27Al NMR Study of the pH Dependent Hydrolysis Products of Al2(SO4)3 in Different Physiological Media

Soluble inorganic aluminium compounds like aluminium sulfate or aluminium chloride have been challenged by the European Chemical Agency to induce germ cell mutagenicity. Before conducting mutagenicity tests, the hydrolysis products in water and in physiological solutions should be determined as a function of the concentration and pH. We used different 27Al NMR spectroscopic techniques (heteronuclear Overhauser effect spectroscopy (HOESY), exchange spectroscopy (EXSY), diffusion ordered (DOSY)) in this work to gain the information to study the aluminium species in solutions with Al2(SO4)3 concentrations of 50.0, 5.0, and 0.5 g/L and their pH and time dependent transformation. At low pH, three different species were present in all physiological solutions and water: [Al(OH)n(H2O)6 − n](3 − n)+ (n = 0–2), [Al(H2O)5SO4]+, and [Al2(OH)2(H2O)8]4+. Increasing pH reduced the amounts of the two monomer species, with a complete loss at pH 5 for solutions with a concentration of 50.0 g/L and at pH 4 for solutions with a concentration of 5.0 g/L. The dimer species [Al2(OH)2(H2O)8]4+ is present in a pH range between 3 and 6. Less symmetric oligomeric and probably asymmetric aluminium species are formed at pH of 5 and 6. The pH value is the driving force for the formation of aluminium species in all media, whereas the specific medium had only minor effect. No conclusive information could be obtained at pH 7 due to signal loss related to fast quadrupole relaxation of asymmetric aluminium species. A slight reduction of the content of the symmetric aluminium species due to the formation of oligomeric species was observed over a period of 6 weeks. Reference 27Al NMR experiments conducted on saturated water solutions of AlCl3 and those with a concentration of 50 g/L show that the type of salt/counter ion at the same concentration and pH influences the hydrolysis products formed.


Introduction
The REGULATION (EC) No 1907/2006 "REACH" (Registration, Evaluation, and Authorization of Chemicals) [1] dated 18 December 2006 can be considered one of the most complex legislation issued by the European Union [2]. One of the major purposes of the REACH regulation is to improve and secure a high level of protection of human health and the environment, and to enhance the alternative (non-animal based) test methods to determine the hazards of chemical substances. Manufacturers and importers of chemical substances are accountable to gather substance relevant information on physical chemical properties, toxicity, and eco toxicity, which have to be reported through a technical dossier to the European Chemical Agency (ECHA).
Information on the toxicity, especially for humans, shall be generated either with methods other than vertebrate animal tests, through qualitative, quantitative structure-activity relationship models, or based on information from structurally related substances wherever possible [3]. If such investigations are performed on inorganic substances, the behavior in water or water-based solutions has to be determined. Therefore, it is critical to identify the hydrolysis products as a function of the pH and confirm their stability over time in water or water-based solutions.
One of the key concerns of REACH beside data generation and evaluation is the identification of potential CMR effects (carcinogenity, germ cell mutagenicity, and toxicity for reproduction) of substances. This is done by using standardized and internationally accepted test methods, i.e., following the OECD (Organisation for Economic Co-operation and Development) test method guidelines. According to the Guidance on Information Requirements and Chemical Safety Assessment [4], tests to determine mutagenicity are carried out in vitro [5][6][7][8] or in vivo if no non-testing methods such as structure-activity relationship (SAR), quantitative structure-activity relationship (QSAR) [9], or read-across approaches are available.
The Fraunhofer Institute for Toxicology and Experimental Medicine published, in 2017, a final report on a project on behalf of the Federal Institute for Occupational Safety and Health [10], where the solubilities of six different inorganic materials (nano SiO 2 , µ-TiO 2 , nano-TiO 2 , µ-Zr 2 O 3 , Eu 2 O 3 , µ-BaSO 4 ) in two different artificial lung fluids (Gamble's solution [11] and artificial lysosomal fluid [12]) were compared at pH 7.4 and 4.5. The purpose of the study was to define a method to identify potential granular biopersistent particles. One important conclusion was that the solubility varies not only from material to material but may differ even for water-based biological media, for example, when complexation or hydrolysis equilibria are involved. The prediction of the solubility of an inorganic compound in non-innocent solvents is difficult. As solubility with respect to inorganic compounds is associated with a change of structure when a substance comes into contact with water or with biological media, structural investigations are important to derive chemical and structural analogies. These investigations will allow a better prediction of how the structure of substances will change before being tested in in vitro or in vivo.
