Effects of pH and Oxidants on the First Steps of Polydopamine Formation: A Thermodynamic Approach

We present a general thermodynamic top-down analysis of the effects of oxidants and pH on dopamine oxidation and cyclization, supplemented with UV–vis and electrochemical studies. The model is applicable to other catecholamines and various experimental conditions. The results show that the decisive physicochemical parameters in autoxidation are the pK values of the semiquinone and the amino group in the oxidized quinone. Addition of Ce(IV) or Fe(III) enhances dopamine oxidation in acidic media in aerobic and anaerobic conditions by the direct oxidation of dopamine and, in the presence of oxygen, also by the autoxidation of the formed semiquinone. At pH 4.5, the enhancement of the one-electron oxidation of dopamine explains the overall reaction enhancement, but at a lower pH, cyclization becomes rate-determining. Oxidation by Cu(II) at reasonable rates requires the presence of oxygen or chloride ions.


■ INTRODUCTION
Melanins are a group of ubiquitous natural pigments with diverse structures and biological functions, widely distributed in all kinds of organisms. 1 The most studied of them are the dark eumelanins found in the skin, hair, and eyes of animals, including humans. They are heterogeneous polymeric materials that are formed by the enzymatic or chemical oxidation of various nitrogen-containing natural phenols, mainly the amino acid tyrosine and the neurotransmitters dopamine (DA) and L-3,4-dihydroxyphenylalanine (L-dopa). In biological systems, all the roles of melanins are not known, but they act, for example, as pigments, photoprotectors, and antioxidants. 2,3 Eumelanins have received attention because of their versatile physical and chemical properties. 4,5 They are nontoxic, biodegradable, semiconducting materials obtained from natural sources or easily synthesized and chemically modified. They effectively bind many metals and scavenge free radicals and can turn practically all absorbed light into heat. A wide range of applications have been suggested for melanin-type materials, for example, drug release, 6 heavy-metal capture, 7 radical scavenging, 8 irradiation protection, 9 energy storage 10,11 and harvesting, 12 structural coloration, 13 biosensors and bioanalysis, 14,15 bioimaging, 16,17 and photothermal therapy. 18,19 A synthetic material called polydopamine (polyDOPA), possessing practically identical properties to those of eumelanin, can be obtained via the chemical oxidation of dopamine by dissolved oxygen in basic aqueous solutions. 20 Both polydopamine and eumelanin have complex and still debated structures, in which the major structural moieties are assumed to be catecholic or quinonoid 5,6-dihydroxyindole (DHI) units, their derivatives, and oligomers. 11,21−29 The structural components are bound together via covalent or noncovalent interactions; neither eumelanin nor polydopamine can be considered true polymers but represent a mixture of various oligomeric species with structural and redox disorder. 30 Although the structure of the final material is not known for certain, the initial steps leading to polydopamine formation are well established (Scheme 1). The autoxidation of dopamine (DA) by oxygen in a basic medium proceeds via dopamine semiquinone (DSQ) to dopaminequinone (DQ), which undergoes intramolecular cyclization to leucodopaminechrome (DAL). Further oxidation yields dopaminechrome (DAC), which can rearrange to 5,6-dihydroxyindole (DHI). Subsequent oxidation and coupling reactions then lead to the final insoluble product.
The synthesis conditions may influence the properties of the melanin-type materials. In cells, melanogenesis takes place enzymatically within a biological matrix, and a similar biomatrix has been used as a template for the spontaneous oxidation of DHI, too. 31 The autoxidation of dopamine by dissolved oxygen produces an aqueous suspension of nanoparticles and a thin globular film on immersed surfaces. 20 However, many applications would require the material mainly in the form of a thin film, and smoother melanin-type films have been prepared by dissolving nanoparticles in concentrated ammonia or by using oxidative layer-by-layer multilayers. 32,33 Oxygen produces highly reactive radicals, which can attack the material, and decomposition due to the splitting of the catechol moiety has been observed in natural and synthetic melanin. 34,35 The autoxidation of dopamine is carried out in basic solutions, which ensures the deprotonation of the amino group required for the intramolecular Michael addition. 36 On the other hand, autoxidative polydopamine formation has recently been reported even at pH 1.5 under hydrothermal conditions. 37 Oxidizing salts, like ammonium persulfate, sodium periodate, and sodium chlorate, can also be used for polydopamine synthesis. 38,39 In addition, several metal ions have been shown to catalyze or initiate dopamine oxidation. Redox-active transition metals, such as Ce(IV), 33 Cu(II), 40,41 Fe(III), 42 and Mn(III), 43 especially can be used for dopamine oxidation, in either the presence or absence of oxygen. This is a kind of biomimetic approach because many enzymes that catalyze the oxidation of phenols (e.g., tyrosinase and catechol oxidases) contain copper in their active sites. 44 It is noteworthy that oxidizing salts and transition-metal ions can also induce dopamine oxidation in weakly acidic solutions. 33,38,40−42 The effect of the oxidant on the properties of the polydopamine nanoparticles and thin films has been studied, especially with Cu(II) and oxidizing salts as the oxidants, and those products have been compared with the autoxidized products. 39−41 A lot of work on dopamine (or dopa) autoxidation or metalinduced oxidation can be found in literature, including thorough kinetic analysis, but most of it is within a rather limited neutral or mildly alkaline pH range. 29,35,45−49 Our previous work was based on using Ce(IV) as an oxidant in mildly acidic conditions (pH 4.5). 33 We were also interested in the reported unexpected formation of polydopamine at very low pHs. 37 Aiming at the production of good-quality melanintype films with oxidative multilayers, we are interested in using more benign biocompatible metals, such as iron or copper. Therefore, in line with the suggestions in the literature, we feel that a more thorough investigation of the effects of pH and oxidants on the dopamine oxidation, ring closure, and polydopamine formation is needed. 23,36 As the processes leading to polydopamine are extremely complicated, we limit our analysis only to the details of the initial steps in the overall pathway (Scheme 1) but in a wide pH range of 1−10 (in more alkaline solutions, quinones are not stable and participate in side reactions). Our approach is based on a thermodynamic analysis of the initial processes, upon which general models are built. The goal is a pH profile of the triggering events in the autoxidation reaction and a firmer understanding of the role of transition metals. The models devised can also be applied to other catecholamines under a variety of conditions. We supplement the theoretical modeling with experimental studies using UV−vis spectroscopy and electrochemistry for characterizing the initial steps in the process.

