Photodegradation of methotrexate in aqueous solution: degradation kinetics and identification of transformation products

Methotrexate is an antineoplastic folate analog of high environmental concern, due to its low biodegradability and toxicological properties. This study focused on its photodegradation under two irradiation conditions, aiming to be representative of environment (300–450 nm) and drinking water treatment (254 nm). The photodegradation experiments were conducted at two pH, to vary the methotrexate ionization state and to produce a large variety of transformation products (TPs). The degradation kinetics determined through LC-UV monitoring were contrasted according to pH and irradiation wavelength. However, the quantum yields were independent of ionization state at 254 nm and the changes in kinetics at higher wavelengths were attributed to a change in the degradation mechanism. The TPs formed during the reactions were identified by UHPLC-MS/MS, using both the positive and negative modes. Among the eleven proposed structures, five were described as methotrexate TPs for the first time. The TPs result from N-demethylation, glutamic acid oxidation, and C–N cleavage, all of them leading to further degraded photoproducts presenting modified or lost glutamic acid part. This was made possible thanks to the negative mode, which allowed the exploration of the glutamic acid moiety modifications. Cytotoxicity assessment on A549 cancer cells demonstrated that all photoproducts formed at pH 7 were less toxic than the parent compound.


Introduction
Pharmaceuticals and their degradation products are detected in surface waters, wastewater, and groundwater at a range of ng/ L to mg/L (aus der Beek et al. 2016;Morasch et al. 2010) and even in drinking waters (Benotti et al. 2009;Mompelat et al. 2011). Because more than 2000 active ingredients are approved in Europe, Perazzolo et al. (2010) proposed a priority list of 57 compounds for pharmaceutical risk assessment. This list was based on screening of their potential harm, analytical feasibility, and previous detection in water. These occurrences are mainly due to their pharmacokinetics since pharmaceuticals are excreted by patients either as the parent compound (60-95% - Turci et al. 2003) or as metabolites before entering domestic wastewaters. Discharges from municipal sewage effluents are considered as the main source of pharmaceuticals in the environment (Garcia-Ac et al. 2009a) since they represent 80% of their release against 20% for hospital wastewaters (Besse et al. 2012). Castiglioni et al. (2005) detected 19 pharmaceuticals in concentrations ranging from 0.5 to 2000 ng/L in Waste Water Treatment Plant (WWTP) and Vergili et al. (2019) detected 21 and 32 compounds in urban wastewaters from Turkey and Germany, respectively. Parent compounds, metabolites, and transformation products (TPs) formed by oxidative techniques such as ozonation, UV/ hydrogen peroxide treatment, and chlorination can enter in surface waters due to their incomplete degradation during wastewater treatment. A lack of knowledge remains about the ecotoxicity of certain pharmaceuticals (Orias and Perrodin 2013). In addition, Escher and Fenner (2011) and Responsible Editor: Christian Gagnon González-Burciaga et al. (2021) have shown that awareness about TP and metabolites can be relevant for environmental risk assessment because they can be more persistent, more mobile, or have higher toxicity than the parent compound.
Among various classes of pharmaceuticals, antineoplastic agents are of particular environmental concern due to their highly potent mechanism of action: they are cytotoxic, genotoxic, mutagenic, and teratogenic (Zounková et al. 2007). They have been detected in source waters at pg/L range for oxazaphosphorines for example (Buerge et al. 2006). The most investigated anticancer drug is cyclophosphamide (Besse et al. 2012), but another anticancer drug, methotrexate ((2S)−2- [(4-{[(2,4-diaminopteridin-6yl)methyl](methyl)amino}benzoyl)amino] pentanedioic acid, MTX), is one of the most widely prescribed antineoplastic agents worldwide (Kosjek and Heath 2011). It has been used for more than 50 years (Lai et al. 2017) in treatments of meningeal leukemia, psoriasis, and rheumatoid arthritis (Sabry et al. 2003). It competitively inhibits dihydrofolate reductase, the enzyme that reduces dihydrofolates to tetrahydrofolates. The reduced folate derivatives are important coenzymes in the biosynthesis of amino acids and the purine and pyrimidine units of DNA (Breithaupt et al. 1982). Perazzolo et al. (2010) have selected MTX on its priority list for pharmaceutical risk assessment.
