Introduction

Noninvasive measurement of local pH is of important significance since abnormal changes of pH are closely associated with cellular dysfunctions and pathologies such as ischemia, oxidative stress^[1], apoptosis^[2] and tumor size progression^[3]. Electron paramagnetic resonance (EPR) coupled with the use of exogenous spin probes has been an indispensable technique to measure pH both in vitro and in vivo due to its noninvasiveness, high sensitivity and appropriate magnetic penetration depth^[4-6]. Nitroxide radicals initially reported in the early seventies^{[7, 8]} are the most commonly used EPR probes for pH measurement. Typically, pH-induced changes in EPR spectra are assessed by measuring the isotropic hyperfine splitting (hfs) constant for nitrogen (α_N) or, sometimes, for phosphorus (α_P), although the electronic g-factor is also affected by proton exchange reaction. The reliability of method can be enhanced by simultaneous analysis of EPR spectra recorded at different pH environments, when the formation constants of mass-balance equation and the EPR parameters are revealed by a two-dimensional (2D) analysis^[9]. The method was originally developed for studying copper coordination but proved to be also useful for the pH sensitive nitroxides^[10]. The pH-sensitive nitroxides have found numerous applications in in vivo measurement of pH inside the gut^[11], release processes in biodegradable polymers^[12], drug-induced changes in the stomach acidity of living rats^[13], and tumor extracellular pH^[14]. However, rapid bioreduction of nitroxides with loss of signal and the relatively broad multiple signals from nitrogen and proton hyperfine splittings greatly limit their applications in in vivo systems due to low EPR sensitivity.

In the past two decades, fully substituted tetrathiatriarylmethyl (TAM) radicals such as OX063 and CT-03 (Fig. 1) have accepted intense attention in the EPR field owing to their sharp EPR single lines, good water solubility and high stability^[15-18]. TAM radicals are a class of air-stable carbon-centered radicals in which the unpaired electron is predominantly located at the central carbon and further stabilized by delocalization into three aromatic ring systems. Full substitution of each phenyl ring by alkylthio and carboxylate groups not only eliminates the hyperfine splitting from protons in TAM radicals, but also enhances their stability by electronic and steric effects. In 2007, Khramstov et al.^[19] pioneered the use of TAM radicals for measurement of pH based on their pH sensitivity to the magnetic parameters (e.g., α_H and g factor). One of the authors in this study also reported a complex of dendritic TAM radical with copper (Ⅱ) as pH probe^[20]. Thereafter, the amino and triphosphonate derivatives of TAM radicals were also synthesized^[21-23] and their spectral patterns exhibited high sensitivity to pH in the physiological range due to nearly neutral pKa values. However, these probes had much complicated EPR signals due to hyperfine splittings from various paramagnetic nuclei (e.g., ¹⁴N, ¹H and ³¹P) in the molecules. Despite this, the triphosphonated TAM probe was used to measure the extracellular pH in tumors^[24]. Recently, in order to simplify the EPR signals of pH probes, Khramstov and coauthors^[25] have reported a monophosphonated TAM probe (p₁TAM-H) and its deuterated derivative (p₁TAM-D) (Fig. 1) which is so far the most ideal pH and O₂ dual probe owing to its simple doublet EPR spectral pattern and improved water solubility required for in vivo applications. However, the synthetic efficiency of these probes was rather low with a yield of < 2 % for p₁TAM-H/p₁TAM-D (about 10 mg) from the triarylmethanol (530 mg). As such, it is challenging to obtain sufficient amounts for their applications in in vivo EPR imaging. Moreover, due to the relatively hydrophobic nature, these pH probes may potentially bind with albumins which are present at high concentrations in blood^[26], resulting in a drastic decrease of the EPR signal intensity and preventing their application as pH and oxygen probes^[27].

In the present work, we developed an efficient procedure for the synthesis of p₁TAM-H and p₁TAM-D with a yield of 25% from the triarylmethanol, allowing us to obtain sufficient amounts for their further derivatization. Thereafter, two PEGylated derivatives (i.e., POP and dPOP) of the phosphonated TAM radicals were also synthesized. Effects of PEGylation on pKa, albumin binding, and stability as well as O₂ sensitivity of these two probes were explored.

