Interaction of [V IV O(acac) 2 ] with Human Serum Transferrin and Albumin

: [VO(acac) 2 ] is a remarkable vanadium compound and has potential as a therapeutic drug. It is important to clarify how it is transported in blood, but the reports addressing its binding to serum proteins have been contradictory. We use several spectroscopic and mass spectrometric techniques (ESI and MALDI-TOF), small-angle X-ray scattering and size exclusion chromatography (SEC) to characterize solutions containing [VO(acac) 2 ] and either human serum apo-transferrin (apoHTF) or albumin (HSA). DFT and modeling protein calculations are carried out to disclose the type of binding to apoHTF. The measured circular dichroism spectra, SEC and MALDI-TOF data clearly prove that at least two VO– acac moieties may bind to apoHTF, most probably forming [V IV O(acac)(apoHTF)] complexes with residues of the HTF binding sites. No indication of binding of [VO(acac) 2 ] to HSA is obtained. We conclude that V IV O–acac species may be transported in blood by transferrin. At very low complex concentrations speciation calculations suggest that [(VO)(a-poHTF)] species form.


Introduction
Bis(acetylacetonato)oxidovanadium(IV), or [V IV O(acac) 2 (1), is one of the most remarkablev anadium compounds.I nt he solid state it consists of discrete molecules of [V IV O(acac) 2 in which vanadium(IV) presents an almost square pyramidal geometry ( Figure 1). [1] It forms quite stable V IV O-complexes, [2,3] the reporteds tability constants in aqueous solution being log b 1 = 8.73 and log b 2 = 16.27,respectively. [4] [V IV O(acac) 2 ]h as found uses in multiple areas. Namely,i th as been used as the vanadium precursor for the preparation of many vanadium complexes where one [5][6][7] or both [8,9] acac À li-gands are replaced. [10] It has also been used directly as catalyst precursor in organic chemistry,f or example, in aerobic oxidations, [10,11] epoxidations (e.g. of allylic alcohols, geraniol), sulfoxidations [12] in combination with tert-butylhydroperoxide or H 2 O 2 , [13,14] or immobilized in solid supports. [15] [V IV O(acac) 2 ]e xerts interesting biological effects;n amely it exhibits insulin-enhancing properties, in that in can stimulate the phosphorylation of protein kinase B( PKB/Akt)a nd glycogen synthase kinase-3( GSK3). [16] It has also been shown to inhibit tyrosine phosphatases (PTPases), such as PTP1B. [16] Thee ffects of [V IV O(acac) 2 ], [V IV O(Et-acac) 2 ]a nd of af ew other V IV Ocompounds on the glycemiao fs treptozotocin-induced diabet-ic rats (STZ-rats) were examined. [17,18] [V IV O(acac) 2 ]w as also reported to have antiproliferative effects [2,[19][20][21] and to have potential as both antidiabetic and antipancreatic cancera gentt o preventort reat patients suffering from both diseases. [19] Another property of [V IV O(acac) 2 ]i si ts remarkable activity in degrading plasmid DNA in the absence of any activating agents, air and photoirradiation. [20,21] The nature of the pH buffer wasf ound to be determinant in the nuclease activity, and in phosphate-buffered mediums ingle-strand cleavagei s clear for concentrations of 1 as low as 1.2 mm (corresponding to ar i = 0.08), and in the presence of activators it is much more extensive. [20,21] The mechanism is oxidative and mainly associated with the formation of reactive oxygen species (ROS). [21] Hydrolytic cleavage of the phosphodiester bond is also promoted by 1,b ut at much slower rate, not competing with the oxidative mechanism. [20] These properties mayh ave important implications in the interpretation of the biological activity of [V IV O(acac) 2 ].
[V IV O(acac) 2 ]t hus exhibits insulin-enhancing properties and severalo ther relevant biological effects. [2,16,22,23,24] Since human serum albumin (HSA) serves as ad rug transportc arriera nd human serum transferrin (HTF) is an important metal ion blood carrier,t he understanding of the interactions that may be establishedb etween [V IV O(acac) 2 ]a nd these plasma proteins is of major importance to understand its pharmacokinetics and pharmacodynamics. Additionally,m ost proposals explaining the insulin-enhancing properties of vanadium compounds involve formation of protein-vanadium complexes, namely in blood, [21] thus, the understandingo ft he speciation of [V IV O(acac) 2 ]i nb lood and how this might be important in the uptake of vanadium by cells, are relevant issues.
Regarding the transporto fv anadium compounds in blood in vivo, most studies agree that these are predominantly transported bound to serum proteins,p articularly HTF. [24][25][26][27][28][29][30][31][32] More recently the uptake of V-species by red blood cells (RBCs) and its possible role was highlighted, [33][34][35] but at the low total vanadium concentrations expected in vivo in the blood following oral administration of Va ntidiabetics, most probably significantly less than ca. 20 mm even upon oral treatment with aV C, [30,36] HTF appears to be the main vanadium transporter and RBCs possibly play some role in the pharmacokinetics of these compounds and in providing ar educing effect, eventually allowing the formation of V III ,w hich may strongly bind to HTF. [37] The binding of [V IV O(acac) 2 ]t oa lbuminw as studied by EPR and ENDOR spectroscopies, [38] and Mustafi et al. [39] used isothermalt itration calorimetry (ITC) and spectrofluorometry to determine the binding constant of [V IV O(acac) 2 ]t ob ovine serum albumin (BSA). It was reported that glucose uptake by 3T3-L1a dipocytes was significantly higherw hen using [V IV O(acac) 2 ]i nt he presence of HSA (when the albumin: [V IV O(acac) 2 ]r atio ! 1.0). [2,38] These authors also stated that serum albumin enhanced the insulin-like activity of all studied V-complexes,i ncluding 1,m ore than human serum transferrin, [40] and that the characteristics described are relatedt ot he high stabilitya nd capacity of [V IV O(acac) 2 ]t or emain intact upon binding to HSA. Moreover,t he specificityo f1 for highly glycolytic cells, its low toxicity [25] and its anti-tumorigenic properties [23,24,41,42] contribute to make 1 ap otentially useful contrast agent for early tumor detection, and possibly also for treatment. The advantage of [V IV O(acac) 2 ]f or this purpose would result from its binding to serum albumini ncreasing its serum half-life and allowing its transporta nd delivery to its targeted tumor site. [39] Garribba and co-workers [27,29,[43][44][45] studied extensively the binding of V IV O-complexest ob lood components, mainly by using EPR spectroscopy.N amely,t he anisotropic EPR spectra of solutionsc ontaining V IV O 2 + -HTF,V IV O 2 + -acac,V IV O 2 + -apoHTFacac and V IV O 2 + -holoHTF-acac, in the presence and absence of other blood serum components, were recorded at pH 7.4. Globally the authors report that two different species could be distinguished, one in higherc oncentration,i dentified as (V IV O) 2 apoHTF,a nd the 2 nd one as [V IV O(acac) 2 ], and no V IV Oacac-apoHTF species were identified. These authors further stated that the EPR spectroscopic data suggest that no mixed complexes are formed, this being in agreement with acetylacetonate not acting as as ynergistic anion. [29,45] In similar studies with HSA, the authors report that only [V IV O(acac) 2 ]i sd etected,a nd that the EPR spectra of solutions containing [V IV O(acac) 2 ]a nd 1-methylimidazole are practically indistinguishable from those of the binary systemV IV O 2 + -acac, [29] no V IV O-acac-HSA speciesb eing identified. [29,44] Similarc onclusions were made from the measurement of the anisotropic EPR spectra of solutionsc ontaining [V IV O(acac) 2 ]a nd holoHTF:n o [V IV O(acac) 2 (holoHTF)] speciesw ere detected, [45] and again it was considered that [V IV O(acac) 2 ]h as as quare-pyramidal coordination geometry and has no tendency to form complexes of the type cis-[V IV O(carrier) 2 (protein)],t his being allegedly corro- borated by the conclusion that no ternary species are formed by V IV O 2 + ,a cetylacetonate and 1-methylmidazole. Similarc