groups for Zr inducing breakdown, [21] again confirming the competing coordination of the drugs.A ss uch, this postsynthetic process was not considered viable for multiple drug loading.
Modulation with asingle drug, either a-CHC, IBU or AL, results in crystalline MOFs with the UiO topology, [22] as confirmed by PXRD (see SI, Section S3.2 for full characterisation), although in the case of AL@UiO-66,n ew Bragg peaks could be observed in the PXRD pattern, suggesting some incorporation of the bis-phosphonate AL as alinker and subsequent minor structural alteration. 1 HNMR spectra of the acid digested samples confirmed the presence of modulators,a nd their content is tabulated in Table S2. However, scanning electron microscopy (SEM) showed that all the individual drug-modulated MOFs consisted of aggregates of various sizes and shapes,a nd hence could not be used for biomedical application, as monodisperse colloidally stable samples are imperative for drug delivery. [23] Adding DCA as aco-modulator of each MTVM synthesis overcomes the sample aggregation issue,r esulting in crystalline MOFs with Bragg reflection peaks characteristic of the UiO-66 topology in all cases (Figure 2a), although AL/ DCA@UiO-66 had poorer crystallinity with some additional reflections again suggestive of structural change.T he modulator content (see SI, Table S3) can be measured in mol % compared to the linker BDC by 1 HNMR spectroscopic digestion;w hile this allows an assessment of relative incor- Molar loading values are in concordance with the modulators expected affinities for Zr. a-CHC/DCA@UiO-66 contained 6.7 mol %o fa-CHC (ca. 1 a-CHC per 12 BDC), and % 35 mol %o fDCA (ca. 1 DCA per 3B DC) whilst IBU/ DCA@UiO-66 showed as maller degree of incorporation of IBU (2.9 mol %compared to BDC) due to its higher pK a , [13a] and similarly % 37 mol %o fDCA.A lendronate had the highest incorporation, again likely representative of its ditopicity and the Zr-phosphonate affinity, [21] as AL/ DCA@UiO-66 contained 38.1 mol %( ca. 1 AL per 2.5 BDC), suggesting its possible role as al inker,a nd 18.8 mol %ofDCA (ca. 1 DCA per 3.5 BDC). Thermogravimetric analysis of samples showed complex overlapping thermal degradation processes,o ften with lowered thermal stabilities compared to pristine UiO-66, confirming that the drugs are anchored to the MOF structures but not allowing quantification of drug loading using this technique.
Thes amples were highly porous (Figure 2b), confirming, together with TGA and FT-IR spectroscopy (See S3.3 for full characterisation) the attachment of the modulators to the Zr positions.Infact, as ageneral trend for the DCA-modulated IBU and a-CHC samples,p orosity was notably enhanced (S BET increased from % 1000-1100 m 2 g À1 to % 1500 m 2 g À1 ) compared to the single-drug modulated analogues,whereas in the case of AL/DCA@UiO-66 the porosity of the sample is reduced (S BET decreased from 1245 m 2 g À1 to 369 m 2 g À1 )upon DCA addition, likely as ac onsequence of the much higher incorporation of alendronate and poorer overall crystallinity indicated by PXRD.U nfortunately,d ue to the defective nature of the samples,inwhich modulators replace the linkers in the structure,e xact structural determination is trouble-some.The MOFs were easily dispersed in phosphate-buffered saline (PBS 10X, pH 7.4), with dynamic light scattering (DLS) measurements (Figure 2c)i ng reat agreement with the particle size determined by SEM (ca. 100 nm, Figure 2d), confirming,together with our previous results, [14c, 15] that DCA co-modulation serves as as ize control protocol, that also enhances the colloidal dispersion of the samples. [12] This versatile synthetic protocol to introduce drugs during synthesis using DCA as aco-modulator was further explored to introduce three drugs (a-CHC, AL and DCA; IBU was not further investigated as it is not an anticancer drug) during synthesis into asingle MTVM MOF structure to give a-CHC/ AL/DCA@UiO-66.Asthe Cl (DCA)and P(AL)content of the samples can be determined by ICP-OES,and the a-CHC content by UV/Vis spectroscopy of digested samples,t he corresponding drug loadings by mass are tabulated in Table 1, as these values are more relevant for subsequent cytotoxicity experiments.