Soluble inorganic aluminium compounds like aluminium sulfate or aluminium chloride are challenged by ECHA to induce germ cell mutagenicity in 2015 [13]. Read-across approaches taken by industry to demonstrate the similarity of the soluble aluminium compounds towards the less soluble aluminium hydroxide were not accepted because of different solubility and, thus, the biological availability was expected to be different. While it is well known that soluble aluminium compounds like aluminium sulfate form insoluble aluminium hydroxide under neutral conditions, specific analytical methods have to be applied to determine the structure of the aluminium species in biological test media for mutagenicity tests. To conclude on these aluminium species in biological test media used for in vitro mutagenicity tests, commercially produced aluminium sulfate was used for this purpose. A literature review was performed to compare published results with the new data using industrial grade aluminium sulfate dissolved in water to a final concentration typically used in in vivo test methods.
Aluminium sulfate Al 2 (SO 4 ) 3 is used in the form of (Al 2 (SO 4 ) 3 ·16H 2 O, or more generally alums AM(SO 4 ) 2 ·12H 2 O, where A is a monovalent cation (e.g., K + or [NH 4 ] + ) and M is a trivalent cation (e.g., as Al 3+ ), as a flocculating agent for drinking water purification and waste water treatment. Al 3+ is added to water to irreversibly destabilize stable suspensions (formed by charged particles) and to induce aggregation with the consecutive precipitation of the aggregates [14]. The mechanism of the flocculation is related with the positive charge of Al 3+ ,which reduces the repulsion forces between the electric double layers surrounding the dispersed particles and ions until a critical "zeta-potential" is reached and precipitation occurs [15]. In water, aluminium sulfate hydrolyses to form-depending on the pH-different aquo acid species [16,17]. The process is considered a hydrolysis-polymerization with the formation of distinct complexes at low pH values which develop into polynuclear hydroxo-Al complexes and polymers at neutral pH. There is still no agreement on the mechanism of this process, although a variety of analytical techniques have been used to characterize and quantify polynuclear hydroxyl Al species [18]. The well-known larger aluminum hydroxide molecule is the "cage-like" [Al 13 O 4 (OH) 24 (H 2 O) 12 ] 7+ Keggin-Al 13 polycation, which was isolated in the form of its sulfate and selenate salts [19] with a Keggin-type structure resembling that of the 12-phosphotungstic acid [20]. The associated model, developed in the 50 s both experimentally [21,22] and theoretically [23], assumes that hydroxylated Al 3+ undergoes changes from monomeric to polymeric species according to the hexameric ring model. After the formation of the largest gibbsite-fragment of Al 54 (OH) 144 18+ in the sol-state, a gel precipitation of Al(OH) 3 having the sheet structure of gibbsite occurs [24]. On the other hand, the "Cage-like" model assumes that Al 3+ in solution exists only as (hexaaquo) monomer, dimer, Keggin-Al 13 structure (AlO 4 Al 12 (OH) 24 and larger polymerized Al species [24,25]. However, both models cannot explain the formation of various (metastable) polymeric aluminium species. 27 Al NMR spectroscopy is widely applied in analysing the hydrolysis of aluminium salt solutions [24,26,27]. Such studies are mostly related with more concentrated solutions and forced hydrolysis due to the lower sensitivity of the NMR spectroscopy compared to other methods [26]. It was found [24,[28][29][30] that in aqueous solutions containing Al 3+ , over a wide range of pH values, only highly symmetric or moderately distorted complexes could be monitored. These species include the hexa-coordinated monomer [Al(OH) n (H 2 O) 6 − n ] (3 − n)+ (n = 0-2) with 27 Al signals in the shift range of 0-4 ppm and a number of tetra-coordinated AlO 4 in the ε-(63 ppm) and γ-isomer (76 ppm) tridecamers [AlO 4 Al 12 (OH) 24  Here we report an 27 Al NMR study on the hydrolysis products of Al 2 (SO 4 ) 3 in four different physiological media as a function of concentration and pH value. At lower pH values it was possible to identify the hydrolysis products using solution NMR. At neutral pH, no signal was detected due to the very fast quadrupole relaxation of Al in a non-symmetric environment. A broad band solid state NMR spectrometer at very low magic angle spinning (MAS) and an excitation range of ca. 3.2 MHz was used to check for the existence of very broad 27 Al signals, however, with no success.