onclusionsw ere made by the same group in studies with Immunoglobulin G. [43] Notwithstanding, contradicting the initial statements of Selbin, [46] [V IV O(acac) 2 ]i s aw eak Lewis acid, and for example, upon dissolution in organic solvents or in solutions containing potentiall igands, the vanadium mayc oordinate ad onor ligand in the vacant site, forming complexes expressed as [VO(acac) 2 (L)],s ometimes designateda sa dducts, with for example,a mide, amine, sulfoxide and pyridine solvents/ligands, [3] and this process [Eq. (1)] has been studied by severalm ethods. [3,47,[48][49][50][51][52][53][54] ½V IV OðacacÞ 2 þ L Ð½V IV OðacacÞ 2 ðLÞ ð1Þ In earlier work, solvente ffects were explained by coordination to the vacant axial position, but this was later questioned in severalp ublications [3] and the available literature is not always in agreement. In fact, in solution, while some authors state that [V IV O(acac) 2 ]b ind solvents in trans position, [49,50] several studies also showed that somes olvents bind to the V IV equatorially,a nd other publications report either cis-o rtransbinding depending on the ligand added. [51][52][53] With some solvents, the interaction (e.g. hydrogen bonding) with the Ooxido ligand has also been proposed. [55][56][57][58] For example, IR studies with pyridine derivatives demonstrated the existence of adducts with the sixth ligand bound cis rathert han trans to the oxido O-atom (Figure 1), [52,54] and ENDOR studies showed that whilef or example, methanol binds to the vacant axial 6 th position, it was shown that either cis or trans isomersa re formed with some substituted pyridines. [51] In fact, although this may result from solids tate effects, as ingle crystal X-ray diffraction (XRD)s tudy of the 4-phenylpyridine (Phpy) adduct 3 confirmed the possibilityo fcis binding of substituted pyridines ( Figure 2). [59] On the other hand, the structure determined by single crystal XRD of [V IV O(acac) 2 (4-methylpyridine)] consists of discrete molecules with 4-methylpyridine in the trans position and short V = O( 1.557 ) and long V-N( 2.447 )b ond lengths. [60] In the adduct formed with dioxan, [V IV O(acac) 2 (dioxan)], [61] the dioxan bridges two [V IV O(acac) 2 ] molecules axially (trans)t ot he Vc enters;i n[ V IV O(acac) 2 (2-pyri-done)] the 2-pyridone is bound axially, [62] as pyrazole( Pz) in [V IV O(acac) 2 (Hpz)] (2). [63] Overall, taking into account the literature data, the interaction of ligands( solvento radissolved molecule) with [V IV O(acac) 2 ]m ay involve: [3] (i)coordinationi nas ixth position trans to the oxido O-atom, (ii)coordination in as ixth position cis to the oxido O-atom, and (iii)hydrogen bonding to the Ooxido or atoms of the ligand (e.g. with chloroform [51,56] ).
HTF has 8t ryptophan residues (3 in the N-lobe, 5i nt he Clobe) and 25 tyrosines (14 in the N-lobe, 12 in the C-lobe). [64] The two iron bindings ites are located near the junction of two domainsf ormed by Cys-117 to Cys-194 bond in the N-terminus. Potential amino acid residues that may act as ligandsf or V IV in the iron binding sites are Tyr188, Tyr95, His249 and Asp63 (N-lobe) and Tyr426, Tyr517, His585a nd Asp392 (Clobe). Side groups of other residues of HTF may bind V IV , namely imidazole donors from histidines, and His-14, His-289, His-349,H is-350, His-473,H is-606 and His-642 have been considered as good candidates. [45] The structure of human serum albumin consists of as ingle polypeptide chain with 585 amino acid residues, with a3 -dimensional structure normally described in terms of 3h omologous chains (I, II and III), each of them formed by two subdomains( Aa nd B). HSA has one tryptophanr esidue (Trp214 located in subdomain II-A) and 18 tyrosine residues.
As described above,t here are many reports on in vitro and in vivo experiments provings everal relevantb iological effects of [V IV O(acac) 2 ]. It has been repeatedly emphasized [2,4,45,65]  In aqueous solutions of [V IV O(acac) 2 ]C rans and coworkers [17,65] reported the observation of three species by EPR spectroscopic measurements at room temperature whose concentrations were time, pH, temperature and salt dependent. The three com- Mustafietaland Makinen et al. [2,38,58] provided adistinct speciation diagram where [V IV O(acac) 2 ]i sc onsidered stable in aqueous solutions of pH 2. However,t his contradicts other observations, [4,17] considered compatible with the formation of [V IV O(H 2 O) 5 2 2 ]a st he pH is increased. In the SI section we discussw hy we consider Figure 3 to correspond to the correct speciation, also demonstrating that [V IV O(acac) 2 ]i sn ot stable in aqueous solutions of pH 2.
As mentioned above, there are severals tudies in the literature reporting the coordinativeb inding of amines and of other monodentate ligands to [V IV O(acac) 2 ]i ns olution, namely including the determination of binding constantsK[Eqs. (1) and (2)]. [47,48] K ¼ ½VOðacacÞ 2 ðLÞ ½VOðacacÞ 2 ½L ð2Þ Most of these studies were done by electronic absorption measurements in organic solvents (e.g. benzene, nitrobenzene, dichloromethane);f or example, for pyridine the values determined in CH 2 Cl 2 and benzene were 867 AE 53 [48] and 56 AE 5, [47] respectively.T he value of Ko bviously also depends on the amine considered;f or example, for piperidine and pyrrolidine, the values determined in benzene were 1400 AE 400 and 2900 AE 1000, [47] respectively.I na queous media the Kv alues will certainlyd iffer from these, and the degree of formation of the adducts [V IV O(acac) 2 (L)] will also depend on the pH of the solution and pK a of the amine.
We measured the spectra of aqueous solutions containing [V IV O(acac) 2 ]a nd several potentially monodentate ligands( see experimental and SI sections for details): imidazole, 2-Me-imidazole, benzimidazole, pyrazole, 4-t-buthylphenol at pH 7.0 AE 0.2. Considering the speciation diagram of the V IV O-acac system with C VO = 2mm and C acac = 4mm (Figure 3), in the pH range ca. 6.2-7.2 the amount of [V IV O(acac) 2 ]i sm aximized, while maintaining its hydrolysis at ar easonable low level (less than ca. 5% of total vanadium).   It is clear that, in the presenceo fi midazole (or 2-Me-imidazole or pyrazole),asignificant globali ncrease in the e values in the 380-900 nm range takes place, which can only be explained by the binding of the N(3) atom of imidazole (or Natoms of 2-Me-imidazole or pyrazole) to [V IV O(acac) 2 ]. It may be observed that for Im:[V IV O(acac) 2 ]r atios of 1:2o r1 :4 the spectra do not differ much from the one recorded at a1 :1 ratio, this meaning that at this ratio most of the imidazole was already bound to [V IV O(acac) 2 ]. The possibility of non-coordinating interactions explaining the increase in absorption values cannotb er uled out. However,t he change in the spectra observed is similart ot hose observed in the monodentate binding of O-carboxylate of several hydroxycarboxylic acids [66] and amino acids, [67][68][69][70][71][72][73] where the monodentatec oordinative binding was demonstrated by several spectroscopic methods and pH-potentiometric titrations.
To furtherp rove the possibility of monodentate binding of these monodentate ligandst o[ V IV O(acac) 2 ], rather similar experiments werec arried out with l-lactic acida tp H% 6.0;t hese are described in the SI section( SI-8.4). Figure  There are some subtle changes in the absorptions pectra of the solutionsa tl max % 560 and 820 nm akin to those observed in the similar experimentsd escribed above (e.g. Figure 4). Importantly,w eak but clearly measurable CD spectra were recorded with the same samples which The additions of benzimidazole and of 4-tert-butylphenol weree xamples where this effect of absorption increased id not depict ac lear trend (see SI section). In the case of benzimidazole possibly because of steric hindrance;i nt he case of 4-tert-butylphenol probably also because of its relatively high pK a value (10.2), thus at pH 7.0 the phenolicO Hr emains protonated andi ts propensity to bind [V IV O(acac) 2 ]i sl ow.

