Synthesis and Physicochemical Characterization of the Impurities of Pemetrexed Disodium, an Anticancer Drug

A physicochemical characterization of the process-related impurities associated with the synthesis of pemetrexed disodium was performed. The possibility of pemetrexed impurities forming has been mentioned in literature, but no study on their structure has been published yet. This paper describes the development of the synthesis methods for these compounds and discusses their structure elucidation on the basis of two-dimensional NMR experiments and MS data. The identification of these impurities should be useful for the quality control during the production of the pemetrexed disodium salt.


Introduction
Pemetrexed (1a, Figure 1) is an antifolate antineoplastic agent that exerts its action by disrupting folate-dependent metabolic processes essential for cell replication. It acts by inhibiting three enzymes used in purine and pyrimidine synthesis de novo-thymidylate synthase (TS), dihydrofolate reductase (DHFR), and glycinamide ribonucleotide formyltransferase (GARFT) [1,2]. By inhibiting the formation of precursor purine and pyrimidine nucleotides, pemetrexed prevents the formation of DNA and RNA, which are required for the growth and survival of both normal and cancer cells. A pharmaceutical product containing pemetrexed disodium (1a) as the active ingredient is used for the treatment of malignant pleural mesothelioma (MPM) in combination with cisplatin and as a second line agent for the treatment of advanced or metastatic non-small cell lung cancer (NSCLC). Currently, the drug is used as a single agent or in combination with other chemotherapeutic agents for the treatment of other types of cancer, such as breast cancer, bladder cancer, colorectal carcinoma and cervical cancer [3,4].
The product was originally developed by Taylor and co-workers [5] at Princeton University and is available on the market under the brand name ALIMTA ® (Lilly). It is a sterile lyophilized powder for intravenous infusion.
The U.S. Food and Drug Administration (FDA) [6] and the European Medicine Agency (EMA) [7] require complete physicochemical characteristic not only for an active pharmaceutical ingredient (API), but also for its key synthetic intermediates. In addition, the determination of a drug substance impurity profile, including known, especially pharmacopeial impurities [8], as well as other unknown impurities, can have a significant impact on the quality and safety of drug products.
The health implications of impurities can be significant because of their potential teratogenic, mutagenic or carcinogenic effects. Therefore, the International Conference on Harmonization (ICH) sets a high standard for the purity of drug substances [9]. If the dose is less than 2 g/day, impurities over 0.10% are expected to be identified, qualified and controlled. If the dose exceeds 2 g/day, the qualification threshold is lowered to 0.05%. It is therefore essential to control and monitor the impurities both in the APIs and the finished drug products. It is also a crucial issue in drug development and manufacturing.
This paper describes a study on identification, synthesis and characterization of the impurities formed during the pemetrexed disodium synthesis. The study will help to understand the formation of the impurities in the pemetrexed disodium synthesis and provide a clue on how to obtain a pure compound.

Synthesis of Pemetrexed Disodium
Convergent synthesis of pemetrexed disodium heptahydrate from key synthetic intermediates (Scheme 1) is well documented and involves firstly the preparation of the p-toluenesulfonic acid salt (5a) [10]. The acid 2 is activated for coupling by reaction with 2-chloro-4,6-dimethoxytriazine (CDMT) in presence of N-methylmorpholine (NMM) to form an active ester 3 and then reacted with diethyl L-glutamate 4. The product of peptide coupling 5 is isolated as p-toluenesulfonate 5a and then saponified to produce a free acid form of the drug substance (1). Finally, the pH is adjusted to pH 8 and the crystalline disodium salt 1a is isolated as the heptahydrate form (1a·7H 2 O).
However, we have found a new method for the preparation of pemetrexed disodium 1a in an amorphous form which involves the deprotonation of pemetrexed diacid (1) in the presence of sodium methoxide under anhydrous conditions [11]. During the study of process developing we observed that the product, pemetrexed disodium, contained a number of impurities; six of them were identified (Table 1).
At early stage of development process [11] we found that impurity level in some batches of pemetrexed varied in the range from 0.05% to 0.5% (HPLC). Further study was undertaken to find out if the measured impurities limits comply with the documentation requirements and if the product may be registered as the API. One of the requirements is to prove that the substance meets the characteristics described in the Pharmacopoeia, i.a., the impurities limits not exceeded specified values.
For routine process controls of batches of the API as well as for development of analytical methods the possession of impurities of appropriate quantity and quality is required. Some impurity standards may be obtained commercially (at a high cost), some of them, mainly those which have not been previously described, are not available commercially. In both cases the elaboration of methods for their synthesis is very advantageous.