The distribution of the resulting hydrolysis products at different pH values in water were compared for three different concentrations. Based on the conducted 27 Al NMR experiments as well as earlier studies we identified most of the hydrolysis products. We observed the presence of three different aluminum species at pH 3 in all physiological media and water. The dominant species is the highly symmetric [Al(OH) n (H 2 O) 6 − n ] (3 − n)+ (n = 0-2) with only small amounts of [Al(OH) n (H 2 O) 5 − n SO 4 ] + (n = 0-2) and the dimer [Al 2 (OH) 2 (H 2 O) 8 ] 4+ being present besides the major component. At pH 4, the situation changes significantly. The main hydrolysis product is still the symmetric monomer, but the content of the less symmetric dimer characterized by a 27 Al NMR shift of ca. 5 ppm [16] increased compared to that at a pH value of 3. Changing the pH to 5 and 6 results in the appearance of new significantly broadened and overlapping 27 Al NMR resonances detected in lower field in the range of ca. 5-12 ppm. The signals can be identified as, most likely, octahedral complexes with lower symmetry including dimer, trimer, and higher oligomers. At pH 7, no aluminium signal was detected. The formation of a residue indicates the formation of colloidal hydrated alumina. The solution samples remained stable for 6 weeks at ambient conditions with only minor changes in the ratio of the hydrolysis products that slightly shift towards the formation of oligomeric species.

Results and Discussion
Aqueous solutions of Al 2 (SO 4 ) 3 in the pH range from 3 to 7 with three different concentrations (0.5, 5.0, and 50.0 mg/L) were prepared in different physiological media (Table 1) and studied by 27 Al NMR spectroscopy.The pH was adjusted with the buffers as compiled in Table 2 (Materials and Methods). Roswell Park Memorial Institute Medium commonly referred to as RPMI 1640 (L1) and Dulbecco's Modified Eagle Medium (DMEM; L2) are two of the commonly used cell culture media for growing mammalian cells. To allow a precise comparison of potential side reactions, both media are used with and without additives. RPMI 1640 was selected in pure form and added with Penicillin, Streptomycin, and Phytohemagglutinine. The enriched medium is referred in Table 1 as L3. Penicillin, Streptomycin, and Phytohemagglutinine are used as inhibition agents to minimize growth of microorganisms. The added Horse serum in DMEM is a typical promoter of proliferation of the cells. The enriched DMEM medium is referred as L4. L4 is the typical medium for the in vitro gene mutation test using the hypoxanthine-guanine phosphoribosyltransferase (HPRT) genes [6].
For all media, the solutions with concentrations 0.5, 5.0, and 50.0 g/L at pH 3 and 4 were clear, whereas those at pH 5 and 6 were turbid with an incipient formation of a precipitate with time. In all samples with pH 7, a gel-like precipitate was observed after the initial dissolution. Before each NMR experiment, the solutions were homogenized by shaking the NMR tube. pH Composition 3 7.74 g citric acid (waterfree) + 3.49 g NaCl + 206 mL NaOH (c = 0.1 mol/L) 4 10.75 g citric acid (waterfree) + 2.57 g NaCl + 68 mL NaOH (c = 1 mol/L) 5 18.52 g citric acid (waterfree) + 196.4 mL NaOH (c = 1 mol/L) 6 11.46 g citric acid (waterfree) + 159.6 mL NaOH (c = 1 mol/L) 7 3.52 g KH 2 PO 4 + 7.26 g Na 2 HPO 4 · 2H 2 O 27 Al NMR spectroscopy was used to identify and characterize the hydrolysis products of aluminium sulphate. The 27 Al nucleus is magnetically active (I = 5/2) with good receptivity relative to 13 C (D = 1170) and 100% natural abundance. Being a quadrupole nucleus, it interacts with the external magnetic field and the electric field gradient generated by the surrounding environment [31]. The 27 Al NMR spectra are broadened due to the quadrupole interaction. They remain, however, very informative as the shift range-from ca. −100 to 300 ppm-is large. Furthermore, there is a direct correlation of the chemical shift and the coordination number of the aluminium cation [26,27].

Identification of the Hydrolysis Products of Al 2 (SO 4 ) 3 at Different pH Values
The hexa-coordinated aluminium cations are mostly shielded and resonate in higher fields (shift range −30 to 30 ppm relative to the reference Al(H 2 O) 6 3+ at 0 ppm), whereas a low field shift was observed for the penta-coordinated aluminium species. The tetra-coordinated aluminium species is characterized by the lowest field shifts compared to the former two classes. The drawback of 27 Al NMR spectroscopy, although it is widely applied, is that it can detect only aluminium in a symmetric environment. The loss of symmetry, as in the case of large complexes and polymeric species, leads to a fast quadrupole relaxation. Therefore, Al spectra of such species are broadened beyond detection in NMR spectroscopy.