Mass spectrometric studies.
To further confirm the formation of [V IV O(acac) 2 (L)] species in aqueous media, solutionsofthe same potentially monodentate ligands:i midazole, methylimidazole, benzimidazole, pyrazolea nd phenol (this insteado f4 -tertbutylphenol) were prepared in 10 mm NH 4 CH 3 COO aqueous solution (previously set to pH 6.5). Accurately measuredv olumes of each of these solutionsa nd of the NH 4 [74,75] we explaino ur our data considering coordinative binding. Non-coordinative binding is much morep robable with solvent or buffer molecules, present in much higher amounts. 25 depicts the low field range of these spectra.  [31,67,76,77] The typeo fs pectrum obtained for V IV O-compounds in the visible range depends on the particularb inding and environment of the chiral donors around the V IV Oc enter. If more than one chiral V IV O-complex forms in solution, the measured CD spectrumi st he weighted sum of the CD spectra of all V IV Ocomplexes formed in the system being studied, each corresponding to as et of distinct De(l)v alues (see SI-section). Figure 5i ncludesC Ds pectra in the 400-1000 nm range of solutions containing apoHTFa nd V IV OSO 4 (blue lines). It is known and accepted [26,27,29,31,34,64,68,78] that in such solutions the V IV O 2 + is bound by donora toms of residues of the iron binding site of HTF,f orming either 1:1o r2 :1 (V IV O:apoHTF ratio) complexes.
The fact that non-zero De values are measured in the visible range for solutionsc ontaining [V IV O(acac) 2 ]a nd apoHTF ( Figure 5) and that the pattern of CD spectra obtained differs from those of solutionsc ontaining V IV OSO 4 and apoHTF( containing (V IV O) n HTF complexes), means that: (i)significant amounts of V IV O-acac species bind to HTF and (ii)there is no extensive hydrolysis of the V IV O-acac species leadingtot he formation of (V IV O) n HTF complexes.
It is also noteworthy that the order of magnitude of the measured De values for the system [V IV O(acac) 2 ]a nd apoHTF is the same as those measuredf or solutions containing (V IV O) n HTFc omplexes.F rom the CD spectra shown in Figure 5 it cannot be concluded with certainty what type of V IV O-acac-HTF complexes are formed, but the relativelyh igh De values measured suggestst hat as ignificant fractiono ft hese complexes correspond to [V IV O(acac)(HTF)] species, where the V IV O:acac molar ratio is 1:1, and the V IV is bound to more than one chiral residueo fa poHTF, [26] as otherwiset he De values would be significantly lower,a sw as the case of those measured in solutions containing [V IV O(picolinato) 2 ]a nd lysozyme, [79] or in solutions containing V IV O 2 + and amino acids with monodentate coordination of the a-COO À moiety to V IV . [67][68][69][70][71][72] At least two distinct types of CD spectra may be distinguished in Figure  :apoHTF molarr atio of 3d oes not produce significant changes in the CD spectra measured. Thus, if more than two V IV O-acacs peciesb ind to apoHTF,t his is not clearly visible in the CD spectra measured.

Small-angle X-ray scattering( SAXS).
Data from native apo-transferrin were evaluated for concentration dependence using Primus, [80] and extrapolated to zero concentration. Using this method, ar adius of gyration of 33.2 was derived from the experimental data, with approximate molecular weight of 74 kDa, andP orod volume 137 000 3 .T his is in agreement with the known molecular weighto f% 79 kDa. The Kratky plot indicates ag lobular,m ultidomain protein, as expectedf rom the published PDB structures. TheP (r) functiond erived from the data indicates D max (the maximum atomic distance vector in the system)t ob e 112 as shown in Figure SI-9-1. Shape reconstruction using the data pipeline described in the SI sectionw as carried out, with the filtered and refineds tructures shown in FigureSI-9-2. The data indicates interesting differences between the solution state structure and that derived from X-ray crystallography. Using the PDB model 2HAV, [81] at heoretical R g (radiuso fg yration) of 31 was obtained using the CRYSOLp ackage. [82] The theoreticalf it, shown in Figure SI-9-3, is satisfactory,s uggesting that, despite the described R g difference, the overall envelope shape of the solution state and crystallographic structure is similar.
As imilarm ethodology was followed for the apoHTF-[V IV O(acac) 2 ]s ample. Comparing the results obtained from both samples,t he sample containing apoHTF and [V IV O(acac) 2 ] showedi mmediate, detectable differenceso nt he SAXS length scales (Figure SI-9-4). In particular, the incubated sample was shown to possess significantly smaller radius of gyration at 30.8 ,a nd smaller derived molecular weight, at 66 kDa. Moreover,e ven if the indicated globular structure is conserved, the normalized Kratky plot appears to indicates ignificant differences in domain structure from the native apoHTF.N ot surprisingly,t here is also as ignificant difference in the P(r) function between the native and the incubated sample ( Figure 6) with the concomitant discrepancy in D max values (112 and 104 , respectively). As hape reconstruction is shown in Figure SI-9-6. Therefore, the analysis of the SAXS data strongly suggests ac lear modification on the conformation of apoHTF upon [V IV O(acac) 2 ]b inding, suggesting the existence of relatively strong protein-complex interactions.

Maldi-TOF Mass spectrometric data.
As described in the experimental sectiont he samples for MALDI-TOF MS were prepared with apoHTF:[V IV O(acac) 2 ]m olar ratios of 1:0, 1:1, 1:2, 1:3a nd 1:5b ym ixing different volumes of the stock solutions with aN H 4 HCO 3 buffer (pH 7.4, 25 mm). The results hereby reported were obtained with samples prepared by the Dried Droplet procedure, and each final spectrum was the accumulated result of at least 1000 laser shots that were obtained from 10 different manually selected regions of the same sample, over the range 14 000-160 000 Da.
In as et of experiments, the average mass obtainedf or apoHTF was 79 247(AE 20), while the average mass obtained with samples containing 1:3a poHTF:[V IV O(acac) 2 ]m olar ratios was 79 581(AE 20). Thus, the difference in masses is significant ( % 334). We assign this difference to the binding of two [V IV O(acac) + ]m oietiest oa poHTF,w hich corresponds to % 332 Da. Figure 7d epicts two representative Maldi-TOF mass spectra.
In ad istinct set of experimentst he [V IV O(acac) 2 ]:apoHTF molar ratios were increased and the differences in masses were 201 (for 1:1), 346 (for 2:1) and 346 (for 3:1), this agreeing with the binding of one [V IV O(acac) + ]m oiety to apoHTF for the 1:1m olar ratio, and two for the 2:1a nd 3:1m olar ratios. Interestingly,i ns imilarM ALDI-TOF experimentsc arriedo ut with 5:1[ V IV O(acac) 2 ]:apoHTF molar ratios the obtained average difference in mass between the samples of apoHTF and [V IV O(acac) 2 ]:apoHTF was 653(AE 60, considering the several spectra obtained for different samples of these solutions). This is consistent with the bindingt oa poHTF of for example, two By doing experiments( described in the experimental section) doing laser shots in separate but very close spots, one containing the apoHTF,t he other containing the solutiono f [V IV O(acac) 2 ], where the m/z peaks obtained were equal to those of apoHTF alone, the existence of false positives can be ruled out (or at least the probability of their existence considered extremely low). We cannott otally rule out the possibility of non-covalenti nteractions explaining the increase in mass. However,i ts hould be highlighted that the peak shapes (of m/ z = 1a nd 2) are both similar and sharp. If severald ifferent protein-complex interactions were established (as for non-covalent interactions), this would lead to broadeningo ft he peaks. [83]   To further confirm the binding of [V IV O(acac) 2 ]t ot ransferrin, solutionsc ontaining apoHTF (ca. 140 mm)a nd different amountsof[V IV O(acac) 2 ]w ere prepared, with molar ratios ranging from 1t o8 ,a nd eluted with at ris buffer solution through desaltingc olumns( PD-10 size exclusionc olumns, see experimental and SI sections). When using these columns the fractions first eluted contain apoHTF (and speciesb ound to it), while the small molecules not bound to the macromolecule are retained inside the pores of the packingo ft he column, and are only eluted after significantly larger volumes of buffer pass throughit. The eluate of the first eluted fractionswere analyzed by inductively coupled plasma-atomica bsorption spectroscopy (ICP-AES). The dilution effect from the elution ( % 67 %) was determined taking into account the absorbance of an apoHTF solution before anda fter elutingi t, in experiments with no addition of [V IV O(acac) 2 ]. Results of the ICP-AES analysisa re included in Ta ble 2. In this table, the similarly obtained values for HSA are also included (see section2.3.3).
From these experimentsi ti sc leart hat the eluted apoHTF fractionsc ontain boundv anadiumc omplexes.T he vanadium:apoHTF molarr atios determined in these fractions( 3 rd columno fT able 2) are lower than those in the corresponding solutionsa dded to the top of the size exclusion columns, the differences increasing with the increase in the initial [V IV O(acac) 2 ]:apoHTF molarr atios. In fact, for the 1.01:1 ratio the differenceb etween the ratios V:HTF in the added and eluted solutions is % 11 %, this meaning that the V IV O-complex is quite strongly bound to apoHTF.F or the 2.05:1 molar ratio, ca. 73 %o fv anadium remains bound to the protein, whilef or the 7.89:1 molar ratio only ca. 39 %o ft otal vanadium initially added remained bound to apoHTF.T his suggests that up to a[ V IV O(acac) 2 ]:apoHTF molar ratio of ca. 2t he vanadium complex binds quite tightly to the protein, while highera mounts of V IV O-acacspeciesdonot bind so strongly.Thus the equilibrium: Vanadiumb ound to apoHTF is, in proportion, more shifted to the right as the [V IV O(acac) 2 ]:apoHTF molar ratio of the solutionsp reparedi ncreases, this indicating that the "first" two moles of vanadium bind to distinct sites (thus forming distinct and more tightly boundc omplexes)f rom those that start being "occupied" only when having higher molar ratios.
In these experiments,w here the vanadium amount is determined by ICP-AES,n oi nformation may be obtained regarding which V IV O-containings pecies bind to apoHTF. Notwithstanding, the datao btained by mass spectroscopy are compatible with those obtained with the size exclusion columns: at otal of at least three V IV O-moieties might bind to apoHTF.T he mass spectrometric data suggests that when taking excess of [V IV O(acac) 2 ], two [V IV O(acac) + ]m oieties may be bound to apoHTF residues, fore xample, of the Fe binding sites, as well as one [V IV O(acac) 2 ]( more weakly bound). We cannotr ule out the possibility that the binding of VO-acacs pecies to apoHTF might be non-covalent, but the pattern of results obtained and the fact that almostn ob inding was found for HSA indicates that at least two VO-acacs peciesb ind to apoHTFi nvolving coordinativebonds.