Structure of Impurities
The compounds shown in Table 1 were taken into account as potential impurities of the final pemetrexed disodium: Only the impurities (R)-1 (Impurity E), 6 (Impurity A), 8 (Impurity D), 10 (Impurity B and C) have been documented in the European Pharmacopoeia [8]. A detailed synthetic processes and structure confirmation have not been reported for 8, 9, 10. Compounds 6 and 7 were mentioned previously, but without full characterization [12,13]. Detailed HPLC analyses for 6, 8, 10 were also described [14].
The impurities collected in Table 1 were synthesized and fully characterized by different techniques (IR, NMR, MS, HPLC, DSC).

Source and Preparation of Impurities
The N-methyl impurity 6 of pemetrexed is formed while condensing the benzoic acid 2 with diethyl L-glutamate 4 in the presence of CMDT and NMM. Kjell and coworkers [12] suggest that the decomposition of the excess CDMT·NMM complex produces a methylating agent which is capable of methylating the N1-nitrogen of dezazaguanine moiety and giving 6 as the final result.
In the first route, derivative 6 was generated by alkylation of 1a with methyl iodide in the presence of triethylamine, followed by the treatment with 1N NaOH and purification by preparative TLC chromatography on the silica gel. However, this simple attempt gave 6 with very low yield (only 5%). We have exploited a different concept for the synthesis of 6 by alkylation of diester 5a with methyl iodide in the presence of triethylamine to N-methyl diester 13 and its subsequent saponification (Scheme 2). Scheme 2. Syntheses of impurities 6 and 7. Conditions: (i) CH3I, Et3N, DMF; (ii) (a) NaOHaq, (b) HClaq; (iii) DMF-DMA, p-TSA, DMF.

N,N-Dimethylformamide Impurity 7
Recently, it has been reported [13] that when the condensation reaction between acid 2 and diethyl L-glutamate 4 is performed in the presence of N,N-dimethylformamide (DMF) as a solvent, it results in the formation 14 which subsequently after saponification results in the formation of impurity 7.
In order to prepare derivative 14, we adopted the procedure described in literature [13] which involved reacting 5a with dimethylformamide-dimethylacetal (DMF-DMA) in the presence of p-toluenesulfonic acid (p-TSA) at 60 °C. We found that high excess of p-TSA can be decreased from 10 to 1 equivalent (coming from p-TSA salt 5a). However, the addition of DMF and anhydrous conditions are required. After these modifications compound 14 was prepared in good yield and was converted to 7 by basic hydrolysis with 1N NaOH at room temperature followed by the acid treatment (Scheme 2) [15].

Enantiomer of Pemetrexed (R)-1
Compound (R)-1, mentioned in Pharmacopeia, is a D-enantiomer of pemetrexed disodium. The presence of this impurity can be detected by a chiral HPLC method [8]. It probably arises from the trace amounts of D-enantiomer in commercial diethyl L-glutamate 4 or is formed during the hydrolysis of the ethyl esters of 1 in an alkaline medium at ambient or higher temperature (e.g., >30°) [13]. This impurity was prepared starting from diethyl D-glutamate, following a synthetic process analogous to the synthesis of pemetrexed disodium (Scheme 1). Impurity (R)-1 is characterized by the same 1 H-NMR, 13 C-NMR and mass spectrum as pemetrexed disodium.

γ-Dipeptide Impurity 8
In the course of our investigation on the pemetrexed disodium synthesis, we realized that the source of impurity 8 is α-ethyl L-glutamate 11 which can be present in the starting diethyl L-glutamate 4. We envisioned that triacid 8 could result from the condensation of monoester 11 with the acid 2, followed by the formation of 19 through coupling α-monoester of pemetrexed 18 with 4 and the saponification of the ethyl ester groups.
Our synthesis of impurity 8 includes first the preparation of α-ethyl L-glutamate 11. The N-protected α-ethyl ester 17 was isolated as the dicyclohexylammonium salt and purified by crystallization [16,17]. The decomposition of the salt with sulfuric acid and the removal of benzyloxycarbonyl group (Cbz) by catalytic hydrogenolysis finally gave 11.
The conversion of 11 to 19a was achieved through conventional peptide coupling with CDMT/NMM and the subsequent formation of the p-TSA salt. The final hydrolysis of the triester 19a with 1 N NaOH followed by the acid treatment provided triacid 8 (Scheme 3).