Al 2 (SO 4 ) 3 was dissolved in concentrations of 0.5 g/L, 5.0 g/L, and 50 g/L in four different physiological media, L1-4, and water for comparison at pH values ranging from 3 to 7. This concentration range is typical for in vitro testing. All recorded 27 Al NMR spectra for the different physiological media were similar for the same pH. Therefore, only representative spectra of Al 2 (SO 4 ) 3 dissolved in the L1 medium in a concentration of 50 g/L (Figure 1) are discussed.
In the 27 Al NMR spectrum of Al 2 (SO 4 ) 3 in L1 medium at pH 3, two resonance signals at ca. 0.9 ppm and at ca. −2.4 ppm with an integral ratio of 9:1 are clearly detected. It should be noted that at these conditions, no peak at ca. −6.7 ppm related with the formation of a bi-sulfato complex [26] was detected. Such a resonance signal has been observed earlier, also in the case of a forced hydrolysis of aluminium through the thermal decomposition of urea for the [Al(H 2 O) 4 (urea) 2 ] 3+ at −4.47 ppm in addition to the signal of the monosubstituted urea complex [Al(H 2 O) 5 (urea)] 3+ at −2.67 ppm [32].
In the current study, a shoulder at ca. 5 ppm at the base of the peak related with the hexa-aquo complex at 0.9 ppm being observed. The shifts of three detected 27 Al resonances are within the shift range of hexa-coordinated Al 3+ [26,27]. We attribute the signal at 0.9 ppm to the presence of [Al(H 2 O) 6 ] 3+ in solution [16]. It is worth mentioning that the measurements were conducted without 2 H spin lock. Therefore, a slight deviation from the 0 ppm value is not unexpected. Hexa-coordinated aluminium nuclei, which possess one [SO 4 ] 2− anion according to some sources in the first [16,33] or the second [34] coordination shell, resonate slightly up-field.  27 Al NMR solution spectra of Al 2 (SO 4 ) 3 in L1 medium recorded at pH 3 to 7 and scaled to the signal with the highest intensity at 0.9 ppm (left) with the spectral region characteristic for the hexa-coordinated Al magnified (right). All spectra were recorded without 2 H spin lock as no deuterated solvent was used. For all spectra, the background 27 Al signal due to the glass of the NMR tube was recorded in a separate experiment with the same number of scans and receiver gain and subtracted from the spectrum of the respective sample.
To check the presence of a coordination shell of water molecules around the Al 3+ cations, 2D 27 Al heteronuclear Overhauser effect spectroscopy (HOESY) experiments with mixing times of 10, 30, 50, and 100 ms and 1 s were carried out ( Figure S1, Supplementary Information).The HOESY experiment brings information about the spatial connectivity between heteronuclear spins, i.e., the signals in the spectrum corresponding to the through space correlation between dipolar coupled spins situated within a distance smaller than 5 Å. Figure 2 (top) presents the 27 Al HOESY spectrum of Al 2 (SO 4 ) 3 in solution L1 at pH 3 recorded with a mixing time of 30 ms. The intense correlation signal at the Al shift is characteristic for octahedral aluminium in [Al(H 2 O) 6 ] 3* and water molecules at 4.3 ppm. This intensive signal was detected also in all 27 Al HOESY spectra recorded with 10 ms, 30 ms, 100 ms, and 1 s mixing times. Water molecules are present in close vicinity, i.e., in the coordination shell of the aluminium species. It should be noted that at longer mixing times, there should be a contribution coming also from the water molecules of the second coordination sphere and from the solution. A correlation signal for the low intensity resonance at −2.4 ppm was detected in the spectrum recorded with 1 s mixing time. Also, a through space connectivity with water molecules was confirmed. However, a water coordination of this species could not be confirmed unequivocally. There are indications that the species responsible for the signal at 5.2 ppm (Figure S1), and which appears as a shoulder in the 27 Al spectrum, has water molecules in close proximity as well. It is, however, difficult to confirm this beyond doubt, as the signal has only low intensity and overlaps with the base of the resonance of [Al(H 2 O) 6 ] 3+ .  27 Al exchange spectroscopy (EXSY) spectra recorded with 0 ms mixing time (left) as a reference spectrum and 10 ms mixing time (right). 27 Al T 1 relaxation measurements conducted by applying the inversion recovery method brought further information about the observed Al species. The spin-lattice relaxation times determined for the signals at 0.9 and −2.4 ppm were 52 ms and 48 ms, respectively. A comparable T 1 relaxation time was observed earlier for a 1.5 M water solution of AlCl 3 [35]. Such relatively long T 1 relaxation times for the quadrupole aluminium nucleus imply a symmetric environment. This is consistent with the suggested octahedral coordination for both complexes. Additionally, these comparable T 1 times suggest a chemical exchange between these two Al species. The Al 3+ ions typically carry ligands; therefore, a chemical exchange of aluminium ions is not possible. However, the surrounding ligands may exchange, which may lead to different chemical environments, i.e., different chemical shifts for the 27 Al nuclei. 27 Al exchange spectroscopy (EXSY) NMR could be used as evidence for an equilibrium ligand exchange between Al 3+ species according to [Al(H 2 In addition, the exchange rate constants for both reactions could be estimated [16,26]. They were determined at the coalescence point of both signals by measuring variable temperature (VT) 1D NMR spectra [16]. A drawback of this method is that the viscosity is reduced at higher temperatures, which affects the rotational correlation time of the respective complexes. It has an immediate effect on the line width of the aluminium signal as well. In addition, high temperatures may affect the lifetime of the complexes.