Electron Paramagnetic Resonance
Our present CD data clearly confirmt hat V IV O-acac-apoHTF speciesd of orm in solutionsc ontaining [V IV O(acac) 2 ]a nd apoHTF,c ontradicting previousc onclusions [29,44] mainly based on EPR spectroscopicm easurements, where no V IV O-acac-apoHTF species were identified. In studies with HSA [29,38] and with Immunoglobulin G [43] it was also reported that only  Hamiltonian parameters calculated theoreticallyu sing DFT methods and considering either water or methanol as solvent both yieldedA z = 167.3 10 À4 cm À1 ,t hus giving good agreement with the experimental values obtained in this work or with those of Garribbaetal. [49] and otherauthors. [2,51] The EPR spectra of the solutions containing [V IV O(acac) 2 ]a nd apoHTFf rozen ca. 10 min. and 24 ha fter preparation are identical ( Figure SI-5-2). Figure SI and apoHTF (2:1 and4:1 molar ratios). Noteworthy is the observation that while the EPR spectra fors olutions containing [V IV O(acac) 2 ]:apoHTF molar ratios of 4:1( typically the conditions used in the experiments of Garribba and co-workers [29,44] ) are almost identical to those of [V IV O(acac) 2 ], that of the sample with 2:1m olar ratio differss ignificantly.T his is emphasized in Figure 8, where an amplification of the high field range of the same spectra is depicted. The species designated by D( with g z = 1.949;A z = 171.5 10 À4 cm À1 )d iffers from those of [V IV O(acac) 2 ]. It resembles the spin Hamiltonian parameters of the V IV O-apoHTF species B( Ta ble 3), butt he corresponding CD spectra differ significantly,t hus speciesDshould be assigned to aV IV O-acac-HTF complex.
The spin Hamiltonian parameters determined for solutionsc ontaining [V IV O(acac) 2 ]i nt he presenceo r absence of apoHTF are quite similar, but not identical;t hey are also not much different from those obtained for the V IV O-apoHTF system (Table 3). Regarding the identification of the V IV O-acac-apoHTF species that form, namely if either [V IV O(acac) 2 -apoHTF] or [V IV O(acac)-apoHTF] speciesform, it is not possible to indicate definite formulations from the EPR spectra measured. In fact, although frozen solution EPR spectra are an extremelyu seful tool to characterize and distinguish distinct V IV O-species that may be present in ac ertain medium, [84][85][86][87][88][89] EPR spectra of frozen solutions may not be at otally reliable guide for judging the molecular structure of V IV O-complexes, [51,84] particularly when several speciesm ay form which yield similars pin-Hamiltonian parameters, as is the case here (see also below).
The formation of [V IV O(acac)(apoHTF)] speciesi s akin of the formation of [V IV O(acac)(L)] compounds. Such type of complexes have been often reported, some of them with molecular structure characterized by single-crystal XRD and with data of frozen solution EPR spectra (see for example, Figure 9). In most of the complexes depicted in Figure9 the ligandsL have donor atoms resembling those available in HTF, namely in its iron bindingsites. Ta ble SI-8.5-1 includes EPR spectroscopy data for several distincts ystems containing V IV O-acac species, with data including either [V IV O(acac) 2 (L)] complexes (e.g. with L = Mepy or py) [51] or [V IV O(acac)(L)] (e.g. with L = bzpy-tch [6] (4) and sal-dmen (6), [92] Figure 9), where the spin Hamiltonian parameters are quite similar,s ome of them involving the acac À ligand with one of the O acac donor atoms bound cis to the O oxido .
Considering complexes [V IV O(acac) 2 (4-Phpy)] (3)a nd [V IV O(acac) 2 (py)] (9), while the DFT calculated energies of the trans-a nd cis-isomersi nM eOH solution are similar( the cisisomer is more stable than the trans-one by 3.3 and 7.5 kJ mol À1 ,r espectively), their corresponding theoretical A z values are also almosti dentical, that is,1 60.6 10 À4 (3-trans), 161.2 10 À4 cm À1 (9-trans), 159.5 10 À4 (3-cis)a nd 159.7 10 À4 cm À1 (9-cis). These values are lower by % 4% than those observed experimentally for the system bearing [V IV O(acac) 2 ] and pyridine (165.5 10 À4 cm À1 , [51] Ta ble SI-8.5-1), the latter being closert ot he A z value of the penta-coordinated complex [V IV O(acac) 2 ]( 165-168 10 À4 cm À1 ,T able 3). The calculations also suggest that the coordination of Phpy or py to [V IV O(acac) 2 ]i sn ot favorable thermodynamically,b eing both According to the data of Ta ble 2, at least three V IV Oc entres may bind to apoHTF.I ft wo bind to residues of the two iron binding sites (e.g. as the example of Figure 10 A), then it is plausible that at least one [V IV O(acac) 2 ]m ay be bound to apoHTF at aside group of an availableimidazole, amino or carboxylate moiety of HTF residues (see Figure 10 At this point it is also worth to mention that the immobilization of [V IV O(acac) 2 ]o nto solid supportsh as been reported [15,[95][96][97] and the binding established has been considered to involve: (i)hydrogen bonding between the pseudo p system of the acac À ligand and the silanolp rotons of the supports, and/or (ii)ligand-exchange with the formation of ac ovalent bond between the Vc entre and an O-atomf rom the support, (iii)interactionsb etween the complex vacant orbitals and the p electrons of the polymer benzene rings of polystyrene supports, [97] (iv) Schiff base formation by condensation between the carbonyl group of the acac À ligand and the free NH 2 groups previously grafted ontot he support'ss urface. [95] Namely,t he immobilizationo f[ V IV O(acac) 2 ]o nto silican anoparticles functionalized with 3-aminopropyltriethoxysilane (APTES) was reported to involveacovalentb ond of the N-aminoa tom with the V IV centre, as depicted in Figure 11. [15] This grafted complex was applied in the catalytic epoxidations and itc ould be recycled and reused four times, with similar catalytic activity and regioselectivity.T his meanst hat the bond established is strong enough for the complex to remainattached to the solid support.
Thus, whatever the correct formulation of these anchored V IV O-acacs peciest os olid supports, besides the V IV O-acac-apoHTF speciesd iscussed above (probably bound to some of the amino acid residues of Fe binding sites), the formation of [V IV O(acac) 2 (apoHTF)] complexes most probablya lso occurs, but correspond to weakeri nteractions. In theses pecies, besides the monodentate coordination of donora toms of available imidazole, amino or carboxylate moieties of HTF residues, probablys everal types of intermolecular interactions are also operating, for example, hydrogen-bond formation.
The formation of aS chiff base between -NH 2 of amino acid residues,f or example, al ysine, and acac ligands probablyd oes not occur,a sn oe lectronic transitions due to imine bonds were detected in the range 300-400 nm;a dditionally,t he Gibbs free energy of the reaction depicted in Scheme 1s uggests that the process might not be thermodynamically favored.    The calculations also indicatet hat model complex 3T is thermodynamically more stable than 1T,t he DG s value of the reaction 1T + T!3T + Ab eing À22.6 kJ mol À1 .T he stability of complex 2T relative to 1T or 3T cannot be estimated with any reasonable accuracy due to different overall charge of these species. Note that in 1T-3T the ligandsa re not linked with each other.
Ta king into account the rather small energy differenceo f variousg eometricali somers, the binding mode of [V IV O(acac) + ] to apoHTF is conceivably controlled by the secondary structure of the protein and intermolecular interactions with groups from the protein, rathert han by the thermodynamic stability of ap articular ligand configuration in the metal coordination sphere. Noteworthy is the findingt hat in all these simulated [V IV O(acac)(L1)(L2)(L3)] structures, those found more stable involve one of the O acac donor atoms bounda xially.I ts hould also be emphasized that hydrogen bonds to for example, the O oxido or O acac atoms mayc hange significantly the relative ener-gies andt he A z values corresponding to calculated 1T-3T model structures.
The calculatedh yperfine coupling constantsA z for the various isomers of 1T-3T are in the range of 153.6-161. 6 10 À4 cm À1 (Figure 12) and they are lower than the values found experimentally for the system [V IV O(acac) 2 ] + apoHTF (Table 3). However,i ns imilars tructures including one water molecule bound equatorially,i nstead of for example, aO -Tyr donor atom, the correspondingA z values increase by (3 to 6) 10 À4 cm À1 (Figure 13).
The calculated DH s values of the reactions:[ V IV O(acac) 2 ] + T + A + H!acac À + 1T (isomer 6) and VO(acac) 2 + 2T + H! acac À + 3T (isomer 4) are significantly positive (62.8 and 41.0 kJ mol À1 ,r espectively) indicating that the formation of complexes 1T and 3T from VO(acac) 2 is thermodynamically unfavorable. This again suggestst hat the binding of [VO(acac) + ] to apoHTF should be stabilized also by intermolecularH -bonding rather than by simple coordination of T, Aa nd Ht ot he metal. These effects are not taken into account in the calculations corresponding to Figures 12 and 13, neither possible effects from rearrangemento ft he apoHTFc onformation upon binding of VO-acac species. In the next section these effects are somewhat taken into account.  [98,99] and were checked against B3P86/6-311g DFT geometries. [100] The N-lobe of HTF was modelled using (a) the closed conformation (PDB ID 1a8e), [101] (b) the oxalate-bound conformation (PDB ID 1ryo) [102] (3), whichi sn ot as tightly packed aroundt he iron site as the closed conformation,a sw ell as using (c) the open conformation (PDB ID 1bp5); [103] in the latter,t he Vatom was located at the geometric centreo ft he iron-coordinating residues (see experimental and SI sections).