Numerous studies detected MTX in municipal WWTP influents with concentrations ranging from 2.6 to 59 ng/L in Canada (Garcia-Ac et al. 2009b) and Spain (Martín et al. 2014;Negreira et al. 2014). Castiglioni et al. (2005) detected MTX at 12.6 ng/L in one of eight samples of WWTP effluent, showing partial removal of this contaminant. Around hospital areas, the level of concentration could be higher: in hospital effluent, maximum concentrations can reach concentration levels of 4689 ng/L in China (Yin et al. 2010), around 4756 ng/L in Spain (Olalla et al. 2018), 28,700 ng/L in France (Catastini et al. 2008), and 300 ng/L in Taiwan (Lin et al. 2014). These high concentrations are not only related to hospital with specialized oncology area, because MTX is also used for other treatments. Catastini et al. (2008) detected even 2000 ng/L in WWTP influent, which was connected to hospitals and 30 ng/L in the WWTP effluent, showing incomplete removal by the wastewater treatment process. Thus, MTX and some known transformation products could enter in the environment, but there is a lack of knowledge about their occurrence in drinking and surface waters.
MTX can be partially degraded through various processes, particularly during wastewater or drinking water treatment. Kiffmeyer et al. (1998) reported the biodegradation of MTX in 8 days by an activated sludge system with an efficiency of 95% and identified 7-OH-MTX as the main TP. The concentration of this TP increased with the same rate as the MTX concentration decreased, indicating that it did not undergo further biodegradation. Recently, Isidori et al. (2016) observed the transformation of MTX into 7-OH-MTX in hospital effluent transiting to a WWTP by potential biotic conversion. Kosjek et al. (2015) have shown that MTX was readily biodegradable by activated sludge and transformed into several TPs. They have identified 8 new TPs resulting from demethylation, oxidative cleavage of amine, cleavage of C-N bond, aldehyde to carboxylate transformation, and hydroxylation. DAMPA was the most abundant and persistent TP. These results contrast with a previous study by Henschel et al. (1997) with OECD 301 biodegradability test, which classified MTX as not readily biodegradable. After using the same OECD standardized procedure, Lutterbeck et al. (2015) proceeded to photo irradiation at 200-400 nm wavelengths on the biodegradation mixture. They identified nine products including 7-OH-MTX and DAMPA, indicating that they did not undergo further degradation. Calza et al. (2014) identified seven photoproducts resulting from irradiation at 360 nm, with transformation involving decarboxylation and hydroxylation or the molecule cleavage. The addition of titanium dioxide favored a supplementary transformation pathway by the elimination of succinic acid from MTX. Kanjal et al. (2020) compared kinetics of photodegradation with and without peroxymonosulfate and identified common TPs with Calza et al. (2014). Lai et al. (2017) studied the role of bicarbonate anions in MTX photocatalysis by UV-TiO 2 treatment. Transformations included the addition of atomic oxygen, hydroxylation, deamination, C-C cleavage, and C-N cleavage of the aniline moiety, this last one being proposed as the primary degradation pathway. Upon the 16 TPs identified in this study, six were new whereas the others were common with previous photodegradation studies (Calza et al. 2014;Lutterbeck et al. 2015;Kanjal et al. 2020). Hsu et al. (2019) studied the self-sensitized degradation mechanism and pathways through direct photolysis and identified 10 TPs common to the study by Lai et al. (2017). Roig et al. (2014) identified chlorination products during the treatment water process. Some other studies focussed on degradation kinetics without identifying TPs, using various processes as catalytic ozonation (Alinejad et al. 2019), UV and gamma irradiation  or hydrolysis (Toński et al. 2021).