We conducted 27 Al exchange spectroscopy (EXSY) on the sample in L1 medium at pH 3 at ambient conditions with mixing times ranging from 1.5 ms to 50 ms ( Figure S2). Figure 2 (bottom) shows representative 27 Al EXSY spectra measured with 0 ms (reference experiment) and 10 ms mixing times. The spectrum recorded with 10 ms mixing time reveals cross-correlation signals between the 0.9 and the −2.4 ppm 27 Al resonances. These cross-peaks have the same phase as the diagonal signals, which confirms a chemical exchange between both aluminium species. Such correlation signals are detected in all spectra with mixing times below 50 ms. Longer mixing times up to 1 s (spectra not presented) lead to signal decay due to relaxation.
The exchange rate constants k AB and k BA for the reversible reaction in Equation (1) were derived from the intensities in two consecutive EXSY experiments recorded at the same conditions-the reference experiment with 0 s mixing time and the EXSY experiment with 30 ms mixing time (using the EXSYCALC program from Mestrelab) [36,37]. At 292 K, the rate constant for the formation of the sulfo-complex B was determined as k AB = 0.217 s −1 , whereas that for the formation of the hexaaquo-Al-complex A was k BA = 81.669 s −1 .
In order to characterize the motional behavior of both Al species, we applied diffusion ordered spectroscopy (DOSY). In this experiment, the diffusion coefficient is correlated to the chemical shift by applying a pulsed gradient field of varying strength to obtain an exponential decay of the 27 Al-resonances as a result of the molecules diffusing out of the excitation range. The 27 Al signal intensity is attenuated depending on the diffusion time. It is worth mentioning that the signal decay and increasing gradient strength do not follow the "standard" exponential relation. Even a strong gradient field did not attenuate the Al-signal significantly. The obtained diffusion coefficients from the diffusion analysis, 2.4 × 10 −11 m 2 /s for the [Al(H 2 O) 6 ] 3+ complex at 0.9 ppm and 3.4 × 10 −11 m 2 /s for [Al(H 2 O) 5 SO 4 ] + at −2.4 ppm, are too high for small complexes. Such values might describe complexes with a large coordination shell (affected by chemical exchange) or clusters.
Significant changes were observed in the 27 Al NMR spectrum of Al 2 (SO 4 ) 3 dissolved in L1 when the pH was increased to 4 ( Figure 1). The intensity of the signals at 0.9 ppm related to [Al(H 2 O) 6 ] 3+ and the one at −2.4 ppm related to [Al(H 2 O) 5 SO 4 ] + are dramatically reduced to less than 5% of the intensity in the spectrum at pH 3. The resonance at 5.2 ppm assigned to the [Al 2 (OH) 2 (H 2 O) 8 ] 4+ dimer species gains intensity to about 50% compared to that of the resonance at pH 3. A broad shoulder appears at lower fields. The 27 Al intensity corresponding to this shoulder is at least partially associated with the trimer [Al 3 (OH) 4 (H 2 O) 10 ] 5+ [24]. Considering the reported shifts of the 27 Al resonance to a lower field with increasing nuclearity of the species in solution, higher oligomers may contribute to this intensity as well. We observe an overall "loss" in Al signal intensity. This phenomenon has been noted for ssNMR [24] and is related with NMR resonances from Al sites having low symmetry. The result is a large electric field gradient which can broaden the NMR signals beyond detection. Several methods have been proposed to quantify this "invisible aluminium" contribution. Some of them include the shift of the measurements to higher magnetic fields using fast and ultra-fast magic angle spinning (MAS) [38], while others rely on short pulse angles (of about 15 degree) on the resonance to excite quantitatively the principle (−1/2 + 1/2) transition [39]. Methodological developments like different echo techniques [40], Multiple-Quantum Magic-Angle Spinning (MQMAS) [40,41], or TRAnsfer of Populations in DOuble Resonance (TRAPDOR) [42] were proposed as well. In solution, in addition to the large electric field gradients, an important role may also be played by the slow tumbling times for large complexes in viscous media [24]. An elegant way to solve the problem of quantifying the "invisible aluminium" has been proposed [43]. So far, the NMR "invisible" Al species have been identified with the help of electrospray ionization mass spectroscopy (ESI-MS) from the hydrolysis of AlCl 3 [44,45] at pH 4 to 6.4.