Despite the significant incorporation of the multiple drugs into UiO-66 structure (Table 1), a-CHC/AL/DCA@UiO-66  was found to be highly crystalline (Figure 2a,s ee SI, Section S3.4 for full characterisation). Thes ample again maintained ah igh AL loading of 21.6 %( w/w), as well as 3.2 %( w/w) loading of DCA,a nd 2.0 %( w/w) loading of a-CHC.A ssuming binding of monoanions at defects,t his corresponds to a a-CHC:DCA:AL molecular ratio of 1:2.3:8.1, in great agreement with reports showing the high affinity of AL for UiO-66 Zr clusters [20] and of the role of pK a in defect binding.T GA, FT-IR spectroscopy and N 2 adsorption and desorption measurements confirmed the drugs to be attached to the Zr positions,a sa-CHC/AL/DCA@UiO-66 has asurface area of 634 m 2 g À1 despite containing over 25 % (w/w) of drugs in its structure (Figure 2b). Thelower porosity could also be attributed to the high incorporation of AL,a s seen with the analogous AL/DCA@UiO-66 sample.A comparison of the drug incorporation ratios with the surface areas and pore volumes,i ndicative of the level of defectivity and/or structural change,isgiven in the Supporting Information, Table S5.
Once again, size control was achieved resulting in nanoparticles of ca. TheM TVM drug-loaded MOFs were tested for anticancer selectivity against MCF-7 breast cancer and HEK293 kidney cells by the MTS assay (See SI, Section S4) after 72 hours of incubation and compared to the effect of the free drugs,w hich allows comparison with our previous work. [14c] Thesimilar size of all MOFs (ca. 90-150 nm) allows comparison of therapeutic efficacyw ithout concerns over major particle size influence.F or both cell lines,t here is approximately an order of magnitude difference in the cytotoxicities of the free drugs,with IC 50 values in the order AL < 5-FU < a-CHC < DCA (Table S6), and each drug shows as mall selectivity for cytotoxicity towards the MCF-7 cancer cell line compared to HEK293. In order to delineate the effects of the MOF delivery vehicle on the efficacies of the different drugs individually and in tandem, cytotoxicities of the MOFs containing either AL, 5-FU or a-CHC alongside DCA are first assessed (Figure 3).
In great agreement with our previous studies of ultrasmall UiO MOFs for dual delivery of 5-FU and DCA, [15] 5-FU@DCA@UiO-66 was profoundly more cytotoxic than free 5-FU towards MCF-7 cells (Figure 3a), with an IC 50 dose normalised to 5-FU over 35 times lower than the free drug, decreasing cell proliferation to values to ca. 45 %a fter treatment with 25 mgmL À1 of MOF for 72 hours (IC 50 values for free drugs are listed in Table S6, and are tabulated for all MOF samples,n ormalised to the varying components,i n Tables S7 and S8). Its cytotoxic effect towards HEK293 was ca. 21 times that of the free drug, so whilst toxicity also increased towards non-cancerous cells,t he overall selectivity compared to the free drug improved nearly two-fold (Figure 3b). It has previously been reported that DCA enhances the anticancer activity and selectivity of certain drugs, including 5-FU, [24] however, doses at least ten times higher than those used here are usually needed to generate as ynergistic effect when the drugs are not loaded into aDDS. [15] Ap ronounced enhancement of the therapeutic effect of a-CHC towards MCF-7 cancer cells was found for a-CHC/ DCA@UiO-66 (Figure 3c). Despite both a-CHC and DCA having IC 50 doses against MCF-7 cells in the millimolar range, the IC 50 of the drug-loaded MOF towards MCF-7 corresponds to amaximum delivered dose of a-CHC that is 27 times lower than that for the free drug and to adelivered dose of DCA 111 times lower than the IC 50 of the free drug. This dramatic enhancement is not observed for the HEK293 cells,where the MOF is biocompatible up to 1mgmL À1 of MOF (Figure 3d). Taken together,t his again corresponds to an increase in cytotoxicity and selectivity towards MCF-7 versus HEK293.
In contrast, ad ecrease in the therapeutic effect of alendronate towards MCF-7 breast cancer cells was found for AL/DCA@UiO-66 (Figure 3e). Theincrease of the IC 50 of AL upon delivery from AL/DCA@UiO-66 corresponds to amaximum delivered concentration 7times higher than that of the free drug. Thebis-phosphonate structure of AL is likely to be strongly adhered to the Zr positions of the MOF,and so incomplete release may attenuate cytotoxicity and raise the possibility of slower,controlled release to mitigate side effects in vivo.Itisimportant to note that the MOF was not cytotoxic towards HEK293 cells up to 1mgmL À1 of MOF (Figure 3f), aformulation loaded with an AL content 49 times higher than the IC 50 of the free drug, meaning that although there is no enhancement in the therapeutic activity of the drug towards MCF-7 cancer cells,there is again aremarkable increase in its selectivity.