■ EXPERIMENTAL SECTION
Materials and Methods. Dopamine hydrochloride, cerium(IV) ammonium nitrate, and iron(III) chloride (all from Sigma-Aldrich); copper(II) sulfate pentahydrate (Merck); and catechol (Acros Organics) were used as received. All solutions were prepared in water distilled twice in quartz vessels. Numerical calculations were done using the Mathcad software (PTC).
pH Dependence of Dopamine Oxidation. Acidic dopamine solutions (0.1 mM) containing dissolved oxygen with or without metal ions (metal/dopamine ratio of 10:1) were titrated with 0.5 M NaOH. After the addition of each titrant, the UV−vis spectrum was recorded when the pH was stabilized (after a few minutes). For the long-duration oxidation studies, 0.1 mM dopamine solutions with or without metal ions (metal/dopamine ratio of 4:1) at different pHs were allowed to react in open vessels for 20 h at room temperature and then analyzed spectrophotometrically. Before the spectral measurements, the sample solutions were dispersed in an ultrasonic bath (USC500THD ultrasonic cleaner) when necessary. The pHs of the solutions were adjusted using 10 mM sulfuric acid, formate, acetate, phosphate, or citrate phosphate buffers (autoxidation).
Catechol and Dopamine Oxidation. Dopamine (0.1 mM) oxidation with dissolved oxygen was carried out in a stirred solution in a spectrophotometric cuvette at 50°C and pH 8.5 (45 mM Tris buffer), and the oxidation in the presence of metal ions was carried out at pH 4.5 (10 mM acetate buffer). The metal/dopamine ratios were 1:1 for Ce(IV) and Fe(III) and 6:1 for Cu(II). For the anaerobic oxidation experiments, the dopamine and metal-ion solutions were deaerated with nitrogen and added to cuvettes through septum caps. With Cu(II), oxidation was followed also in the presence and absence of 0.1 M NaCl. The oxidation of catechol was carried out similarly in 0.1 M acetate buffer, but the concentrations of catechol and the metal ions were 0.2 mM. All measurements were performed at room temperature (except those for the autoxidation of dopamine).
Characterization. UV−vis spectra were measured using an HP 8453 Diode Array UV−vis spectrophotometer equipped with a Peltier temperature controller. Electrochemical studies

Scheme 1. Initial Steps in Dopamine Autoxidation
The Journal of Physical Chemistry B Article were carried out with a computer-controlled potentiostat (Iviumstat, Ivium Technologies), a conventional threeelectrode cell at room temperature with a 1 mm diameter gold working electrode (Cypress Systems), and a platinum-wire auxiliary electrode. The gold electrode was electrochemically cleaned by cycling in ultrapure 0.5 M sulfuric acid until a stable voltammogram of pure gold was obtained. 50 The reference electrode was a miniature Ag/AgCl electrode, calibrated using a Ag/AgCl (3.5 M KCl) reference electrode, against which all experimental results are given. Its standard potential is +0.205 V above that of the normal hydrogen electrode (NHE), which is used as a reference in the theoretical results. All measurements were done at pH 4.5 (0.1 M acetate buffer containing 0.1 M NaNO 3 ). The concentrations of catechol and metal ions were 1 mM, and the catechol and metal-ion solutions were deaerated with nitrogen for 30 min and added to the cells under a constant nitrogen atmosphere. Cyclic voltammograms were recorded using a sweep rate of 10 mV/ s. In the alternating-current (ac) voltammograms, the potential scan rate was 5 mV/s, upon which an 11 Hz sinusoidal excitation with a 10 mV rms amplitude was added, and the rms ac-current magnitude was recorded.

■ RESULTS AND DISCUSSION
Dissolved Dioxygen as Oxidant. The spontaneous autoxidation of dopamine by dissolved oxygen, the most common path to polydopamine, serves as a reference system for our discussion. The autoxidation is carried out in basic aqueous media, often at slightly elevated temperatures to speed up the relatively slow process. The spectral evolution during the autoxidation of a 0.1 mM dopamine solution in a Tris pH 8.5 buffer at 50°C is shown in Figure S1 in the Supporting Information. The spectra show the formation of a transient species with an absorption maximum at ca. 480 nm, identified as dopaminechrome (DAC), which is formed from the cyclic leuco form (dopamineleucochrome, DAL) produced by intramolecular cyclization. 35 The further reactions of dopaminechrome to 5,6-dihydroxyindole (DHI) and indolequinone (IQ) are slow, which leads to its accumulation in the solution. 35,51 Dopaminechrome has another absorption band at 300 nm, which hides the possible signal due to the rapidly decaying dopamine semiquinone with a band at 305 nm. 51 From the beginning, absorption increases also at 300−350 nm. Dopaminequinone has an absorbance maximum at longer wavelength (395 nm), 35 we attribute these changes to the formation of the cyclized products, dopaminechrome and, especially, leucodopaminechrome. The leuco form has its absorption maximum at ca. 285 nm and a tail of higher absorptivity than that of dopamine. 52 Later in the reaction, the absorbance increases over the whole observed spectral range, and the spectra begin to resemble the characteristic featureless polydopamine−melanin spectrum, masking further detailed evolution. In this discussion, we concentrate only on those initial triggering steps leading to the formation of the first cyclized products over a large pH range.