1 Experimental section 1.1 Instruments and reagents

All manipulations involving air- and moisture-sensitive compounds were carried out under an atmosphere of purified argon by using standard Schlenk-line techniques. Solvents were dried with appropriate drying agents and degassed before use. All commercially available reagents were used as received without further purification. NMR spectra were recorded on a BRUKER Avance Ⅱ 400 MHz NMR spectrometer. Chemical shifts (δ) were reported and referenced to tetramethylsilane (TMS) or NMR solvents used (δ_H 7.27 for CDCl₃). The following abbreviations were used: s=singlet, d=doublet, t=triplet, m=multiplet. Coupling constants (J) were reported in Hz. High resolution mass spectra (HRMS) were obtained on an MALDI-TOF spectrometer (Bruker Autoflex Ⅲ TOF/TOF).

Xanthine (X), xanthine oxidase (XO), glutathione (GSH), ascorbic acid (Asc), hydrogen peroxide (H₂O₂), ammonium iron (Ⅲ) sulfate hexahydrate [(NH₄)₂Fe(SO₄)₂·6H₂O], diethylenetriaminepentaacetic acid (DTPA), nitrilotriacetic acid disodium salt (NTA), bovine serum albumin (fatty acid-free, BSA), tetramethylethylenediamine (TMEDA), butyl lithium (1.6 mol/L in hexanes), dimethyl carbonate, diethyl chlorophosphate, trifluoromethanesulfonic acid, trimethylsilyl bromide (TMSBr) and cesium carbonate were commercially available. All other chemicals used throughout the experiments were of analytical grade. The dendrons D1 and dD1were synthesized as previously described^[28]. Phosphate buffer (PBS, 1 mmol/L, pH 7.4) was prepared from sodium dihydrogen phosphate and disodium hydrogen phosphate in the presence of DTPA (100 μmol/L).

1.2 EPR spectroscopy

EPR spectra were recorded on a Bruker EMX-plus X-band spectrometer at room temperature. General instrumental settings were as follows: microwave power, 40 μW; time constant, 20.48 ms; conversion time, 10 ms; sweep time, 81.92 s; modulation frequency, 10 kHz; modulation amplitude, 2×10^-6 T; sweep width, 5×10^-4 T; number of points, 8 192. EPR measurements were performed in 50 μL capillary tubes. In addition, EPR measurements under anaerobic conditions were carried out using a gas-permeable Teflon tube (i.d. = 0.8 mm). Briefly, the experimental solution was transferred to the tube which was then sealed at both ends. The sealed sample was placed inside a quartz EPR tube with open ends. Argon gas was allowed to bleed into the EPR tube and then EPR spectrum was recorded.

1.3 EPR simulation

Spectral simulation was carried out using the simulation program written by Professor Rockenbauer^{[9, 10, 29]}. As part of the EPR simulation program, the 2D model "EPR pH titration" was employed to analyze the dependence of pH on the fraction of two ionization states of the pH probes in our work. Different EPR parameters including α_P, g tensor, resolved and non-resolved proton or deuterium splitting, line width and pKa value were adjusted by combining iteration and least squares procedures, with a search for the minimum square deviation between experimental and computed spectra. It was worth noting that 2D EPR analysis and superposition analysis of EPR spectra originates from the same simulation program just with different modes.

1.4 pH titration

The aqueous solution of the TAM probe was titrated by addition of a small volume of NaOH or HCl with the final dilution of sample less than 1%. pH was controlled by electrode which was calibrated at 25 ℃ using pH values for reference solution. Temperature of reference and titrated solutions during pH measurements was keep the same data using type of integral temp sensor attached to Mettler-Toledo Lab pH electrode FE20. Anoxic conditions were maintained using a gas-permeable Teflon tube (i.d. = 0.8 mm) which was placed inside a quartz EPR tube with open ends and argon gas was allowed to bleed into the EPR tube.