Modeling of the binding of VO-acacs pecies to HTF
No meaningful structures of V IV O-acacc omplexes with the closed conformation of HTF (after removing the Fe III and carbonatei ons) were found due to the small volumea vailablet o bind these complexes. For the other protein forms,t he most energetically favourable protein complexes were formed with the [V IV O(acac)(H 2 O) 2 ]s pecies. For the association of all V IV O complexes,a fter re-optimization of protein conformation, computed heats of formationw ith the open form of the protein is always energetically preferred when compared with the association with the oxalate-bound form. Ta ble SI-10-1 includes the computed heats of formation fort he various systems before carrying out the re-optimization of protein conformation. Table  SI- . In each structure refinedt he VO-acacc omplexes are tightly held in place by al arge range of interactions, including water bridges, hydrogen bondsa nd metal cation-p interactions.A dditionally,t he processesw here the lower energies were obtainedg ave rise to at ypeo fb ind-ing of vanadium to at yrosine residuew hich was not anticipated. In fact, in the structures refined, apart from the V = Ob ond lengthso fc a. 1.6-1.7 ,a nd the bindingt ot wo O-acac atoms at ca. 1.7-1.8 ,t he V IV is bound to the aromatic ring of at yrosine, with V-Cd istances in the 2.30-2.40 range, resultingi n half-sandwich complexes; this type of binding has been found for vanadocene(IV)-type compounds [105,106] (see more detailsi n the SI section, namely FiguresSI-10-1,-2a nd -3). Coordinating residues,a nd other relevant residues in the vicinity of the ligands,i nvolve the Fe-binding residues,A sp63, Tyr95, Tyr188, and His249, buta lso other protein residues, in particular Lys206,S er125, Ala126 andP ro247, also participate in the ligand-protein interaction, which is in agreementw ith previous results on the VO interaction with the Nl obe of hTF in the presence of carbonate. [104] However,t he A z values calculated for the model structures, upon freezingi ts coordinates similarly to what was done in [104]a re rather low,c a. 141 10 À4 cm À1 , so we rule out this type of binding of V IV O-acac species.
The structure (model A, Figure SI