There are two separate critical steps in the formation of dopaminechrome and its further-rearrangement products. The first is the initial oxidation of dopamine to dopaminesemiquinone and its subsequent oxidation to dopaminequinone. These reactions are basically reversible. The second step is the practically irreversible intramolecular cyclization to form leucodopaminechrome, which transforms to dopaminechrome. This irreversible step can drive the overall reaction even under unfavorable conditions. It is known that hydrogen peroxide is produced upon the oxidation of dopamine to dopaminequinone. 53 This suggests that in the beginning, the two-electron redox systems O 2 (aq)/ H 2 O 2 and dopaminequinone/dopamine are crucial to the overall redox reaction leading to dopaminequinone. Protons are involved in these redox reactions, and the speciation of the compounds depends on pH. The effect of solution pH can be taken into account by calculating the formal standard redox potentials using the reported redox potentials and the pK values for the species involved. 48,54,55 Although the formal twoelectron redox potentials of the redox pairs O 2 (aq)/H 2 O 2 and DQ/DQH 2 (DQ denotes the catechol/quinone moiety) would suggest that a two-electron oxidation of dopamine by the oxygen/hydrogen peroxide redox pair is thermodynamically favorable over the whole pH range (Figure 1), molecular oxygen is actually a poor oxidant. 56,57 The ground state of oxygen is a triplet state ( 3 Σ g − ), and its direct two-electron redox reactions with organic species, which are usually in singlet states, are spin-forbidden. 55,58 This creates a large activation barrier and makes the reactions slow. Therefore, the peroxide pathway must proceed through two one-electron steps with a superoxide radical anion, O 2 ·− , or a perhydroxyl radical, HO 2 · (pK HO 2 · = 4.8), 55 as an intermediate, depending on the pH. However, the standard redox potential of the aqueous O 2 , which also makes the one-electron reduction of oxygen thermodynamically unfavorable in most cases. However, the superoxide radical anion, O 2 ·− , and perhydroxyl radical, HO 2 · , are much stronger oxidants. 42 On the other hand, dopamine has an inverted order of the one-electron redox potentials, which is typical for quinones in aqueous media. 48 It should be noted here that care has to be taken when comparing the redox-potential values given in the literature as many of them refer to the biochemical standard state (pH 7). In addition, the strictly thermodynamic standard state for oxygen is an ideal gaseous substance at 1 bar, The Journal of Physical Chemistry B Article whereas a more appropriate standard state for reactions of dissolved oxygen refers to 1 M aqueous solutions (as here), leading to a value that is 0.15 V higher.
On the basis of kinetic modeling, it has been suggested that the rate-determining step in dopamine oxidation is the formation of monodeprotonated dopamine. 42 However, proton transfer to and from oxygen and nitrogen is very fast, and in the discussion of proton-coupled electron transfer, the protonation reactions are generally assumed to be in equilibrium. In accordance with this assumption, we discuss the two oneelectron redox processes required to oxidize dopamine to dopaminequinone and reduce oxygen to hydrogen peroxide, using the square-scheme model suggested by Laviron,59 which is the standard model used to describe proton-coupled redox processes. 60 Hydrogen peroxide is a weak acid (pK H 2 O 2 = 11.7); 55 dopamine and its semiquinone have several protonation states in the normal pH range, and both redox systems can be presented by incomplete 3 × 3 square schemes (Scheme 2). The total two-electron two-proton (2e − , 2H + ) redox processes of both systems can be described by two one-electron redox reactions, and their formal standard potentials are drawn in the form of a Pourbaix diagram in Figure 1 as a function of pH (the calculation details are in the Supporting Information). The curves have been calculated using the value pK DQH · = 4.7 for dopamine semiquinone 48,54 and the first macroscopic pK of dopamine for the deprotonation of the first catecholic hydroxyl (vide infra). The effect of the protonation state of the amino group is neglected.
In the following, the generic acronyms SO, HP, DHQ, and DSQ (Scheme 2) will be used for superoxide, hydrogen peroxide, dopaminehydroquinone, and dopaminesemiquinone species, respectively, irrespective of their protonation state (dopaminequinone, DQ, is not protonated in the normal pH range). It is evident from Figure 1 that the first one-electron step in the autoxidation of dopamine is always thermodynamically highly unfavorable, although it becomes less so at pH > 5.
In polydopamine formation, the second critical reaction step involves the cyclization of the nonprotonated amino side chain to form a 5-ring. This Michael N-addition requires that the catechol moiety be in the quinone form. Therefore, the reaction must produce dopaminequinone from the semiquinone form. After the initial oxidation to form superoxide and the semiquinone radicals, the reaction can be thought to proceed in several possible reversible paths, for example: The conditional equilibrium constants for each reaction can be calculated as a function of pH using the one-electron formal standard potentials in Figure 1 and are presented in Figure 2.
The first reaction is the same as the initial oxidation by molecular dioxygen, but when the reaction products accumu-   The Journal of Physical Chemistry B Article late, the backward reaction has to be considered too. The first and second reactions are thermodynamically unfavorable over the whole pH range. Reaction 3 is favorable because superoxide radicals are much better oxidizing agents than dioxygen, and the reaction between superoxide and semiquinone radicals (reaction 4) is highly favorable over the whole pH range. The last reaction (reaction 5) is the semiquinone disproportionation equilibrium, which always lies on the side of the products. Dopaminequinone is formed in reactions 2, 4, and 5, which probably all contribute to its formation during the reaction. However, initially, the concentrations of superoxide and dopaminesemiquinone radicals are very low because of the slowness of the initial oxidation reaction. Therefore, reaction 4, which is between SO and DSQ and is thermodynamically highly favorable, should be the most important one because the reactants, born in the initial oxidation reaction, are very close to each other, possibly still within the encounter complex. Later, when more DSQ and SO have been generated in the solution, the significance of dopamine oxidation by SO and the disproportionation increase. The disproportionation reaction can, in effect, produce an effect known as self-acceleration of hydroquinone oxidation because of the chemical feedback it provides. 53 The pH dependence of dopamine autoxidation and cyclization was studied experimentally using two approaches ( Figure 3). First, 0.1 mM dopamine solutions at different pHs were allowed to react in open vessels for 20 h and analyzed spectrophotometrically. A long reaction time was used in order to also see possible changes in acidic solutions, in which the reactions are slow. However, the long time scale means that the intermediate species may already have disappeared. Second, an acidic dopamine solution was titrated with concentrated NaOH, and the spectrum was recorded after each addition when the pH was stabilized. It should be noted that this titration method is not well-defined in a thermodynamic or kinetic sense, but we have previously used it successfully to study the Ce(IV)-induced oxidation of dopamine. 33 The time scale of titration is short (minutes), helping the observation of intermediate species, but each addition causes a local, transient pH rise, which can affect the reaction, especially at low pHs.