α-Dipeptide Impurity 9
Similarly, if starting diethyl L-glutamate 4 contains some amount of γ-ethyl L-glutamate 12, then during the preparation of pemetrexed disodium α-dipeptide impurity 9 may appeared. At the beginning for the synthesis of dipeptide 9 a similar synthetic procedure to that described above for the preparation of dipeptide 8 was used (Scheme 4).
The reaction of γ-ethyl ester 12 (contamination of the main compound 4) with 2 led to γ-ethyl ester 20, which after further condensation with 4 gave p-toluenesulfonate 21a. This compound, after hydrolysis with NaOH at ambient temperature followed by acidification with HCl, gave triacid 9. Surprisingly, the HPLC analysis of 9 revealed the presence of two separated equivalent peaks. For compound 8 such phenomenon was not observed. LC-MS showed that the molecular weight for both peaks of 9 was the same. In the NMR spectra of 21a and 9 we also observed doubled signals (for details, see Section 2.4).
One possible explanation of the observed data is the formation of diastereoisomeric mixture. Most probably monoester 20 (in fact the substituted benzoyl-(S)-α-amino acid) undergoes a racemization during the activation/coupling step with diester 4 (Scheme 4, step iii) [18], thus a mixture of diastereoisomeric triesters 21 (S,S-and R,S-configuration) is formed which after hydrolysis give the mixture of diastereoisomeric triacids 9 (S,S-configuration and R,S-configuration), respectively.
A simple change in the coupling conditions to eliminate racemization in coupling 20 with amine 4 failed. Using HATU [19] instead of CDMT/NMM also resulted in the mixture of diastereoisomeric triesters 21. In order to resolve this problem and obtain standard samples, we decided to prepare independently both diastereoisomers of impurity 9 (S,S-and S,R-) employing a different synthetic route (Scheme 5).
In this approach N-Cbz-protected (S or R) glutamic acid derivatives were used for coupling leading to dipeptides 22 (S,S-and S,R-) which were subsequently converted into 9 (S,S-and S,R-). It is well known that contrary to N α -acyl protected amino acids which racemize readily during the activation/coupling of the carboxyl group for the amide bond formation, in the case of the urethane-type amine protecting groups (as Cbz) the tendency of racemization is largely suppressed [20]. The γ-monoesters 12 (with S-or R-configuration, respectively) were N-Cbz-protected [21] and then coupled with diester 4 to obtain enantiomerically pure protected dipeptides 22 (S,S-and S,R-). After cleaving the Cbz-group, each of the enantiomerically pure amines was coupled with acid 2 to give both diastereoisomeric triesters 21, which after hydrolysis led to diastereoisomeric triacids (S,S)-9 and (S,R)-9, respectively.
The NMR, MS and HPLC analyses confirmed that (S,S)-9 and (S,R)-9 were pure single compounds (in HPLC designated by different retention times 27.0 and 26.6 min for (S,S)-9 and (S,R)-9, respectively).

Dimer Impurity 10
Dimeric impurity 10 (as diasteroisomeric mixture) of pemetrexed might be formed during the basic hydrolysis of 5a to the pemetrexed disodium salt. According to the European Pharmacopoeia [8], in order to prepare impurities 10, pemetrexed disodium is dissolved in 0.1 M NaOHaq and heated at 70 °C for 40 min. After cooling to room temperature, the mother solution was diluted with water to obtain the "reference solution".
We have modified this procedure and developed a purification method for impurities 10. Pemetrexed disodium was dissolved in 0.1 M NaOHaq and heated under reflux for 3 days (TLC control). Then, the mixture was cooled and 10% HClaq was added to adjust pH ≈ 3. A formed precipitate was filtered and purified by chromatography to get the mixture of 10. The structure of the investigated compound was confirmed by NMR (discussed hereinafter).