Fluorometric assays
Intrinsic protein fluorescencei sm ainly duet ot ryptophan and tyrosine residues.F luorescence assays have been used to monitor the binding of several compounds to proteins,n amely to HTF and HSA. The study of the binding of [V IV O(acac) 2 ]t oB SA by fluorometric measurements was reported before, [39] butn ot to HSA.
Tryptophan residues are the least common amino acids in proteins but normally dominate their fluorescencep roper- ties, [107][108][109] and are the most commonly used intrinsic fluorophores.T he fluorescencee mission from Trpr esidues is very sensitivet oc hanges in local environment [110] and this sensitivity has been used extensively to monitorn umerous biological processes.H owever, often it is not possible to pinpoint the precise causes of changes in the fluorescencey ield, thusl imiting the usefulness of the fluorometric measurements. If there is more than one Trpi naprotein,a si st he case of HTF,f urther complications arise when trying to interpret the changes in fluorescencea tt he molecularl evel. However,i tw as reported that emission from the N-lobe of HTF is dominated by Trp264. [108] Except one study with vanadocened ichloride, [111] to our knowledge fluorometric techniques were not previously reported for studies of binding of vanadium compounds to HTF. The binding of V IV O-complexes to apoHTF,n amely at the iron binding sites, leads to the presence of the V IV O-speciesn ot far from aT rp residue at each site;m oreover,t he hydrodynamic volumeo ft he protein may change. Both thesee ffects may change the fluorescenceintensity. Figure 14 depictsf luorescencee mission spectra of the apoHTF-[V IV O(acac) 2 ]s ystem, when using the excitation wavelength (l ex )o f2 95 nm. ApoHTF demonstrates strong fluorescence emissionw ithamaximum at % 322 nm. As the complex concentration increases the HTF fluorescenced ecreases;t hus, the fluorescenceq uenching is concentration-dependent and apparently [V IV O(acac) 2 ]b inds closee nough to the tryptophan residues,n amely Trp264,t oq uencht heir fluorescence. Under our experimental conditions, no fluorescencee mission in the range 295-550 nm was displayed for the studied compounds and therefore there was not any interference with the fluorescence of apoHTF.
It is commont ou se fluorescenceq uenching measurements to evaluateb inding constantso fc ompounds to proteins.I n the case of organic compounds, the conditions required for such methods to be reliable for use for this purpose have been discussed. [112,113] If these conditions are nor fulfilled, fluorescence quenching measurements should not be applied to evaluateb inding constants of compounds to proteins, but in practicem any studies are published without verifying the validity of the procedures.I nt he case of metal complexes there are furtherr equirements that will be discussed below.
The methodology typically used in many publications was used in this work and is described in the SI section( SI-6); namely the fluorescence quenching measurements were made with ac oncentration of protein (apoHTF) of ca.1 0 À6 m,a nd varying the metal complex concentrationf rom ca. ca. 10 À6 to ca. 10 À5 m (see Figure 14).
Upon applying the Stern-Volmer equationt he corresponding quenching constant K SV was obtained ( Table 4). Following the typical calculation methodologies used in the literature, the quenching mechanism was considered to be static (due to bindingo faVO-acacc omplex to apoHTF) and the binding constant K BC given by equation4 ,a nd number of binding sites (n) were determined (Table4).
n ½V IV OðacacÞ 2 þ protein Ðf ½ V IV OðacacÞ 2 g n -protein ð4Þ When excitation was made at 280 nm (for which the other protein fluorophores, Phe and Tyr, may also be excited) ah igherq uenching %w as obtained (45 %), as well as ah igher K SV constant (3.9 10 4 m À1 ). The calculated binding constant was: K BC = 1.4 10 4 m À1 .
Considering the procedure described above and in the SI section (SI-6), from the fluorescencem easurements we would conclude that [V IV O(acac) 2 ]i sa ble to bind apoHTF,a lthough the quenching of the fluorescence is only moderate. Around 30 %q uenching of the Trpf luorescencei so bserved, the K SV being 1.9 10 4 m À1 ,t he binding constantK BC estimated in these measurementsb eing 1.0 10 4 m À1 .
Comments to the use of fluorescence quenching measurements to calculations of binding constants. The fluorescencea nd its quenching is ar ather indirect measurement of the interaction of [V IV O(acac) 2 ]w ith apoHTF,a st he Trpr esidues may not be close to the binding site responsible for the quenching effect, and there might be bindinga ts ites which do not affect the fluorescencee mission. Moreover,i nt he case of metal complexes, besidest he requirementsd iscussed in [112,113] for the validity of use of fluorescencee mission to calculate binding constants, there are further aspects/issuesr elatedt ot he speciation of the systems at low concentrations of labile metal com-   (Table 4), it is clear in Figure 15  It is thus concluded that the methodology used above leading to the values presented in Ta ble 4i sn ot valid to determinate the binding constantso f[ V IV O(acac) 2 ]t oa poHTF or to HSA. This conclusion may probably be equally appliedt omany other labile (and hydrolysable) metal complex-protein systems previously reported with binding constantsd etermined by fluorescence quenching measurements. We do highlight that researchers should carefully check which speciesm ay be presenta tl ow protein and metal concentrationsb efore applying these methodologies.
Even if ab inding constant K BC of 10 12 is used for this system, possibly much higher than would be reasonable to expect, the speciation obtained( see Figure SI As mentioned above, the binding of [VO(acac) 2 ]t oB SA was studied by ITC and fluorescence spectroscopy. [2,39] Namely,i t was reported that a1:1 adduct is formed between [V IV O(acac) 2 ] and bovine serum albumin (BSA), with dissociation constants (K d )o f2.6 10 À7 (ITC) and 6.1 10 À7 (fluorescence), [39] these correspondt ob inding constants (1/K d )o f3 .8 10 6 and 1.6 10 6 , respectively;t his was considered an important factor for its biological activity in vivo, particularly its insulin-enhancing effect. [2] The binding of V IV O 2 + to human serum albumin wass tudied by several techniques, namely by EPR and ENDOR spectroscopies, [27,38,44,89,115,116] CD and visible absorption. [116] The results show that V IV Oo ccupies two types of binding sites in HSA;i n one of the sites the resulting V IV O-HSA complex hasaweak CD signali nt he visible and its EPR spectrum may be easily measured;t his was assignedt oa mino acid side chains of the amino terminal site, Asp-Ala-His-, known as ATCUN site. [116] The other bindings ite depicts stronger signals in the CD in the visible range, but has ah ardly measurable EPR spectrum;i tw as assigned as involving residues of the multi metal binding site (MBS) of HSA. [116] Studies with fatted and defatted albumin showed [116] that the bindingo ff atty acids decrease the ability of V IV Ot ob ind albumin. Figure 16 depicts CD spectra measured with solutions containing[ V IV O(acac) 2 ]a nd defated HSA in the visibler ange. It is not cleari fDe ¼ 6 0a re obtained at all;i fy es, the bands are extremely weak. Even using higher [V IV O(acac) 2 ]:HSA ratios no clear bands were recorded. Zn II -complexes do not absorb radi- ation in the 400-1000 nm range and do not have unpaired electrons;t herefore they are both CD and EPR silent. The Zn 2 + ions bind strongly to the MBS binding site of HSA, so if any V IV O-acac-HSA speciesb inds at this site, if aZ n II salt is added they will be removed from it. [116][117][118] Upon addition of ZnCl 2 to as olution of [V IV O(acac) 2 ]a nd HSA with am olar ratio of 2, the very weak bands apparently observed in Figure 16 in the 700-900 nm range are no longer seen;t his could suggest that Zn 2 + ions are substituting V IV O-acac species bound at the MBS site; however, in all CD spectra the signal-to-noise ratio is so low that no definitec onclusion can be made. The Globally,w em ay conclude that the EPR and CD spectroscopic data do not clearly rule out or confirm the presence of V IV O-acacs pecies bound to HSA;i ft he bindingt akes place, probablyt hese correspond to [V IV O(acac) 2 (HSA)] complexes involvingm onodentate coordination of an N-amino, N-imidazole or O-carboxylate from amino acid residues of HSA, otherwise much stronger CD spectra would be measured, as was the case of V IV O-maltolato complexes. [117] 2.3.2. Use of size exclusion columns. The binding of [V IV O(acac) 2 ]t oH SA is weak, thusi tm ight happent hat part of the [V IV O(acac) 2 ]i nitially bound to HSA was lost during elution throught he size exclusion column,b ut clearly this data indicates that in solutions containing [V IV O(acac) 2 ]:HSA molarr atios of 5, not more (most probably less) than one V IV O-(acac) moiety is bound to each protein molecule.

Fluorometry experiments
While BSA has two Trpr esidues, HSA has only one. The intrinsic fluorescenceo fH SA is mainly due to the Trp214 residue, because of the very low quantum yield of the Phe and Tyrr esidues. [119,120] When the excitation wavelength( l ex )o fH SA is selected at 280 nm both Trpa nd Tyrr esidues contribute to the fluorescencee missions. However,a tl ex = 295 nm, the emission observed is only due to Trp214 which displays astrong fluorescence emission peak with l max % 330 nm. [119] Under our experimental conditions, no fluorescencee mission in the range 295-550 nm was displayed for the studied compounds and therefore there was not any interference with the Trp214 fluorescence of HSA. Figure 17    We followed as imilarm ethodology to the one appliedi n the [V IV O(acac) 2 ]-apoHTF system,t oo btain the Stern-Volmer constanta nd ab inding constant of [V IV O(acac) 2 ]t oHSA (see SIsection), and the results are showni nT able4.H owever,a se xplained above, this methodology is not reliable or valid to obtain the bindingc onstanti nt his system (and in many other systemsr eported in literature), and we include this information in this text to highlight this fact.