In the 20 h spectra, absorbance only slightly increases above pH 5 at ca. 300 nm, and after that, an increase is observed over a large wavelength range, with no clear band attributable to dopaminechrome. In the titration spectra, the first change below ca. pH 3 takes place also around 300 nm, and a separate weak band assigned to dopaminechrome forms after pH 3.5 at 480 nm. 35 On the other hand, dopaminequinone should have a band at 395 nm, which is not observed. 35 The absorption increase at ca. 300 nm can be assigned mostly to the formation of cyclic leucodopaminechrome, on the basis of the UVspectrum of the corresponding form of dopa. 47,52 However, we attribute these first observed changes in the titration spectra to local pH transients. Clear changes take place only above ca. pH 5, and the onset of spectral changes at 480 nm is shifted to pH 6−7, and rapid growth takes place at ca. pH 9. On the other hand, in the case of the long reaction time, all dopaminechrome has reacted further, and the spectra are obscured by other products (DHI and polydopamine), as is the case in the titration spectra at high pH, too. However, both results show that the effective onset pH for autoxidation is ca. pH 5, but the reaction rate increases markedly only at higher pHs. This is well in accordance with the calculated pH profiles for the process (vide infra).
In order to quantitatively discuss the pH dependence of dopamine autoxidation and intramolecular cyclization during the initial stages of the reaction, we describe the system by the following simple model (DAL = leucodopaminechrome, DAC = dopaminechrome): This model describes the triggering events in dopamine autoxidation, and we ignore the generation of dopaminequinone via disproportionation and other reactions that are important at later stages in the process. Here, k 1f , k 1b , k 2f , and k 2b are the average apparent one-electron-forward-and backward-charge-transfer rate constants for the first and second redox steps, respectively, and k c is the apparent first-order rate constant for the ring closure (vide infra). Considering the highly unfavorable thermodynamics of the reaction between oxygen and dopamine, the first one-electron transfer should initially determine the rate of dopamine autoxidation. In order to describe the pH dependence of the redox rates, we use the classical Marcus theory to estimate the redox rate constants (see the Supporting Information). 61 The electron-transfer rate constants are given by where ΔG 1,2 θ′ = −FΔE 1,2 θ′ . Here, ΔG 1,2 θ′ and ΔE 1,2 θ′ are the pHdependent conditional Gibbs-free-energy change and the formal potential difference, respectively, for redox steps 1 or 2; λ = λ is + λ os is the reorganization energy, given as the sum of the inner-and outer-sphere energies; and k app 0 is the maximum apparent rate constant including the equilibrium constant for encounter-complex formation. The backward-electron-transfer The Journal of Physical Chemistry B Article rates are calculated using the conditional equilibrium constants for the steps, as k 1,2;b (app) = K 1,2 ′ /k 1,2;f (app) . Unfortunately, no reorganization energies for the relevant reactions were found in the literature, and we used the values reported for the selfexchange reactions of O 2 ·− /O 2 and the p-benzoquinone species as rough approximations. 62,63 These yield a value of λ ∼ 1.5 eV, 61 which is really a crude estimate. However, it turns out that the value of the reorganization energy does not have a marked influence on the form of the pH profiles calculated for the process.
The next step in the initial pathway to polydopamine is intramolecular cyclization. It takes place as a Michael Naddition of the unprotonated amino group to the oxidized catechol ring, which is subsequently reduced to hydroquinone. Nitrogen with its lone-pair electrons attacks the positively charged C5 atom in dopaminequinone (Scheme 1). The C−Nbond formation is followed by the loss of two protons and the rearomatization of the catechol ring. The nitrogen electrons act as reducing agents in this process. The macroscopic-pK values for dopamine are pK 1 = 8.87 or 8.89, pK 2 = 10.63 or 10.41, and pK 3 = 13.1, from two independent sources. 64,65 However, because of the zwitterionic nature of dopamine, the protonation and deprotonation process should be described using the microscopic-protonation constants (Scheme S1 in the Supporting Information). The distribution diagram ( Figure S19) shows that the first macroscopic constant corresponds mainly to the deprotonation of the first hydroxyl group, and pK 3 clearly represents the deprotonation of the second hydroxyl group. Therefore, the macroscopic-pK 1 value has been used to draw the Pourbaix diagram for dopamine above. However, the macroscopic-pK 2 value does not correspond to the acidity constant, pK NH 3 , of the protonated amino group in the oxidized dopaminequinone, for which a value of pK NH 3 = 9.58 has been obtained using the dimethoxy derivative. 45 The dopaminequinone-intramolecular-cyclization reaction is relatively rapid (the first-order rate constant is k c 0 = 25.6 s −1 at 25°C), 45 which can be attributed to the facile formation of 5membered rings via an exo-trig reaction. 66 In all-carbon systems, this is due to favorable ring strain and entropy contributions at the transition-state geometry, and with catecholamines, the cyclization reaction is strongly favored by entropic factors, too. 45,67 The cyclization rate of catecholamines increases strongly with pH. The further oxidation of the cyclic leuco form to the aminochrome form is very fast with dopa (3,4dihydroxyphenylalanine), and the cyclization rate is identical to the observed rate of dopachrome formation. The pK NH 3 value in dopa is ca. 2 units lower than in dopamine, which we suggest to be due to intramolecular hydrogen bonding, producing a ca. 10-fold difference in the cyclization rates. The oxidation of the leuco form to the aminochrome form (dopachrome or dopaminechrome) by dopa(mine)quinone is very fast, but the oxidation by dissolved oxygen is spin-forbidden and rather slow. 49,51 However, because the cyclization reaction is practically irreversible, the further steps do not affect the modeling of the initial oxidation and ring-closure steps. The further rearrangement of dopaminechrome (DAC) to 5,6-dihydroxyindole (DHI) and its oxidation are relatively slow at physiological pH, leading to the accumulation of dopaminechrome under those conditions. 35,68 Several kinetic studies carried out in mildly basic solutions suggest that the rate of dopamine oxidation is inversely proportional to the hydrogen-ion concentration, which has been explained by assuming the deprotonation of the hydroxyl groups or the charged amino group to be the rate-determining step. 45 However, the proton-transfer kinetics of aliphatic amines is close to diffusion control in aqueous media. 