The 2D experiments conducted at pH 4 were partially successful due to the fast quadrupole relaxation of the asymmetric Al sites. It was possible to determine the T 1 relaxation times for the main resonances at −2.4 ppm (31 ms), 0.9 ppm (18 ms), and at ca. 5.6 ppm (1.6 ms). The reduction of the T 1 relaxation time of the signal at 0.9 ppm indicates that [Al(H 2 O) 6 ] 3+ is transformed into higher oligomers and polymers. At the same time, the intensity loss for [Al(H 2 O) 5 SO 4 ] + appeared related to the transformation to the hexa-aquo complex, which shifts the equilibrium to the polynuclear species.
At pH 5 and 6, the solution is no longer clear. Incipient precipitation was observed, which is related to a change in the degree of hydrolysis (i.e., the [OH − ]/ [Al 3+ ] ratio), which describes the extent to which the acidic buffer capacity of pure salt solutions is consumed [46]. For [OH − ]/[Al 3+ ] ≈ 2.5, the solution is fully hydrolyzed and precipitation occurs [17,46].
At pH 5, the critical [OH − ]/ [Al 3+ ] ratio has been exceeded, and the aluminium content is distributed between the precipitate, the "invisible aluminium", and the dissolved species. Information about the latter was obtained from the 27 ) were no longer detected, and a broad signal covering a shift range of ≈30 ppm was observed. This shift range is still related to hexa-coordinated aluminium, the broadening being due to species with some asymmetry in their environment.
At pH 7, a significant amount of Al(OH) 3 sol precipitated [27]. No 27 Al NMR resonances could be detected due to extreme signal broadening. 27 Al NMR experiments were conducted on samples dissolved in L1 on a broad band solid state NMR spectrometer, exciting a bandwidth of ca. 3.2 MHz. The recorded spectra (not shown) did not reveal additional signals. The signals are scaled relative to the one at 0.9 ppm. No significant intensity differences are observed. Thus, the main effect on the spectra is related to the pH of the solutions.

Comparison of the 27 Al Signals in the Spectra Recorded for Different Media
The 27 Al spectra of Al 2 (SO 4 ) 3 dissolved in RPMI 1640 medium (L1) and water are very similar in terms of signal intensities and ratios. The resonances at 5.2 ppm are detected as shoulders of the main peak at 0.9 ppm. The addition of antibiotics and the lectin phytohemagglutinin (PHA consisting of two closely related proteins PHA-L and PHA-E, which agglutinate leucocytes and erythrocytes, respectively) in the RPMI 1640 medium slightly enhances the formation of more symmetric species at pH 3.
In the 27 Al NMR spectra of Al 2 (SO 4 ) 3 solutions in DMEM medium, the three Al resonances (mentioned above) were observed. The signal intensities and the respective signal ratio of the signals at 0.9 and −2.4 ppm are comparable to the RPMI 1640 medium. Clearly, the DMEM medium promotes dimer formation at the expense of the [Al(H 2 O) 6 ] 3+ hexaaquo and the [Al(H 2 O) 5 SO 4 ] + complex. This may be related with a slight increase of the pH value. Additionally, the intensities in the spectrum in the L4 medium are lower than those in the DMEM medium. The addition of the antibiotics has only a minimal influence on the 27 Al NMR spectrum. The addition of horse serum to the DMEM medium (L4) is responsible for a decrease of the 27 Al intensity compared to the pure DMEM medium.