1.5 pKa value calculation

The values of fraction, f, of each ionization state of the TAM probe were obtained by computer simulation of experimental spectra. The fraction values were then fitted by a standard titration equation to yield the corresponding value of pKa. For example, titration of dPOP is described by the equation:

$ f({\rm{dPO}}{{\rm{P}}^{2 - }}) = 1 - f({\rm{dPO}}{{\rm{P}}^{1 - }}) = \frac{1}{{1 + {{10}^{ - {\rm{pH}} + {\rm{p}}K{\rm{a}}}}}} $

(1)

2 Results and discussion 2.1 Comparison of synthetic methods and synthesis of main probes

The synthesis of monophosphonated TAM radicals (i.e., p₁TAM-H and p₁TAM-D) was originally reported by Khramtsov et al.^[25]. As shown in Fig. 2(a), the key intermediate 2 was synthesized previously by complete deprotonation of the triarylmethanol 1 using a large excess of n-BuLi (10 Equiv.) and TMEDA (10 Equiv.), followed by treatment with the mixture of diethyl carbonate (DMC) and diethyl chlorophosphate. Due to the harsh deprotonation condition and subsequent nonselective reaction of the resulting aryl anion with the two electrophiles, the total yield of p₁TAM-H/p₁TAM-D was very low (1.6%, from the triarylmethanol 1) and the isolation was troublesome due to the presence of multiple byproducts.

Li et al.^[30] reported the seminal work for one-step synthesis of asymmetric TAM radicals by nucleophilic substitution of the TAM cations with various nucleophiles in aqueous solution. This approach was further extended to the synthesis of the monosubstituted TAM radicals in organic solvents^{[31, 32]}. In the present work, we attempted to synthesize p₁TAM-H and p₁TAM-D using this novel approach with moderate modifications [Fig. 2(b)]. Firstly, the triarylmethanol 1^[33-35] in tetrahydrofuran (THF) was partially deprotonated using n-BuLi (3 Equiv.) and TMEDA (3 Equiv.), followed by addition of DMC (30 Equiv.) to afford the diester 3 in a yield of 60%. Then, the diethyl phosphonated TAM radical 4 was obtained in a yield of 50% over two steps by CF₃SO₂OH-induced conversion of the diester 3 into the corresponding TAM cation, followed by selective nucleophilic reaction at the site of the aromatic H with sodium diethyl phosphite which was prepared by reaction of diethyl phosphite with NaH. It was worth noting that the reaction mixture mainly contained two compounds (i.e., the compound 4 and the radical form of the diester 3)^{[31, 32]}, thus enabling facile purification of the compound 4 by column chromatography on silica gel. Finally, the compound 4 underwent two continuous hydrolytic reactions using LiOH and TMSBr to give the water-soluble p₁TAM-H and p₁TAM-D in a yield of 85% over two steps, which were conveniently purified by column chromatography on reversed-phase C-18. The use of the relatively weak base LiOH instead of KOH improved the yield from 65% to 85%. Compared to the previously reported procedure^[25], our new approach exhibits the advantages of nearly 16-fold higher total yield (25% vs. 1.6% from the triarylmethanol 1) and more convenient purifications.

With p₁TAM-H and p₁TAM-D in hand, we started to synthesize their PEGylated analogues in order to improve their biocompatibility and eliminate their potential binding with albumins^{[28, 36]}. As shown in Fig. 3, The PEGylated TAM probes POP and dPOP were conveniently obtained in almost quantitative yields by SN2 reaction of the dicarboxylates 5-D/5-H with dD1/D1 in the presence of Cs₂CO₃ and catalytic amount of KI, followed by TMSBr-induced hydrolysis. The probes POP and dPOP were characterized by EPR, ¹H NMR and high resolution mass spectroscopy. POP: HRMS (MALDI-TOF), a broad peak from 3 100 to 4 500 (average, n∼7). ¹H NMR (400 MHz, CD₃OD), δ 4.24 (m, 12H), 3.64 ~ 3.74 (m, 149H), 3.53~3.55 (m, 12H), 3.36 (m, 18H), 2.59 (d, J=2.56 Hz, 12H). dPOP: HRMS (MALDI-TOF), a broad peak from 3 300 to 4 200 (average n∼7). ¹H NMR (400 MHz, CD₃OD) δ 4.25 (m, 12H), 3.61~3.73 (m, 149H), 3.53~3.56 (m, 18H), 3.36 (m, 18H), 2.59 (d, J=3.4 Hz, 12H).