69 For ethylamine, the protonation and deprotonation rates have been determined to be 1.4 × 10 7 s −1 and 3.2 × 10 10 M −1 s −1 , respectively. For aliphatic amines, the rates are almost constant, although steric hindrance can decrease the protonation rate. Applying steady-state approximation to the deprotonated form and using the protonation rate of ethylamine as an approximation for the amino group in dopaminequinone (see the Supporting Information), the cyclization rate can be expressed as [H ] where [DQNH 2 ] is the concentration of dopaminequinone with a deprotonated amino group, C DQ is the total concentration of dopaminequinone, and k c 0 is the first-order rate constant (estimated from the literature). 45 The general kinetic model above can be discussed using different levels of approximation. In this work, we focus on the pH dependence of the initial reactions that are required to trigger the process. Therefore, we make the simplifying assumption that the concentrations of dopamine and dissolved oxygen are constant; that is, [DHQ] = C Q and [O 2 (aq)] = C O 2 (≈ 0.25 mM at 25°C in air). This also approximates the real reaction conditions in an open vessel at a low pH (slow reaction) and with a high dopamine concentration. Under this assumption, the kinetic scheme allows an analytic solution (for details, see the Supporting Information). The unknown preexponential factors, k 1f 0 and k 2f 0 , in the redox rate constants were estimated using the reported electron-transfer rates at pH 7.4. 49 This is an approximation, too, because in the pH range considered here, differently protonated species participate in the reactions, and the rate coefficients may vary accordingly. The exact expressions are quite complicated, but they show that the DQ concentration remains low and is inversely proportional to the cyclization rate, decreasing with a pH-dependent time constant of τ = 1/k c . The DQ formation rate is given by and that of leucodopaminechrome by The rate of DAL formation increases with time, because the DQ concentration grows, and is zero in the beginning as no DQ is present. In order to compare the oxidation and cyclization kinetics in the early stages of the process, we looked at the initial DQ-formation rate (ξ DQ ) and the pH dependence of the DAL-formation rate at t ∼ 0 (ξ DAL ). The The Journal of Physical Chemistry B Article results, which can be understood as relative probabilities for triggering the processes at different pHs, are plotted in Figure 4 and calculated using a wide range of reorganization energies.
The value of the reorganization energy does not markedly affect the rates and, more importantly, the form of their pH profiles. The calculations predict that the DQ-formation rate will be constant but small below ca. pH 5 but will increase above it. Similarly, the DAL-formation rate accelerates above pH 5 because of the higher DQ concentration, but a significant increase is observed only at high pHs (ca. 8−9). Both results are in accordance with experimental data. There are, therefore, two physicochemical parameters that control the autoxidation process of dopamine and, evidently, other catecholamines, too. They are the pK value of the hydroxyl group in the semiquinone radical (here 4.7) and the pK of the protonated amino group (9.6) in the quinone form.
The model allows us to make some major predictions and interpretations of the autoxidation process. The cyclization rate increases with time but is very low below ca. pH 5. It levels off at high pHs as all amino groups become deprotonated. The dopaminequinone-formation rate depends exponentially on the cyclization rate, and at low pHs, where the cyclization can be neglected, a steady increase of dopaminequinone is observed. Under these conditions, the reaction slowly tends toward an equilibrium determined by the first two redox steps. Therefore, at low pHs, cyclization is the rate-determining step in the process, not dopaminequinone formation. However, if there existed a mechanism for the enhancement of cyclization, then polydopamine could be formed also in highly acidic media. We think that this explains the recently reported synthesis of polydopamine at very low pHs under hydrothermal conditions. 37 In aqueous media, acid dissociation is entropically favored, which lowers the pK value at higher temperatures. In addition, at high temperatures and pressures, the dielectric constant of water decreases, which profoundly affects all processes involving charged species, including protonation reactions. 70 Specifically, the formation of a dopamine zwitterion becomes less favorable. Assuming little effect on dopamine oxidation, dopaminequinone with an increased fraction of neutral amino groups is formed under these conditions, allowing cyclization to take place. In addition, the cyclization reaction is enhanced at high temperatures also because it is entropically favored. 45 The autoxidation model presented here is, of course, an oversimplification and should be applied only to the initial phases triggering the process. Later, disproportionation reactions and the role of superoxide radicals become important. In addition, the rate of the initial one-electron autoxidation of dopamine is very low; therefore, if semiquinone or superoxide radicals can be formed by some other pathway, the reaction can proceed more effectively, also at a low pH. This will be considered in the next section when discussing catalysis by metal ions.
Transition-Metal-Assisted Oxidation. Transition metals are known to catalyze autoxidation reactions and, in fact, metals bound to biomolecules or as traces in buffers have been suggested to be responsible for the observed autoxidation reactions in biological and experimental systems, respectively. 48,53,58 Detailed kinetic analysis of iron-and copperinduced oxidation of dopa and dopamine can be found in the literature. 29,42,48,49 In this work, we have studied dopamine oxidation and dopaminechrome formation under aerobic and anaerobic conditions in a pH range of 1−10 using three redox-active transition-metal ions, Ce(IV), Fe(III), and Cu(II). Under aerobic conditions, the general pH dependence of the reactions with different metal ions was tested using long reaction times and titration with NaOH, similar to that described above for the autoxidation reaction ( Figure 5). The previous results with Ce(IV) are similar to those obtained with Fe(III). 33 The early absorbance increase at 300 and 480 nm is attributed to the formation of leucodopaminechrome and dopaminechrome, respectively, whereas no direct evidence for dopaminequinone can be observed in the spectra. The observed pH shifts in the reaction onset, referring to the autoxidation reaction, are so large (2−6 pH units) that they require explanation in spite of the uncertainties involved.