It should be noted that serum albumin (including horse serum), the most abundant plasma protein in mammals, is characterized by an extraordinary ligand [47] and metal [48] binding capacity. Therefore, it is not surprising that the presence in the solution of another binding functionality could promote formation of non-symmetric (including hexa-coordinated) aluminium complexes. Moreover, it has been confirmed by a number of techniques such as spectrophotometric titration, ultrafiltration, dialysis and gel chromatography [49] that A1 3+ in serum is highly protein bound to both human serum albumin (HAS) and human serum transferrin (HSTF), at the protein concentrations found in blood plasma. Furthermore, pH dependence of the Al 3+ binding has been observed with maximal binding at neutral pH [49]. Based on those results, it was suggested that HSA and HSTF may serve as carriers in the biological transport of A1 3+ in blood plasma [49 and references cited therein]. Considering the relatively small amount of horse serum at this aluminium sulfate concentration and the weaker binding of Al 3+ at pH 3, it is not surprising that there is no striking difference between the RPMI and water solutions. A more pronounced influence on the NMR spectra is expected at lower Al 2 (SO 4 ) 3 concentrations and higher pH values. No precipitation was observed in all five samples. Figure 4 presents the 27 6 ] 3+ was visible only for the water solution at pH 5. Some intensity, associated with the dimer, remained for the water solution and-to a smaller extent-in the L1 media at pH 5. In addition, a new significantly broadened signal appeared with a maximum at ca. 12 ppm. Surprisingly, the Al signals in the L3 medium that were most intense at pH 3 and 4, decayed at pH 5. A precipitate was formed in all solutions (except for the aqueous solution).  27 Al resonance at ca. 12 ppm appears instead. At pH 6, this broad signal remains only in the RPMI and DMEM media. At pH 6, precipitation occurred in all five solutions. No intensity was detected for the L3 and the water solutions. Only broad Al resonances in the media L1 and L2 were detected, where no additives were used. The same signal, but with reduced intensity, was observed in the L4 media as well.
At pH 7, no signal was detected in the Al NMR spectra as the Al 3+ cations are in an asymmetric environment due to the formation of the Al(OH) 3 sol. Figure S3 presents the 27 Al NMR spectra of Al 2 (SO 4 ) 3 with a concentration of 5.0 g/L in the pH range 3-7 recorded at the same conditions for media L1-4 and water. At pH 3, the spectra in all media were comparable as detected resonances with the spectra at higher concentrations. The three signals at 5.2 ppm for the dimer [Al 2 (OH) 2 6 ] 3+ was still observed in the water solution with dramatically reduced intensity. Instead, a new broadened peak appeared at ≈12 ppm. Such a resonance is related with hexa-coordinated aluminium in an asymmetric environment, most probably due to the formation of oligomeric species. At pH 4 to 6, the dominating signal is that associated with oligomeric species. No signal was detected for all samples at neutral pH. Thus, we conclude that at lower concentrations, the formation of oligomeric species occurs at pH 4-6, i.e., the condensation equilibrium [Al(H 2 O) 6 ] 3+ = Al(OH) 3 (H 2 O) 3 + 3H + is shifted to the right.
The spectra of the samples at the lowest concentration of 0.5 g/L were recorded with 1024 scans to enhance the aluminium signal intensity ( Figure S4). Even at these conditions, meaningful spectra could be measured only for pH 3 in water as all three characteristic signals for this pH observed at the higher concentrations were detected. For the L1 medium, the signal at 0.9 ppm was still visible. It is expected that the main hydrolysis product/s in the pH range 4-6 remain as hydrated oligomeric aluminium species.
To evaluate the influence of the type of aluminium salt in relation to the type of counter ion on the 27 Al NMR spectra in aquous solutions as a function of pH, we recorded representative spectra of AlCl 3 and compared it to the spectra of Al 2 (SO 4 ) 3 . Figure S5 Figure S6). Clearly, the stability of the [Al(H 2 O) 5 SO 4 ] + complex is higher compared to the respective chlor-containing analog. Moreover, it is known that N-and O-containing hard Lewis bases (as (SO 4 ) 2− ) form stable complexes with light s-and p-metal cations (such as Al 3+ ). The type of salt/counter ion at the same concentration and pH has an effect on the hydrolysis products formed. The presence of penicillin resulted in a slight reduction of the intensity of the monomer signal at pH 5.
It should be noted that in all Al 2 (SO 4 ) 3 solutions, independent of the concentration, no signal for the A13 Keggin polycomplex was observed. However, the 27 Al NMR spectrum recorded for a saturated solution ofAlCl 3 in water for comparison ( Figure S6) contained the resonance for the tetra-coordinated central AlO 4 in the ε-isomer (64 ppm) and in the polycation [AlO 4 Al 12 (OH) 24 (H 2 O) 12 ] 7+ at pH 4 and 5. This fact confirmed that the type of salt/counter ion at the same concentration and pH influences the hydrolysis products formed.

Stability of the Al 2 (SO 4 ) 3 Solutions
To check the stability of the Al 2 (SO 4 ) 3 solutions we have chosen representative samples with a concentration of 50.0 g/L in the pH range from 3 to 6 in L1 media. The samples were kept in NMR tubes at ambient conditions for 6 weeks. Subsequently, 27 Al NMR spectra were measured under identical experimental conditions using identical datasets. Before each measurement, the samples were homogenized by shaking the NMR tube several times. The recorded NMR spectra shown in Figure S7 were compared with those obtained 6 weeks earlier.