Humancells studies:intracellular distribution of [V IV O(acac) 2 ]
The cytotoxic activity of [V IV O(acac) 2 ]w as evaluated in the A2780 ovarian cells within the concentration range 0.1-100 mm using the colorimetric MTT assay.T he complex presentedm oderate cytotoxic activity with an IC 50 = 66 AE 18 mm,a fter 24 hi ncubation with the cells ( Figure SI-12-1).
The IC 50 value (66 mm)w as the concentrationo f[ V IV O(acac) 2 ] selected to carry out intracellulard istribution studies. As depicted in Figure 18 more than 60 %o fc omplex is retainedi n the membrane, where large complexes of proteins act to carry out vital cellular processes;t his correspondingt oc a. 18.5 ng of V/ millionA 2780 cells. Only as mall percentage was retained in the nucleus, an interesting result taking into consideration the complexr emarkable nuclease activity of [V IV O(acac) 2 ]u sing plasmids (naked DNA). [20,21] Globally the total amount of vanadium analyzed by ICP-MS in the A2780 ovarian cells after 24 ho fi ncubation with [V IV O(acac) 2 ]i sl ow (about 30 %o rl ess of those of Cu and Zn amountsf ound in the same study,c arried out with other types of complexes), [121] and more than 60 %i sl ocalized on its membrane.T he natureo ft he V-species present either in the membrane or in each of the cells' compartmentsi sn ot known, and probablyt he amount found in the nucleusi st oo low to affect it.
The cellsw ere culturedi nR PMI supplemented with 10 % fetal bovine serum (FBS), and exposed for 24 ht ot he complex. Fetal serum albumin, is one of the major components of FBS, and neither this protein nor HSA have specific binding sites for vanadium, thus we would expect rather similar transport capacity.T hus, this study suggests( at least fort he A2780o varian cells), that albumins are not efficient transporters of [V IV O(acac) 2 ], for its uptake inside the cells. However,t he amount of vanadium analyzed in the membranes is enough to affect significantly their metabolism. It is knownt hat for example, membranes contain several types of relevant proteins,a nd the binding of V-species may change/inhibit their biological function.

Conclusions
[V IV O(acac) 2 ]( 1), is one of the most remarkable vanadiumc ompounds and found uses in multiple areas. Namely it hasb een suggested as ap rospective drug for the treatment of diabetes, cancer and in tumor diagnosis. The understanding of how it is transported in blood is of major importance to establish its pharmacokinetics and pharmacodynamics.
Previous reports addressing the interactions between [V IV O(acac) 2 ]a nd plasma proteins are contradictory, [2,29,[43][44][45] but have agreed in the conclusion that [V IV O(acac) 2 ]d oes not bind to human serum apotransferrin. In this work we reportc ircular dichroism and MALDI-TOF data that clearly confirmst hat V IV Oacac speciesb ind to human serum apo-transferrin. The EPR spectra measured, SAXS data and DFT calculations corroborate the plausibility of this conclusion.
Analysis of apoHTF and vanadium contentso fs amples containing [V IV O(acac) 2 ]a nd apoHTF upon elution through size exclusion columns confirmt hat up to three V IV O-acac species may bind to apoHTF.S imilars tudies carriedo ut for the [V IV O(acac) 2 ]-HSA system did not confirm or rule out binding to albumin,b ut the interactions of [V IV O(acac) 2 ]a re much weaker with HSA than with apoHTF.F or example, when using [V IV O(acac) 2 ]:HSA ratios of 5, significantly less than one vanadium containing complexb inds to HSA.
Previous reports [29,[43][44][45] on the non-binding of [V IV O(acac) 2 ] to apoHTF explained by the stability of [V IV O(acac) 2 ], or the absence of water molecules coordinated in the equatorial position to be replaced by donorg roups of amino acid residues, or inability of this binding, are contradicted by the numerous previous reports of this type of binding in solution,b yt he several Circular dichroism spectroscopy. CD spectra were recorded on aJ asco J-720 spectropolarimeter either using the usual photomultiplier (200-800 nm range) or with ar ed-sensitive photomultiplier (EXWL-308) in the 400-1000 nm range. The measurements were normally carried out at % 25 8Cu sing either a2 0mm( chamber volume % 5.0 mL) absorption cells from Hellma ,o raM acro Suprasil 3.5 mm I.D. 50 mm cylindrical quartz cell (Jasco Parts Center, CQ3-50, chamber volume % 800 mL), [78] or a2mm cell (chamber volume % 600 mL) made of Quartz Suprasil from Hellma Analytics. Electron paramagnetic resonance (EPR) spectroscopy. The first derivative X-band EPR spectra of the frozen solutions (frozen in liquid nitrogen) were recorded on aB ruker ESP 300E spectrometer at 77 K. The ESP 300E spectrometer was operated at % 9.51 GHz with af requency modulation of 100 KHz. The samples (250 mL) were placed in 3mmq uartz tubes (Wilmad 707-SQ-250M) and frozen in liquid nitrogen. The calibration of the magnetic field frequency was done using DPPH (2,2-diphenyl-1-picrylhydrazyl) as standard. While keeping the resolution at 4096 points, the microwave power was adjusted to 20 dB and the receiver gain was set between 4.5 10 4 and 7.9 10 4 .T oi mprove the signal-to-noise ratio 5t o1 0scans of each sample were accumulated.
Fluorescence spectroscopy. The fluorescence spectra were obtained on aF luorolog Horiba Jobin Yvon Spectrofluorometer with single proton counting controller-FluoroHub from Horiba Scientific . The samples were excited either at the wavelength of 280 nm (slit width 5nm) and the emission spectra were measured in the 290-500 nm range, or at 295 nm (slit width 9nm) with the emission spectra measured in the 305-515 nm range. All measurements were done in 10 mm fluorescence quartz cells from Hellma Analytics with the chamber volume of % 3.50 mL at room temperature.
Electrospray mass spectrometry (ESI-MS). A5 00-MS Varian Ion Trap Mass Spectrometer was used to measure ESI-MS spectra of solutions in both positive and negative modes. The mass spectrometer was operated in the ESI negative or positive ion mode, typically with the following optimized parameters:i on spray voltage, À4.5 kV;c apillary voltage, À20 V; tube lens offset, À124.99 V, sheath gas (N 2 ), 20 arbitrary units (negative mode);i on spray voltage, 5kV; capillary voltage, 5V;t ube lens offset, 63 V, sheath gas (N 2 ), 20 arbitrary units (positive mode);c apillary temperature, 275 8C. Spectra typically correspond to the average of 20-35 scans, recorded in the range between 100-600 Da.
Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF). Mass spectra were obtained using aB ruker Daltonics Ultraflex MALDI TOF/TOF Mass Spectrometer operating in linear mode with positive ion extracting at 25 000 Va nd ap ulsed ion extraction of 480 ns. Each final spectrum was the accumulated result of at least 1000 laser shots that were obtained from 10 different manually selected regions of the same sample, over ar ange of 14 000-100 000 Da. Prior to calibration, the spectra were processed with Compass 1.3 using smoothing and baseline subtraction for reproducible peak annotation. The spectra were externally calibrated using of 50 pmol of albumin from bovine serum ([M+ +H] + 66,430).

Sample preparation
Human serum apo-transferrin (apoHTF), from Sigma-Aldrich, was dissolved (20 mm Tris-HCl pH 8, 20 mm sodium carbonate and 200 mm sodium chloride) to ac oncentration of 6mgmL À1 .Afirst sample of native apoHTF (500 mL) was then passed through aP D-10 MiniTrap G-25 column according to the previously reported pro-cedure. [26] As econd sample of apoHTF (500 mL) was incubated with a30times molar equivalent excess of [V IV O(acac) 2 ]a troom temperature;a fter one hour of incubation, the excess of unbound ligand was removed using PD-10 MiniTrap G-25 column. In both samples, af inal protein concentration of 3mgmL À1 was obtained for a1 :2 diluted sample.