Forthe triple-drug formulations, a-CHC/AL/DCA@UiO-66 again increased the IC 50 against MCF-7 cells when normalised to AL content, to adose around 1.5 times higher than that for the free drug ( Figure 4). Incubation with 25 mgmL À1 of a-CHC/AL/DCA@UiO-66 decreased MCF-7 cell proliferation to 34 AE 6%,further decreasing to 18 AE 11 % upon treatment with aconcentration of 250 mgmL À1 of MOF. Evaluating the MOFs by the loading of their most cytotoxic drug component, AL,s uggests a-CHC/AL/DCA@UiO-66 exhibits around 57 times the cytotoxicity towards MCF-7 compared to AL/DCA@UiO-66,a nd is also significantly more cytotoxic than comparable doses of a-CHC/ DCA@UiO-66.
Thee ffect towards HEK293 cells,h owever was remarkably different. a-CHC/AL/DCA@UiO-66 was found to be biocompatible up to 1mgmL À1 of MOF (maximum AL delivered dose 38 times higher than the IC 50 of the free drug); the selectivity of cytotoxicity against MCF-7 versus HEK293 is again enhanced compared to AL alone.
Loading the MTVM MOF with af ourth drug, 5-FU, resulted in af urther enhancement of the IC 50 dose of 5-FU@a-CHC/AL/DCA@UiO-66 towards MCF-7 cells,w ith the IC 50 normalised to AL content just 10 %h igher than the free drug. Biocompatibility with HEK293 cells was maintained up to 0.2 mg mL À1 MOF incubation MOF (maximum AL delivered dose 7.5 times higher than the IC 50 of the free drug), suggesting that multimodal drug delivery can synergistically enhance anticancer activity while conserving the biocompatibility towards HEK293 cells at the concentrations studied.

Conclusion
On the whole,wehave demonstrated that different drugs, containing either carboxylates or phosphonates as metalbinding units,can be introduced to the synthesis of UiO-66and potentially any other Zr MOF of the UiO family-as simultaneous modulators.T he drugs become attached to the Zr clusters of the resulting MOF,which has been found to be related to both the pK a of the metal binding unit (the lower the pK a ,t he higher the incorporation) and its chemical functionality (phosphonates have ahigher affinity for Zr than carboxylates and are incorporated more). As the drugmodulators are attached as defects rather than pore-loaded, the resultant MTVM MOFs are highly porous,a nd we have used their porosity to postsynthetically load 5-FU,ultimately resulting in four drugs incorporated in significant quantities into asingle nanovector. We have shown that adding DCA to the drug modulated syntheses also offers size control, resulting in nanoparticles of ca. 100 nm that are welldispersed in PBS (10X), which enables the comparison of their cytotoxicity without concerns over size effects.T he anticancer therapeutic activity of the double drug combinations towards MCF-7 breast cancer cells is highly increased for 5-FU@DCA@UiO-66 and a-CHC/DCA@UiO-66 compared to the free drugs,whereas the MOFs are biocompatible to HEK293 kidney cells even at high doses,e nhancing selectivity.A lthough the therapeutic activity of AL when loaded into the MOFs is reduced in all cases compared to the free drug, likely as it is not fully released from the core of the MOF,increases in cytotoxicity towards MCF-7 cells are noted for treble and quadruple drug formulations as the drug cocktails become more complex. Additionally,adrastic increase in selectivity towards cancer cells is achieved across the formulations; 5-FU@a-CHC/AL/DCA@UiO-66 maintains the cytotoxicity of free AL towards MCF-7 cells yet is considerably more biocompatible towards HEK293 cells than free AL,s uggesting that drug delivery using MOFs as DDSs could overcome the unwanted cytotoxic issues of some drugs.
Our MTVM protocols are highly versatile and reproducible (both carboxylates and phosphonates can be introduced simultaneously into one single phase during synthesis), and could be applied to almost any drug containing potential metal-binding units (e.g.d oxorobucin or paclitaxel) and to other MOF systems,o pening ab road range of possibilities and combinations to design novel drug loaded MOFs.W hile we have used biologically relevant molecules as modulators in this study,weare convinced that the MTVM protocol can be used to introduce modulators to enhance other applications through cooperative effects,s uch as preferential gas capture or heterogeneous catalysis,a nd also to tune the host-guest interactions of the frameworks through pore environment control to favour the uptake of certain gases (gas uptake/ separation), chelating units (water treatment/ pollutant removal) and cooperative units for catalysis,a mong many other possibilities.