2.2 EPR characterization: spectral sensitivity to pH

Fig. 4 shows EPR spectra of the probes POP and dPOP recorded in phosphate buffer (PB, 1 mol/L) at different pH values (8.6, 6.8 and 4.9) under anaerobic conditions. At pH 8.6, POP existed as a form of POP^2- with the phosphonate group being completely deprotonated and exhibits a doublet of quintet (2×5 lines) spectral pattern due to hyperfine splittings from phosphorus (α_P) in the phosphonate and four equivalent protons (α_H) in two ester linkers. EPR spectral simulation showed that POP^2- has the α_P and α_H values of 3.370×10^-4 T and 1.02×10^-5 T (Table 1), respectively. Similar spectral pattern was also observed for the partially deprotonated form POP^1- at pH 4.9 which has a larger α_P value (3.565×10^-4 T) but almost same α_H value (1.02×10^-5 T) compared to POP^2-. At nearly neutral pH (6.8), both of two ionization states are present in aqueous solution and their EPR signals are overlapped to a great extent due to relatively small difference of their α_P values (Δα_P=1.95×10^-5 T) and the presence of multiple proton hyperfine splittings. The 2D EPR analysis also revealed slightly different g factors as 2.003 53 for POP^2- and 2.003 55 for POP^1-. Comparatively, EPR spectral pattern of dPOP is much simple with a doublet signal from the phosphorus hyperfine splitting either at low or high pH [Fig. 4(b)], because the deuteration eliminates proton hyperfine splittings in the ester linkers. At neutral pH, the low field signals from dPOP^2- and dPOP^1- were distinguishable. However, the high field peaks of dPOP^2- and dPOP^1- were merged into one peak with slight line broadening most likely because the field difference (Δα_P) due to α_P values of two ionization states is nearly equal to the shift caused by Δg = 0.000 02.

In order to obtain the reliable pKa values, pH titrations of POP and dPOP in phosphate buffer (1 mmol/L) containing NaCl (150 mmol/L) were carried out using low modulation amplitude (2×10^-6 T) and modulation frequency (10 kHz) under anaerobic conditions. The pH-dependent fractions of POP^2-/POP^1- or dPOP^2-/dPOP^1- were obtained by simulating the individual EPR spectra. These fractions were then fitted by standard titration equation [Equ. (1)] to afford the corresponding pKa values of 6.82 for POP and 6.75 for dPOP (Table 1). In addition, the EPR spectra obtained were also simulated by the 2D EPR approach. This approach enables efficient and accurate determination of pKa by simultaneous fitting of all of spectra in each titration using a whole set of parameters (hfs, g-value, linewidth and pKa). The quality of fit can be significantly improved by taking into account the broadening caused by non-resolved hyperfine splitting of protons or deuteriums. As Fig. 5 shows, smooth fitting of the titration curves provides pKa values of 6.80 for POP and 6.79 for dPOP, which are slightly different from the values obtained by the traditional method. Comparison of pKa of p₁TAM-D (about 6.90)^[25] with that of dPOP (about 6.79) indicates that the esterification using the PEGylated dendron only slightly decreases pKa of the monophosphonated TAM radical.