Structure Elucidation by Analytical Methods
The structures of all studied impurities were identified using the results of various 2D NMR spectra, including the COSY, 1 H-13 C/ 15 N gradient selected HSQC, as well as 1 H-13 C/ 15 N gradient selected HMBC sequences.
Enantiomeric impurity (R)-1: compound (R)-1 in a diacid form as well as the disodium salt are the enantiomers of the main compound 1. The comparison of the multinuclear NMR spectra recorded for (R)-1 with those registered currently and published earlier for (S)-1 [5,22] undoubtedly confirm the structure of the enantiomeric (R)-1 ( Table 2).
N-Methyl impurity 6: The analysis of the results of different NMR spectra, especially 1 H-15 N HMBC experiment, strongly supports the proposed structure 6. The most significant effect was observed for the 15 N-NMR chemical shifts. Unfortunately, compound 6 (acid form) forms a gel in the DMSO solution and that is why we used NMR data of diethyl ester 13 for comparison. In the 1 H-15 N g-HMBC spectrum of 13 a strong correlation peak was noticed between the protons of CH3 group at δ = 3.56 ppm and nitrogen at δ = −274.2 ppm. This "cross-peak" identifies the position of the methylation which is an N3 nitrogen atom. Additionally, in the 1 H-13 C HMBC experiment CH3 protons (introduced onto nitrogen atom) correlate with two carbons (δ = 152.1 (C2) and 139.3 ppm (C4)) and as a consequence strongly support the place of the methylation.
Introducing a methyl group onto the N3 nitrogen atom causes a strong shielding increase noticeable for N3 and C4 nuclei by ca. 66 and 12 ppm, respectively. Similar shielding effects at C4 are visible in the case of p-toluenosulfonic salts (R)-5a, 21a and 19a, (Experimental), which explains the protonation site of the neutral compounds at the N3 atom. The NMR data for 6/13 are given in Table 2.
N,N-Dimethylformamidine impurity 7: The sets of the 1 H/ 13 C-NMR chemical shifts (whose assignment comes from the analysis of the results of more advanced 2D experiments, including 1 H-13 C HSQC/HMBC and 1 H-15 N HSQC/HMBC correlations) unambiguously confirm the presence of an imine part in the molecule of the obtained compound and thereby the structure of impurity 7. The most important and significant effect was observed for the nitrogen nucleus at the C2 atom. The 15 N shielding decrease of ca. 130 ppm between the exocyclic nitrogen at C2 in the structures of 1 and 7 is responsible for the exchange of the nitrogen atom character from the amino to imine group. The comparison of 1 H/ 13 C-and 15 N-NMR data for 1 with that obtained for 7 (change of the NH2 group to N=CH-N(CH3)2), for which the 1 H-NMR spectra are described in literature [13], leads to the observation of a few shielding/deshielding effects on the atoms in close neighborhood of the replacement. The most important one is carbon C2 deshielded by ca. 3 ppm when compared with its position in 1. Moreover, a quite surprising deshielding effect is noticeable at the N1/H1 pair ( Table 2). The H1 proton is deshielded by ca. 0.6 ppm, whereas in the case of nitrogen N1 the same effect is stronger by ca. 13 ppm. Other more distant atoms also experience deshielding effects. For carbons C5, C8 and nitrogen N9 these changes are ca. 2 ppm, 1.5 ppm and 1.5 ppm, respectively ( Table 2).   (Table 2).