The cyclization process is limited by the deprotonation of the amino group, and Figure 5 seems to imply shifts in its apparent pK value. The influence of a nearby charge on the pK value of an ionizable group is a well-known phenomenon in proteins. 71 Similarly, the positive metal ion bound to the catecholate moiety electrostatically affects the protonated amino group in the molecule. This interaction thermodynamically favors its deprotonation and decreases its apparent pK value. Simple thermodynamic arguments (see the Supporting Information) suggest that this would lead to a shift of only ΔpK ∼ −0.45 per unit positive charge, which is, in fact, well in accordance with the difference between the microscopic-protonation constants, pk 2 and pk 12 . However, metal complexation takes place with a catecholate dianion, which lowers the effective positive charge, and the effect on the amino group in dopaminequinone is questionable as a result of the poor complexing capacity of the quinone form. Therefore, electrostatic interactions must be ruled out as the cause of the shifts observed.
We can envision several possible reasons for the enhanced rate at low pHs. First, transition-metal ions may oxidize dopamine to semiquinone. Second, in the presence of oxygen, Third, complex formation with the catechol moiety may enhance either the oxidation kinetics, by alleviating the spin restrictions, 53 or the thermodynamics, by forming a noninnocent complex with a lower redox potential than that of dopamine. In addition, dopamine−metal−oxygen complexes can enhance the reaction. 58 The first alternative depends on the oxidation power of the metals. Dopamine is a strongly complexing ligand, and metals are predominantly attached to the catechol moiety, forming mono-, bis-, or tris-complexes. If the catechol moiety (Cat) is an innocent ligand, the redox properties of the ligand and the metal center can be treated separately. This should be the case at least with catecholates of the first-row transition metals (e.g., Fe, Cu), which can display valence tautomerism (intracomplex charge transfer) with localized charges because of reversible metal−ligand electron transfer. 72,73 Interestingly, the oxidation of catechol-terminated selfassembled monolayers has been reported to shift cathodically in solutions containing transition metals. 74 In order to understand the effects of the metal ions, we first studied the metal-assisted oxidation of catechol, in which case further reactions do not interfere, using low equimolar concentrations in order to slow down the process. The twoelectron redox potential of catechol (+0.32 V vs Ag/AgCl at pH 4.5) is very close to that of dopamine in the whole pH range, but the one-electron potentials (+0.12 and +0.77 V vs Ag/AgCl at pH 4.5) are cathodically and anodically shifted with respect to those of dopamine ( Figure S20). In both the presence and absence of oxygen, mixing equimolar amounts of catechol and Ce(IV) or Fe(III) solutions (at pH 4.5) results in the rapid formation of a band due to o-quinone (at ca. 390 nm), together with a decrease of absorbance of catechol and the metal. Using the reported molar absorbance of the quinone form at 390 nm (1370 M −1 cm −1 ), 75 we can estimate that approximately half of the catechol is oxidized to quinone almost immediately by Ce(IV), and practically all the metal oxidant is consumed. In principle, this can take place either via the one-electron oxidation of catechol molecules and the consumption of all the metal, followed by semiquinone disproportionation, or via successive one-electron oxidations of catechol by the metal. In our previous work with cerium-based oxidative multilayers, the experimental evidence supported the first pathway with the complete loss of the oxidant. 33 With Ce(IV), the o-quinone band decreases with time, and an isosbestic point can be seen at ca. 427 nm before the absorbance increases further above 400 nm, and a rather featureless spectrum is obtained. This further reaction can be attributed to polycatechol formation. 76 Qualitatively similar behavior is seen with Fe(III), but no isosbestic point is observed, and the formation of o-quinone is slower. The band due to the 1:1 Fe(III)−catechol complex forms immediately at 700 nm as approximately one-half of the catechol is complexed. 77 Concomitant with o-quinone formation, the Fe−catechol complex is broken down. In this case, the mechanism probably resembles a combination of the pathways discussed above. Cu(II) does not produce any reaction in the presence or absence of oxygen, and the final spectrum of the solution (after ca. 1 h) is identical to the sum of the component spectra.
Electrochemical experiments on the metal-catechol mixtures yielded interesting results (Table 1 and Figure 6). In the cyclic voltammograms the two-electron redox wave of catechol dominates in every case. The formal standard potential of the catechol/o-quinone pair obtained is +0.34 V vs Ag/AgCl, very close to the calculated value. The cyclic voltammograms display some smaller humps on both sides of the main peak, which

The Journal of Physical Chemistry B
Article cannot be accurately characterized (with Fe, there is also a small peak at ca. 0 V, which can be attributed to the Fe(III) ↔ Fe(II) transformation). Therefore, we have applied acvoltammetry, which is a much more sensitive method, to study these systems. In ac-voltammograms, a clearly defined redox wave can be seen on both sides of the central catechol/oquinone two-electron redox process ( Figure 6B). The formal standard potential of the first of these peaks does not depend on the metal (Fe or Ce; with Cu the possible peak is masked by the Cu(II)/Cu(I)/Cu(0) redox system) in solution, indicating that it is related to the catechol moiety. The spectral measurements showed the generation of o-quinone in the reaction mixture, which must be preceded by semiquinone formation. Therefore, we attribute this more cathodic peak to the one-electron redox process of the semiquinone/o-quinone pair because the observed potentials are very close to the calculated one. The other peaks on the anodic side of the twoelectron catechol/o-quinone wave are more problematic because their potential depends on the metal and is clearly more cathodic than expected for the catechol/semiquinone one-electron process. We tentatively suggest that these peaks correspond to the two-electron oxidation of metal-bound catechol. Spectra show that, in the case of Fe(III), approximately half of the catechol is complexed with metal under these conditions. The positive charge of the metal ion increases the oxidation potential, and its dependence on the metal may also imply some redox-state mixing (ligand noninnocence), especially with the Ce complex. Spectroscopic studies with dopamine at pH 4.5 were done similarly using low equimolar amounts of reactants to limit the contribution from follow-up reactions (see the Supporting Information for the spectra). We have focused on this mildly acidic pH value because it was also used in previous studies by us and others. 33,39−41 In addition, the pH-series study shows that at much lower pHs, no reaction takes place at a noticeable rate, whereas in solutions closer to neutral, autoxidation starts to influence the results. When dopamine is used instead of catechol, a rapid increase of a clear band at 480 nm is observed when equimolar amounts of dopamine and Ce(IV) or Fe(III) are mixed, under both aerobic and anaerobic conditions, followed by a slower decrease, forming a trace typical for an intermediate (Figure 7). No clear spectral band assignable to dopaminequinone could be observed, implying a rapid formation of the cyclized (and tautomerized) form dopaminechrome, which then reacts further. Absorbance slowly increases above ca. 550 nm, for example, at 620 nm, which corresponds to the absorbance maximum of indolequinone. 33 The reported molar-absorptivity value of dopaminechrome (3280 M −1 cm −1 ) 48 allows us to estimate that with Ce(IV), most of the oxidant is consumed. With Cu(II), no reaction took place at this pH without oxygen.