2.3 Effect of dendritic PEGylation on EPR spectra of dPOP/POP and p₁TAM-D/H

Our previous studies showed that dendritic PEGylation endows TAM radicals with multiple advantages including negligible albumin binding, enhanced biostability and prolonged pharmacological kinetics in mice^[28]. To check if the PEGylated probes in this study have similar properties, we firstly investigated and compared the concentration-dependent effect of bovine serum albumin (BSA) on EPR spectra of dPOP and p₁TAM-D in PB (1 mmol/L, pH 7.4). As shown in Fig. 6(a), the EPR doublet signal of dPOP in ambient air seems to be almost symmetric due to moderate O₂-induced line broadening. Upon addition of BSA (0~500 μmol/L), EPR spectral profile of dPOP do not show any change, indicating no marked binding between them [Fig. 6(a)]. Comparatively, p₁TAM-D, like the TAM radical CT-03^[27], binds strongly with BSA as verified by severe distortion of its EPR signal due to the line broadening [Fig. 6(b)]. Interestingly, the low field peaks of both pH probes are gradually shifted to low field possibly because the addition of BSA led to the decrease of pH [see Fig. 6(a) and Fig. 6(b)]. This pH decrease induced by addition of BSA is further confirmed by pH electrode [Fig. 6(c)]. The high field peak intensity of dPOP only slightly decreases with the concentration of BSA and 92% of the peak intensity remained in the presence of 500 μmol/L of BSA while only 25% of the peak intensity can be identified for p₁TAM-D [Fig. 6(d)].

As shown in Fig. 7, the PEGylation modification also enhances the stability of POP toward various reactive species. Almost no signal decay was observed for POP after 30 min incubation with oxidants (e.g., superoxide radical, hydroxyl radical, peroxyl radical, hydrogen peroxide) and reductants (e.g., ascorbate and glutathione). On the contrary, the paramagnetism of p₁TAM-H was quenched to different degrees by these oxidoreductants, especially by superoxide radical and glutathione. Similar reactivity with these reactive species was also observed for other TAM radicals^{[20, 37-39]}.

2.4 pH/O₂ sensitivity of dPOP in the presence of BSA

To confirm the accurate measurement of pH using dPOP as the probe in the presence of BSA which may show great interference in applied system, its EPR spectra were recorded in PB (1 mmol/L) containing 500 μmol/L BSA. As shown in Fig. 8(a), EPR spectral profile of dPOP did not show any distorance compared with the according EPR spectra in the absence of BSA [Fig. 4(b)]. Plotting of fractions of the ionization states of dPOP as a function of pH using equation (1) provides a pKa of 6.79 for dPOP in the presence of BSA [Fig. 8(b)], which is same as the value that obtained in PB alone. Therefore, the PEGylation solves the albumin binding issue of p₁TAM-H/p₁TAM-D and guarantees the application of dPOP in albumin-rich biological systems such as blood. Moreover, EPR signal of dPOP also exhibits O₂-sensitive line broadening which is usually used to measure O₂ concentration. The ratio of the line widths under aerobic and anaerobic conditions is used to indicate the O₂ sensitivity of various probes as did in the previous study^[28]. As shown in Table 1, two ionization states of dPOP have similar O₂ sensitivity with the ratio value of about 1.5, which is higher than the corresponding value of POP (about 1.3). Interestingly, the O₂ sensitivity of dPOP (1.57 for dPOP^1- and 1.62 for dPOP^2-) slightly increases in the presence of BSA (500 μmol/L) possibly due to higher concentration of dissolved O₂ in the solution containing BSA.

3 Conclusion

In summary, p₁TAM-H and p₁TAM-D have been efficiently synthesized and gram-scale synthesis could be conveniently achieved, thus enabling their further derivatization. Unlike p₁TAM-D, its PEGylated derivative dPOP still exhibits good pH and O₂ sensitivity in the presence of high concentration of BSA (500 μmol/L). The PEGylation could also improve the biocompatibility and extend the blood circulation time of dPOP as observed in our previous study^[28]. Thus, dPOP could be administrated by intravenous injection and overcome the limitation of p₁TAM-D and the triphosphonated TAM probe which were injected intratissually^{[25, 40]}. Therefore, our present study provides a novel strategy for the design and efficient synthesis of new pH and O₂ dual EPR probes with improved properties which could significantly advance the application of in vivo EPR spectroscopy and the imaging in biomedical field.