γ-Dipeptide impurity 8 and α-dipeptide impurities 9:
The NMR spectra and HPLC analysis of compounds 8 and 9 showed some unexpected results. As mentioned before, in the case of impurity 8 the HPLC method revealed one compound, whereas for impurity 9 two peaks in HPLC were observed. A detailed analysis of the 2D-NMR experiments for 8 taken in the DMSO solution leads to the 1 H/ 13 C-and 15 N-NMR chemical shift assignment and complete confirmation of the structure of this impurity ( Table 3).
The same detailed analysis of the 2D-NMR data was performed for 9, confirming the structure of this impurity, although a doubled set of 1 H/ 13 C-NMR and even 15 N-NMR signals was noticed there.
At the beginning the observation of the doubled 1 H/ 13 C-NMR signals forced us to check if we were not dealing with rotamers. A well-known method employing 1 H-NMR spectroscopy [23] was useless, because the 1 H-NMR signals overlapped. We decided to raise the temperature and observe the 13 C-NMR spectrum. However, the temperature increase (between 25-100 °C or 298-373 K) resulted in no significant changes in the 1 H/ 13 C-NMR spectra and based on these experiments we ruled out the hypothesis of the rotamers presence.
Another possible explanation of the observed data was the formation of the diastereoisomeric mixture during the synthesis. To prove this hypothesis, both diastereoisomers of impurity 9 (S,S-and S,R-) were independently synthesized employing a substantially racemization-free synthetic route based on the use of the N-Cbz protecting group described in Section 2.3.5 (Scheme 5).
The NMR, MS and HPLC analysis of the synthesized compounds confirmed that (S,S)-9 and (S,R)-9 were pure single diastereoisomers characterized by a single set of the 1 H/ 13 C-and 15 N-NMR signals. The 1 H/ 13 C-NMR spectra recorded for both diastereoisomeric compounds (S,S)-9 and (S,R)-9 were very similar to each other, with minor differences in the narrow ranges of the 1 H/ 13 C-NMR chemical shifts (Table 3). These are visible for the C16-C28 chain and especially for the H17-H20 and H23-H26 protons. Table 3. NMR Data for compounds 8, (S,S)-9 and (S,R)-9 with the correlations observed in the HSQC and HMBC spectra.
All the results strongly support the hypothesis that during the synthesis of impurity 9 according to Scheme 4, the coupling of 20 with 4 leads to the diastereomeric mixture of 22.
Diasteroisomeric dimer impurity 10:This impurity was mentioned in Pharmacopeia but has not been reported in literature. The HR mass spectrum of 10 showed the exact mass m/z = 867.2709, perfectly corresponding with the Pharmacopeia structure. Yet more convincing proof of the structure of 10 comes from the analysis of various NMR experiments. The 1 H/ 13 C spectra for this impurity in DMSO showed double sets of signals which may be considered as an existence of two structurally identical fragments. This observation and further analysis of the long-range 1 H-13 C and 1 H-15 N g-HMBC correlations leading to 1 H/ 13 C-and 15 N-NMR signals assignment, indicates that both "pemetrexed" parts in this structure are connected with each other via the C7/C8 bond.
One part of this dimeric molecule looks identically as the pemetrexed molecule (both C7 δ = 114.2 ppm and C8 δ = 123.0 ppm carbons are aromatic), whereas in the second part carbons C7 and C8 change their character in such a way that C7 (δ = 51.7 ppm) becomes more aliphatic, while the C8 nucleus (δ = 179.6 ppm) is much more deshielded than when they are positioned in the "pemetrexed" part ( Table 2). The above mentioned effects are as follows: the shielding increase of ca. 62 ppm and at the same time the shielding decrease of ca. 55 ppm for C7 and C8, respectively.
The purity of the examined compounds was determined using HPLC/UHPLC methods with the chromatography system UltiMate™ 3000RS UHPLC (Dionex Corporation, Sunnyvale, CA, USA) equipped with an autosampler and a DAD 3000RS detector.
Method A: Gemini C18 column (150 mm × 4.6 mm, 3 µm; Phenomenex, Torrance, CA, USA) was placed in a thermostated column heater at 25 °C. The mobile phases consisting of A (4 g/L dipotassium hydrogen phosphate; pH 5.2) and B (acetonitrile) were used with the gradient mode at the flow rate of 0.9 mL/min. The samples were prepared at a concentration of about 0.5 mg/mL and were diluted in 0.4 g/L of dipotassium hydrogen phosphate. The injection volume was 10 µL. The UV detection at 230 nm was used.
The TLC separations were performed on the TLC silica gel 60 F254 on alumina sheets (Merck and/or Sigma-Aldrich). The visualization was performed by UV light (254 and/or 365 nm) and [24].
The specific rotation [α]D was calculated from an optical rotation measurement performed on the Perkin Elmer 341 Polarimeter (PerkinElmer, Waltham, MA, USA) at the wavelength of 589 nm (D line of a sodium lamp), at 20 °C.
The melting points were determined by differential scanning calorimetry (DSC) carried out by means of the DSC822 with an IntraCooler (Mettler Toledo GmbH, Schwerzenbach, Switzerland).  (13) To the solution of 5a (6.0 g, 9.16 mmol) in DMF (40 mL) triethylamine was added (3.20 mL, 22.96 mmol), followed by methyl iodide (2.06 mL, 33.09 mmol) and the solution was left at room temperature for 72 h. Then CH2Cl2 (80 mL) and water (80 mL) were added. The layers were separated and the aqueous layer was extracted with CH2Cl2 (1 × 40 mL).The combined organic layers were dried over anhydrous MgSO4 and concentrated. The residue was dissolved in MeOH (10 mL) and i Pr2O (50 mL) was added. The resulted precipitate was filtered, washed with i Pr2O (2 × 10 mL), and dried to give 13  (6) Compound 13 (3.50 g, 7.03 mmol) was treated with 1M NaOHaq (30 mL) and stirred at RT for 2 h. The reaction mixture was diluted with EtOH (30 mL) and water (30 mL) and adjusted to pH 3.0 with 1 M HCl. The resulting slurry was heated to 60-65 °C and then cooled to RT. The solid was filtered, washed with EtOH (2 × 10 mL) and dried in vacuo at 40 °C for 24 h. Crude 6 was purified by flash chromatography on a silica gel column using CH2Cl2/MeOH/H2O/NH3 as the mobile phase (40:40:5:2 v/v). The respective fractions were collected and concentrated. The residue was dissolved in water (50 mL) and the pH was adjusted to 2-3 with 1 M HCl. EtOH (220 mL) was added and stirred for 30 min. The suspension was filtered and the solid was washed with EtOH/H2O (20 mL) and dried at 40 °C to obtain 6 (2.51 g, 81%, HPLC purity 99.05%).