Slight changes in the spectra are due to aging. At pH 3, the intensity of all signals decreased by ca. 8%. At pH 4, a comparable decrease was detected for the signal of the [Al 2 (OH) 2 5 SO 4 ] + remained unchanged. The most pronounced differences occured at pH 5 and 6, where the broad signal at 12-13 ppm sharpened. We attribute the broadening to the formation of a better defined (but still asymmetric) hexa-coordinated Al species. At pH 7, no signal was detected.
Based on the results for these representative samples, slow changes occur in the solutions of Al 2 (SO 4 ) 3 over a prolonged period of time. These changes are related to the decreasing Al 8 ] 4+ dimers at lower pH and formation of an oligomeric Al species at pH 5 and 6.

Materials and Methods
Al 2 (SO 4 ) 3 was obtained from Grace Silica GmbH (Dueren, Germany) and used without further treatment. The medium L1 (Roswell Park Memorial Institute Medium (RPMI 1640)) with stable L-glutamine containing 25 mM 4-(2-hydroxyethyl)-1-piperazineethanesulphonic acid (HEPES), 5.5 g/L NaCl, 5 g/L phenol red, and 2.0 g/L NaHCO 3 was purchased from Biochrom GmbH and used without further treatment. Medium L2 (Dulbecco's Modified Eagle Medium (DMEM)) without L-glutamine containing 3.7 g/L NaHCO 3 and 4.5 g/L D-glucose was purchased from Biochrom and used without further treatment. The medium L3 was prepared from L1 with the addition of 2% phytohemagglutinin, 1% penicillin (10,000 U/mL), and 1% streptomycin (10,000 U/mL). Medium L4 was prepared from L2 with the addition of 5% horse serum, 1% penicillin (10,000 U/mL) and 1% streptomycin (10,000 U/mL). The pH was adjusted with the following buffers: All 27 Al NMR experiments were conducted on a Bruker Avance DRX 400 (Bruker Biospin GmbH, Rheinstetten, Germany) NMR spectrometer operating at a 1 H frequency of 400.31 MHz and 27 Al frequency of 104.31 MHz, equipped with a 5 mm inverse two channel probe head with z-gradients. All experiments were recorded without 2 H lock as no deuterated solvent was used. The single pulse excitation experiments were measured with a recycle delay of 1 s, averaging 32 scans. For the 27 Al NMR experiments with background suppression, 128 scans for the 5.0 g/L and 50.0 g/L concentrations and 1024 scans for the 0.5 g/L concentration were averaged with a recycle delay of 2 s. The inversion-recovery NMR method was used to determine the 27 Al spin-lattice relaxation times. For the 27 Al exchange spectroscopy (EXSY) experiments, 2048 data points were recorded in the direct dimension with 1024 increments in the indirect dimension and 16 scans per increment in a phase sensitive mode. The mixing times were set to 0 s (reference experiment) and 1.5, 5, 10, 30, and 50 ms and a recycle delay of 0.5 s was used. The exchange rate constants were determined from the reference experiment and with the 10 ms mixing time using the EXSYCALC software package from Mestrelab. The 27 Al heteronuclear Overhauser effect spectroscopy (HOESY) experiments were conducted with aluminium detection, recording 4096 data points in the direct dimension with 64 increments in the indirect dimension and 256 scans per increment. The recycle delay was set to 1 s. Stimulated echo sequence with bipolar gradient pulses and a longitudinal eddy current delay was used for the 27 Al diffusion ordered spectroscopy (DOSY) experiments. The gradient strength was incremented in 16 steps from 2% to 95% of the maximum gradient strength. The diffusion time and the gradient pulse length for the Al 2 (SO 4 ) 3 sample at pH 3 dissolved in L1 medium were 120 ms and 2.8 ms with 2 s recycle delay, respectively. After Fourier transformation and baseline correction, the diffusion dimension of the 2D DOSY spectra was processed using the Bruker Topspin 1.3 software package (2005, patchlevel 8, Bruker Biospin GmbH, Rheinstetten, Germany). The diffusion analysis was performed using the Topspin T 1 /T 2 relaxation package. 27 Al NMR reference spectra were recorded for the samples dissolved in L1 medium for all pH values using a broad band solid state Bruker Avance DSX 400NMR spectrometer operating at a frequency of 104.2 MHz. A commercial three channel probe head from Bruker at very low magic angle spinning (MAS) of ca. 2 kHz with no spinning frequency control was used for the experiments.