Regarding toxicity assessment, few studies report bioassays with Vibrio Fischeri, showing a decrease in toxicity under direct photolysis (Calza et al. 2014;Lutterbeck et al. 2015), while an increase in toxicity was seen during photodegradation in presence of bicarbonate anions (Lai et al. 2017). Alinejad et al. (2019) reported a decrease in cytotoxicity as Muneer et al. (2020), who reported also a decrease in mutagenicity after gamma and UV treatments.
In summary, various studies focussed on the MTX photodegradation, but none compared its reactivity in terms of degradation kinetics and/or transformation pathways, according neither to the wavelength nor the pH. Furthermore, various TPs issuing from MTX through abiotic (photodegradation or various AOPs) and biotic processes were evidenced and presented a high variability according to the studies.
The goal of this research was to disclose the MTX photodegradation under conditions occurring (1) in natural surface waters, thus receiving solar light (using 300-450 nm emitting lamps, i.e., using the photoactive part of the solar spectrum) and (2) during drinking water treatment, thus using 254 nm irradiation. Focusing on direct photolysis processes, experiments were realized in pure water to allow the identification of TPs. The pH was varied (pH 3 and 7) to study changes in reactivity according to the two ionization states of MTX. In addition, cytotoxicity assessment of TP mixture was conducted during the photodegradation experiment at pH 7 under both irradiation conditions. This study aimed to complement the existing knowledge since the photodegradation of MTX is not documented at low wavelengths, focussing on the degradation kinetics of MTX, the identification of TPs, and the evolution of cytotoxicity following MTX photodegradation. Our analytical strategy included the usage of mass spectrometry in both positive and negative modes to have multiple arguments for structural identification of the TPs.

Photodegradation experiments
The degradation experiments were conducted using an initial MTX concentration of 40 μM by dilution from an alkaline aqueous stock solution prepared every 2 weeks. Selecting a concentration greater than environmental ones allowed monitoring the parent evolution, detecting and identifying the TPs without sample pre-concentration. The pH was set at 7 or 3 by the addition of 0.05 M HClO 4 or NaOH. The solutions were kept in the dark before the experiments. During reactions, samples were withdrawn at specified times and stored in the dark at 10°C until their analysis by liquid chromatography HPLC-UV.
Two wavelength conditions, representative of natural conditions in surface water (300-450 nm) and drinking water treatment (254 nm), were used. For the 300-450 nm treatment, 150 mL of solution was placed in a 3-cm internal diameter tubular pyrex reactor at the center of a cylindrical device equipped with six lamps emitting within 300 to 450 nm with a maximum emission at 365 nm (TLD 15 W, Philips). A fan located outside the bottom of the tube was used to remove heat from the reactor, thus allowing the experiments to be conducted at an ambient temperature of 20−22°C. The fan and lamps were turned on for at least 30 min before the start of each experiment. Light intensity measured by a radiometer (VLX-3W, Vilber Lourmat, Marne-la-Vallée, France) was for each of the six lamps ca 0.1 mW/cm 2 at 312 nm and 1.2 mW/cm 2 at 365 nm. The MTX solution was bubbled by a 150 mL/min flow of air during 20 min before exposition and during the irradiation (Fig. S1). For the 254-nm treatment, a quartz cylindrical 2-cm internal diameter tubular photoreactor containing 60 mL solution was placed parallel to a monochromatic low-pressure mercury lamp (TUV 15 W, Philips) outputting a 254 nm light intensity of 1.5 mW/cm 2 at the photoreactor location. The solution was mixed by magnetic stirring (Fig. S2). To determine the quantum yields (see calculations in SI), the UV spectra of initial solutions were recorded in quartz cuvettes using a spectrophotometer (UV-2401 PC, Shimadzu).