In order to justify the observations, we have constructed the Pourbaix diagrams of the metals, together with dopamine (and oxygen), as shown in Figure 8    The Journal of Physical Chemistry B Article speciation strongly depends on the pH and complexing ligands.
In addition, poorly soluble species can be formed, and a different diagram has to be drawn for each metal concentration separately. All side reactions of the oxidized or reduced forms of the metals can be accounted for by using the side-reaction coefficient, α ox(red) , defined as [ox(red)] tot = α ox(red) [ox-(red)] free . 78 The formal standard redox potential of the redox pair ox + ze − ⇌ red is then given by ox/red red ox (12) where E ox/red θ is the standard redox potential of the ox/red pair. In the UV−vis spectra, the metal concentrations are quite low, and the diagram here corresponds to a total nominal concentration of 0.1 mM for the oxidized and reduced forms of each metal (the methods detailed in the Supporting Information allow the diagram to be drawn for other concentrations and complexing agents). The formal potential varies with pH because of the formation of hydroxo complexes and solid hydroxides or oxides, and the appearance of solid precipitates causes sharp bends in the curves. Dopamine, and catechols in general, are strongly complexing ligands themselves, but assuming reversible complex formation and intracomplex charge transfer, the complexation with an innocent redox-active ligand is thermodynamically identical to the case with no metal complexation (see the Supporting Information). Therefore, the presence of dopamine is not assumed to influence the oxidizing power of the metals in Figure 8. Figure 8 shows that Ce(IV) is highly oxidizing only in acidic conditions. At low pHs, Fe(III) is also a rather good oxidizing agent, but both metals rapidly lose their oxidizing power as the pH increases. The one-electron oxidation of dopamine to semiquinone is favorable only by Ce at low pHs. In the absence of oxygen, this is the first reaction in the dopamine-oxidation process, and thermodynamics suggests rapid semiquinone formation in acidic solutions (and reasonable rates also at pH 4.5, studied here). Semiquinones tend to disproportionate at all pHs, forming dopaminequinone. On the other hand, if not all Ce(IV) is consumed in the initial oxidation of dopamine to semiquinone, the remaining oxidant can rapidly oxidize semiquinone to dopaminequinone. In both cases, half of the original dopamine will be converted to the quinone form. The same conclusions apply to the reaction of Ce with catechol, too. The formal potential of the Fe(III)/Fe(II) pair is much lower in acidic media, and assuming comparable reorganization energies, the oxidation of dopamine to semiquinone should be slower, in accordance with the experimental results. However, disproportionation and semiquinone oxidation can take place easily, as with cerium, leading to the same overall conversion, although at a lower apparent rate. The differences in the rate of the initial oxidation step explain the different dopaminechrome transients in Figure 7. No Fe−dopamine complex is observed, contrary to the oxidation of catechol, which implies fast intracomplex charge transfer. If semiquinone radicals are formed via some other route, both Ce(IV) and Fe(III) can readily oxidize them to dopaminequinone below ca. pH 6−7. In addition, if oxygen is present in the solution, reaction 2 becomes important. Its equilibrium constant is on the order of 10 −5 in the whole pH range (Figure 2), but both metals produce DSQ in acidic and neutral solutions. On the other hand, the products are removed from the solution as DQ quickly cyclizes, and SO easily reacts further because of its high oxidizing potential. Therefore, reaction 2 is driven to the right, which enhances dopaminechrome formation. These factors contribute to the large onset pH shifts in Figure 5. However, at physiological and higher pHs, both Ce(IV) and Fe(III) actually become rather poor oxidizing agents for dopamine.