Synthesis of Impurity
EtOH (180 mL) was added to the oil, followed by the solution of the p-toluenesulfonic acid monohydrate in EtOH (15.96 g in 180 mL) and the resulting suspension was heated under reflux for 2 h. The mixture was cooled to RT, the crystals of (R)-5a were filtered and washed with EtOH (2 × 60 mL). The wet cake was reslurried in EtOH (400 mL), refluxed for 1 h and cooled to RT. The crystals were filtered, washed with EtOH (2 × 60 mL) and dried in vacuo at 40 °C for 24 h to provide (R)-5a (14. Compound (R)-5a (14.4 g, 21.98 mmol) was treated with 1 M NaOHaq(112 mL), the mixture was stirred at room temperature. After 1 h the reaction mixture was adjusted to pH 8.0 with 1N HClaq and heated to 55-60 °C. EtOH (560 mL) was added to the solution. After cooling to RT, the precipitated solid was collected by filtration and washed with EtOH (2 × 80 mL). The wet solid (12.84 g) was dissolved in water (120 mL) and the solution was heated to 55-60 °C. EtOH (500 mL) was added and then the mixture cooled to RT. The solid was filtered, washed with EtOH (2 × 80 mL) and dried in vacuo at 35 °C for 48 h to provide (R)-1 (9.8 g, 87%, 99.4% total HPLC purity, 99.9% chiral HPLC purity). HRMS calcd for C13H15NO6Na m/z = 304.0797, found m/z = 304.0799.
Acid 2 (296 mg, 0.99 mmol) was dissolved in DMF (10 mL), then DIPEA (0.52 mL, 2.97 mmol) was added followed by HATU (498 mg, 1.29 mmol) and the resulting solution was stirred at RT for 30 min. Then the solution of the previously obtained amino-triester (500 mg, 1.39 mmol) in 10 mL DMF was added and the resulting mixture was stirred overnight.

Conclusions
Herein we have developed and described the synthesis of the process-related impurities of pemetrexed disodium, the active ingredient of an anticancer therapeutic. The process related impurities 6, 7 and 10 we synthesized by modified methods. For the impurities 8 and 9 we developed new synthetic methods. We found that during synthesis the impurity 9 a mixture of diastereoisomers can be formed. To avoid this process, we have developed an effective method of the synthesis of 9 where racemization does not occur. Two diastereoisomeric impurities 9 were obtained: S,S-9 and, reported for the first time, S,R-9.
The structure elucidation of all obtained impurities was discussed on the basis of two-dimensional NMR experiments and MS data and their physicochemical characterization was presented. We have developed HPLC methods for the determination of chemical and enantiomeric purity of pemetrexed disodium and its impurities.