Analytical determinations
High-performance liquid chromatography with ultraviolet detection (HPLC-UV) for MTX monitoring The reaction monitoring was performed by HPLC-UV to establish rapidly the MTX degradation kinetics, before HPLC-MS (/MS) to characterize the photodegradation mixture in terms of TPs. The HPLC-UV was carried out using a Waters Alliance (2690/5) system on an HPLC C18 Gravity-SB column 150 × 3 mm, 3 μm particle size (Macherey-Nagel). The gradient profile started at 97% water with 0.05% formic acid and linearly progressed up to 50% in 30 min, the second solvent being acetonitrile, at a constant flow rate of 0.5 mL/min. The detection was performed using a diode-array detector (TUV 2489, Waters) acquiring data in the 230-305 nm range. Data were acquired and treated using Empower software. Quantifications were performed at 305 nm, using the 40 μM initial solution (mean of 4 injections for each experiment) as 100% abundance. All the peak areas were expressed in percentage of initial MTX. The MTX kinetics of degradation were treated as a first-order form. The apparent first-order rate constants k' were evaluated according to Ln(C/C 0 ) = −k't, where C is the MTX concentration at time t and C 0 its initial concentration.

UHPLC-ESI-MS for TPs monitoring
The TPs were monitored by ultra-high-performance liquid chromatography (UHPLC)-mass spectrometry (MS). The device consisted on an Acquity UPLC system (Waters, Milford, MA, USA) composed of a high-pressure binary solvent manager, an autosampler, and a column oven connected to a triple q u a d r u p o l e m a s s s p e c t r o m e t e r ( T Q D , W a t e r s ) . Chromatographic separation was carried out on a UPLC C18 Gravity-SB Column, 100 × 2 mm, 1.8 μm particle size (Macherey-Nagel). The gradient profile started at 97% water with 0.05% formic acid for 5 min, linearly progressed to 50% at 30 min, with a flow rate of 0.3 mL/min. The UHPLC effluent was directed to a UV detector (TUV, Waters) programmed at 305 nm, prior to its introduction into the mass spectrometer via an electrospray source operating either in positive or negative mode. The electrospray conditions were as follows: capillary voltage 3 kV, cone voltage 40 and 54 V respectively under positive and negative modes, source temperature 150°C, desolvation temperature 450°C, extractor voltage 3 V, cone and desolvation gas flow (nitrogen) at 20 and 900 L/h, respectively. LC-MS studies were carried out in full-scan MS over the mass range of m/z 120-550. For quantitation, the integrations were performed on pseudo-molecular ions and the limit of quantitation using a signal-to-noise (S/N) of 5 was estimated to be 200 μg.L −1 corresponding to 1% of residual MTX.

Pre-concentration by solid-phase extraction (SPE)
To perform the structural determinations of the TPs, a preconcentration was necessary. Each reaction was stopped at an irradiation time selected as a compromise between the variety of observed TPs and their relative abundance. The obtained solutions were then submitted to solid-phase extraction (SPE), using various conditions to maximize the TP extraction on the lipophilic/hydrophilic balance Oasis HLB (200 mg, 6 mL) cartridge. Na 2 EDTA was added to the samples with a suitable volume of a 0.1 M solution to achieve a final concentration of 0.1% (gram Na 2 EDTA per gram solution) to improve the extraction efficiency of target compounds according to Ferrando-Climent et al. (2013). The elution was done with 4 mL of methanol.

UHPLC-ESI-MS/MS for identification
Structural elucidation of TPs was achieved with an ultra-highperformance liquid chromatograph Ultimate 3000 (Thermo Fisher Scientific) coupled to a Q Exactive hybrid quadrupole-Orbitrap™ (Thermo Fisher Scientific) mass spectrometer through a Heated Electrospray Ionization (HESI-II) probe. The chromatographic separation of the samples was performed on the same UPLC column and the same gradient as previously in UHPLC-ESI-MS. The analyses were performed in positive and negative modes under the following specific conditions: spray voltage, ± 3 kV; sheath gas pressure: 35 psi, auxiliary gas flow rare: 10 (arbitrary units); ion transfer tube and vaporizer temperature 300°C. Nitrogen was employed as sheath and auxiliary gas.