The behavior of Cu(II) reserves further analysis. In nature, enzymes catalyzing the oxidation of phenols and catechols usually contain copper in their active sites. 44 Tyrosinase and catechol oxidase, enzymes important in melanogenesis, have a dicopper active site that can reversibly bind dioxygen as a bridge between the copper atoms. Catechol also binds to this dicopper core, although the exact bonding mode is unknown. Metal−dioxygen complexes have been suggested as being involved in the other metal-enhanced, nonenzymatic, aerobic oxidation processes of dopamine and catechol, too. 58 Actually, in such complexes, dioxygen is assumed to be reduced to a superoxide radical. 79 Mechanisms involving redox reactions between iron or copper and oxygen species have also been devised for kinetic studies of the metal-assisted aerobic oxidation of dopamine. 48,49 Especially, as Fe(II) or Cu(I) accumulate during the reaction, oxygen may provide a chemical feedback route by regenerating the oxidized transition metals. It is evident from Figure 8 that thermodynamically, under anaerobic conditions, Cu(II) is only a slightly better oxidant than oxygen. However, chloride ions have an enhancing effect, which has been attributed to strong Cu(I)−chloro complexes. 48 In this work, the experimental and theoretical results confirm that chloride ions enhance the rate of dopamine oxidation by Cu(II) and allow us to attribute the effect to the increased oxidizing power of the Cu(II)/Cu(I) pair, in accordance with previous studies. Calculations further show that under the conditions studied, this is due to the stabilization of Cu(I) by soluble chloro complexes above a chloride concentration of ca. 1 mM and not to the formation of insoluble CuCl (see the Supporting Information, Figure S22). In mildly acidic and The Journal of Physical Chemistry B Article neutral solutions containing chloride, Cu(II) can drive reaction 2 to the right, similar to Ce(IV) and Fe(III), as discussed above. Interestingly, calculations indicate that in alkaline and neutral solutions (ca. pH > 6), including at physiological pH, Cu(II) is actually a better oxidizing agent than Ce(IV) or Fe(III) (however, the redox potentials are dependent on all complexing species in the medium). The clear threshold pH of 5−6 in Figure 5 coincides with the increase of the oxidizing power of Cu(II). Furthermore, thermodynamics shows that the stable potential range of Cu(I) is very narrow, 80 which suggests that Cu(II) could actually act as a two-electron oxidant, contrary to what Fe(III) and Ce(IV) can do, especially as the second oneelectron oxidation potential of dopamine is low under these conditions. It is important to notice that, in this study, the observed ineffectiveness of Cu(II) alone in the oxidation of catechol or dopamine is actually a matter of the time scale. The half-life of intramolecular electron transfer in a Cu(II)− catechol complex is reported to be approximately 10 h at room temperature. 81 Previous work on Cu(II)-assisted dopamine oxidation and polymerization have also demonstrated the very slow rate of the process. 39−41 As a summary, the spectral and electrochemical results show that Ce(IV), Fe(III), and Cu(II) in chloride-containing solutions, can oxidize dopamine (Ce and Fe can oxidize catechol, too) to semiquinone, increasing its formation at mildly acidic pHs. When dissolved oxygen is present, the metals enhance reaction 2, which further accelerates the whole process. Semiquinone can disproportionate or be oxidized by oxygen/ superoxide to o-quinone, as in autoxidation, but all the metals can also directly oxidize semiquinone. With catechol, in the time scale of the experiments, the main product is o-quinone (which slowly produces polycatechol), whereas only dopaminechrome and further products are seen with dopamine (at pH 4.5). This suggests that the ring-closure reaction is fast enough to consume practically all the dopaminequinone formed. The metal-assisted pathway suggested by the analysis is shown in Scheme 3, in which the steps shown with dashed lines can take place under aerobic conditions (possible metal−oxygen− dopamine complexes are omitted).
An important question is whether transition metals enhance the cyclization reaction, too. Although the catalysis of the autoxidation reaction by transition metals has been much studied, mostly close to physiological pH, the effects of metals on this subsequent reaction have received much less attention. In organic solvents, metal-catalyzed Michael aza-addition to α,β-unsaturated ketones has been shown to be due to Bronsted acid catalysis by formed protons. 82 The spectra obtained during the anaerobic oxidation of dopamine in the presence of metals at pH 4.5 allow us to estimate the initial rate of dopaminechrome formation. Initially, absorbance at 480 nm increases linearly in the mixture of 0.1 mM Ce(IV) and 0.1 mM dopamine (Figure 7), yielding an initial rate of ca. 1.4 × 10 −8 M s −1 . Under similar conditions, half of catechol is immediately oxidized to o-quinone. The cyclization is a unimolecular process and must be proportional to the quinone concentration. With these assumptions, the value of 2.3 × 10 −4 s −1 is obtained for the observed dopaminechrome-formation-rate coefficient. On the other hand, we obtain a value of 2.0 × 10 −4 s −1 for the cyclization rate at pH 4.5 from eq 8. Considering the uncertainties involved, these two values are in remarkably good accordance with each other and suggest that the unassisted rate of ring closure is enough to cyclize practically all dopaminequinone formed in the Ce(IV)-enhanced oxidation. In case of Fe(III) and Cu(II) with chloride, the oxidation reaction is slower than with Ce(IV). Therefore, at pH 4.5, the enhancement of dopamine oxidation due to thermodynamic factors is enough to explain the positive effects of all the studied redox-active transition-metal ions on dopaminechrome formation. However, this pH lies within an intermediate range between two possibilities. At low pHs, the rate of dopamine oxidation by metals increases, and the cyclization will become rate-determining. On the other hand, at high pHs, the thermodynamic driving force for the oxidation to semiquinone decreases, but the rate constant of cyclization increases.

■ CONCLUSIONS
A general thermodynamic model was developed to explain the pH dependence of the oxidation and cyclization of dopamine both in autoxidation and in the presence of transition-metal ions. Combined with a simple model mechanism, pHdependent relative probabilities could be obtained for the initial triggering steps in autoxidation. This model is applicable to the reactions of other catecholamines, too. In autoxidation, the most important physicochemical parameters are the pK values of the corresponding semiquinone radical and the pendent amino group in the oxidized quinone form.
The addition of oxidizing transition-metal ions (Ce 4+ , Fe 3+ ) to a dopamine solution greatly accelerates the reaction at mildly acidic pHs and shifts the effective dopamine oxidation and cyclization range down by several pH units. A thermodynamic model easily applied to many different cases was created to explain the enhancement. The model and experimental studies show that these two metals can accelerate the oxidation of dopamine to the semiquinone form in mildly acidic solutions in the absence of oxygen. The semiquinone formed is transformed to dopaminequinone, which rapidly cyclizes and tautomerizes to dopaminechrome. In acidic media, the model predicts that the initial oxidation rate is higher with Ce(IV) than with Fe(III), which is also experimentally observed. In mildly acidic conditions, this enhancement of the dopamine-to-semiquinoneoxidation rate can explain the overall rate enhancement in the process leading to dopaminechrome (and further products in the polydopamine pathway). In acidic media, Cu(II) alone is not a good enough oxidizing agent to initiate the process (at a The Journal of Physical Chemistry B Article reasonable rate) but requires the presence of chloride ions or dissolved oxygen. Thermodynamic considerations show that chloride ions increase the formal Cu(II)/Cu(I) redox potential because of the formation of strong Cu(I)−chloro complexes. Interestingly, in neutral (physiological) and basic media, Cu(II) becomes a better oxidant than Fe(III) or Ce(IV).
We are well aware that the reaction is much more complicated than it is discussed in this work, which focuses on the early stages of the process where many complications due to the further reactions of the formed products can be neglected. However, it shows that the thermodynamic topdown approach can clarify the first steps in polydopamine formation over a wide variety of experimental conditions.