Data collection and processing
The two samples, native apoHTF and apoHTF-[V IV O(acac) 2 ], were analyzed by SAXS immediately after passing through PD-10 Mini-Trap G-25 columns. SAXS data were collected at beamline BM29 (ESRF,G renoble, France) using ar obotic sample changer. [128,129] Te n frames of one second each were collected and all the measurements were done at 277 K. Different protein concentration ranges were used:3to 0.19 mg mL À1 for both samples. Data were reduced and analyzed using Scatter (Diamond Light Source, UK) and the ATSAS suite. [80] Each experimental frame was inspected for radiation damage;t hese frames were removed from further consideration and not used for buffer subtraction. Theoretical SAXS curves were calculated using CRYSOL. [82] Real space inversion obtain the P(r) function (pair distance distribution function) was carried out in Scatter.T he P(r) function is the Fourier transform of the scattering pattern, and represents ah istogram distances between all possible pairs of atoms in as tructure. An ensemble of seven low resolution envelope models were generated from the P(r) function in DAMMIF, [130] and then subject to averaging and filtering with DAM-AVER and DAMFILT. [131] The output from DAMFILTw as then passed to DAMMIN as as tart model for final refinement against the experimental curve. Other details are given in the SI section (SI-9). The absorption spectrum of each of these solutions was measured in the range 380-900 nm, the measurement starting always after the same period of time upon preparing each solu-tion. Samples were also similarly prepared with lactic acid, but no buffer was added and the pH was adjusted to 6.0 with solutions of HCl or NaOH. Both visible absorption and CD spectra were recorded (section SI-8.4).
Additional details of preparation of solutions are given in the SI section.

Mass and EPR spectra of solutions of [V IV O(acac) 2 ]and monodentate ligands
As tock solution of [V IV O(acac) 2 ]( 80 mm)w as prepared in methanol (previously degassed with N 2 ). Stock solutions of several monodentate ligands (imidazole, methylimidazole, benzimidazole, pyrazole and phenol) were prepared in 10 mm NH 4 CH 3 COO aqueous solution (previously set to pH 6.5). All these solutions were degassed with N 2 .A ccurately measured volumes of each solution of monodentate ligand and of the NH 4 CH 3 COO aqueous solution were added to six distinct vials using micropipettes, and an accurately For each of the six solutions, ESI-MS (+ /-) was carried out ca. 5m inutes after the addition of [V IV O(acac) 2 ]. Samples of these solutions were also introduced in tubes (for EPR spectroscopic measurements) and immediately frozen in liquid N 2 .

Maldi-TOF spectrometric measurements
Sample preparation. [V IV O(acac) 2 ]s tock solutions were prepared in MeOH with ca. 3mm immediately before mixing. Solutions of apoHTF from PROSPEC were prepared with ca. 300 mm by dissolving the protein in NH 4 HCO 3 buffer (pH 7.4, 25 mm). These solutions were allowed to stand overnight to allow equilibration.

Samples
for MALDI-TOF MS were prepared with [V IV O(acac) 2 ]:apoHTF molar ratios of 0:1, 1.1, 1:2, 3:1a nd 5:1b y mixing different volumes of the stock solutions with buffer.T he % of organic solvent was < 8% (v/v) and the apoHTF final concentration was either 50 mm (two sets of experiments, or 100 mm (one set with = 0:1, 1.1, 2:1a nd 3:1[ V IV O(acac) 2 ]:apoHTF molar ratios). The dried Droplet preparation procedure was used: Dried Droplet preparation. 2 mLo fe ach sample was mixed with 2 mLo fm atrix solution (saturated solution of sinapinic acid in 1mL of 30 %( v/v) acetonitrile and 0.1 %( v/v) aqueous trifluoroacetic acid). Then 1 mLo ft he sample-matrix solution was deposited by duplicate onto aM TP 384 ground steel BC target and allowed to dry at room temperature. The apoHTF concentration of the final samples was thus 25 mm. In one set of experiments two spots were deposited, very close to each other but without contact (at less than ca. 0.5 mm): one containing the apoHTF,t he other containing as olution of [V IV O(acac) 2 ] (see also the SI section). The laser shots were done in az igzag fashion so that both spots were included. In this way it was possible to check if the formation of [V IV O(acac) 2 ]:apoHTF species could be formed in the gas phase, thus giving false positives. The m/z peaks obtained in these experiments were equal to those of apoHTF alone, therefore the existence of false positives can be ruled out (or at least the probability of their existence extremely low).

Humancells studies;intracellular distribution of [V IV O(acac) 2 ].
The intracellular distribution of [V IV O(acac) 2 ]w as evaluated in the human A2780 ovarian cancer cells. For the assays cells (~1 10 6 / 5mLm edium) were cultured in RPMI supplemented with 10 % fetal bovine serum (FBS) and exposed to the complex for 24 h (37 8C, 5% CO 2 and humidified atmosphere) at ac oncentration equivalent to the IC 50 .T he IC 50 was obtained from ad ose-response curve using the MTT cytotoxicity assay.T he culture conditions and procedures were similar to previously described methods. [132] After incubation cells were washed with PBS and centrifuged to obtain ap ellet;t he subcellular soluble protein fractions i.e.,c ytosol (proteins from cytoplasm), membrane/particulate (membrane proteins including organelles), nuclear fraction (nuclear proteins, including the nuclear membrane proteins) and the cytoskeletal fraction (insoluble proteins and genomic DNA) were extracted using ac ell fractionation kit FractionPREP (BioVision) following the manufacturer's recommended procedures.
Each cellular fraction was digested with 0.5 mL of distilled conc. HNO 3 in ac losed pressurized microwave digestion unit (Mars5, CEM) for 12 ha t1 50 8Ci nH P500 vessels and then diluted in ultrapure water to obtain a2.0 %(v/v) acid solution. The vanadium content was measured using aT hermo XSERIES quadrupole ICP-MS instrument (Thermo Scientific). The instrument was calibrated using am ulti-element ICP-MS 71 Cs tandard solution (Inorganic Venture). Indium-115 (10 mgL À1 )w as used as the internal standard.

DFT calculations.
The full geometry optimization of the molecular structures was carried out at the DFT level of theory using B3P86 functional [133,134] with the help of the Gaussian 09 program package. [135] This functional was found to be appropriate for the theoretical studies of structural parameters and 51 VN MR chemical shifts of various V complexes. [92,93,136] The optimization was carried out taking into account solvent effects using the IEFPCM solvation model [137,138] with the UAKS molecular cavity and dispersion, repulsion and cavitation contributions. Water or methanol was used as solvent. No symmetry operations were applied for any of the structures calculated. The geometry optimization was carried out using ar elativistic Stuttgart pseudopotential which describes 10 core electrons and the appropriate contracted basis set (8s7p6d1f)/[6s5p3d1f] for the vanadium atom [139] and the 6-31G* basis set for other atoms. The Hessian matrix was calculated analytically for all optimized structures to prove the location of correct minima (no imaginary frequencies) and to estimate the thermodynamic parameters, the latter being calculated at 25 8C. The single point calculations at the IEFPCM-B3P86/6-311 + G** V(ECP)//6-31G* V(ECP) level of theory were then carried out. The basis set corrected enthalpies and Gibbs free energies in solution (H s and G s )d iscussed in the text, were calculated using the following equations: H s ¼ E s ð6-311 þ G**ÞÀE s ð6-31G*ÞþH s ð6-31G*Þð 5Þ G s ¼ E s ð6-311 þ G**ÞÀE s ð6-31G*ÞþG s ð6-31G*Þð 6Þ The 51 Vh yperfine coupling constants were estimated in gas phase at the single-point calculations using the BHandHLYP functional and 6-311 + G** basis set for all atoms on the basis of the equilibrium geometry obtained at the IEFPCM-B3P86/6-31G*(V-ECP) level of theory.T he anisotropic 51 Vh yperfine coupling constants A x ,A y , and A z were estimated as the sum of the isotropic Fermi contact term and corresponding dipolar hyperfine interaction term. [140] Wave functions were verified for their stability using the keyword STABLE in Gaussian. If necessary,t hey were reoptimized to achieve astable solution.