MS data were acquired in full scan within mass-to-charge (m/z) range of 120-550 m/z at a mass resolution of 70,000 and a scanning speed of 3.7 scan/s. It allowed detecting the pseudo-molecular ions [M + H] + and [M -H] − of the various compounds, and thus their elemental composition. The pseudo-molecular ions were fragmented by higher-energy collisional dissociation (HCD) with normalized collision energy of 35% and then diriged to the Orbitrap where they were analyzed with a resolution of 35,000. Xcalibur™ software (Thermo Fisher Scientific) was employed for instrumental control and data processing.
The cytotoxic activity of MTX was investigated using the M T S ( ( 3 -( 4 , 5 -d i m e t h y l t h i a z o l -2 -y l ) -5 -( 3carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium)) method (Promega): 2.5 10 3 cells (100 μL) were seeded in 96well plates and incubated overnight at 37°C in the presence of 5% CO 2 and 95% hygrometry. Then, the cells were treated with 100 μL of MTX at different final concentrations (from 50 μM to 0.05 nM). After 72 h, the media was aspirated and 100 μL of an MTS solution (10% in media) was added to each well. After a 3-h incubation period at 37°C in the presence of 5% CO 2 and 95% hygrometry, the optical density was measured at 490 nm wavelength using a microplate reader (Infinite M200 Pro, Tecan Trading AG, Switzerland). Untreated cells were used as control. Each concentration was tested in six replicates and the experiment was fulfilled in triplicates. The concentration inhibiting 50% of the cell proliferation (IC 50 ) was determined using GraphPad Prism (GraphPad Software Inc, San Diego, CA, USA). To evaluate the cytotoxicity evolution along the MTX degradation at pH 7, different time-points were chosen for the two irradiation conditions: treatment at 254 nm, time-points: t = 0, 60, 180, and 360 min; treatment at 300-450 nm, time-points: t = 0, 60, 120, and 180 min. Each sample was then evaluated using the method described above with 100 μL of each solution at the final concentration of 100 nM (close value of the IC 50 determined for MTX).

Kinetics of MTX degradation and evolution of TPs
When the MTX aqueous solutions at pH 7 or 3 were placed in the dark, they remained stable for at least a 24-h period. Irradiated at 254 nm during 3 h, the MTX degraded up to ca 60% and 73% at pH 7 and 3, respectively (Fig. 1a). The reactions being conducted in MilliQ water without any photoinducer can be interpreted as direct photolysis. The kinetics followed a pseudo-first-order with regression coefficients greater than r 2 of 0.99, and the constants were determined as 8.8 ± 1.6 10 −5 s −1 and 12.4 ± 0.7 10 −-5 s −1 (mean ± standard deviation, n = 3) at pH 7 and 3, respectively (Fig. 2a). The significantly faster reactivity at pH 3 resulted from a different ionic form, as MTX is a polyelectrolyte with 3 pK a at 3.36, 4.70, and 5.71 (Rubino 2001). The change in photoreactivity with the ionic form has been widely observed for numerous compounds and frequently linked with a change of the quantum yield (Chowdhury et al. 2020). However, the MTX UVabsorption spectrum sharply changes between the two pH, presenting around 254 nm a minimum at pH 3 when MTX is a cation and a maximum at pH 7 when MTX is an anion (SI, Fig.  S3). As a result, the quantum yields appeared to be similar at pH 7 and 3, around 1.63 10 −3 and 1.83 10 −3 , respectively, because of the differences in molar absorptivity at 254 nm (Fig. 2b).
The MTX photodegradation was also performed at both pH 7 and 3 under 300-450 nm wavelength range. At pH 3, the MTX degradation under 300-450 nm irradiation was slower than at 254 nm for the same pH 3, with pseudo-kinetic constant (2.7 ± 0.9 10 −5 s −1 , n = 4) (Fig. 2a) determined with regression coefficients in the 0.94-0.98 (r 2 ) range, and quantum yield (1.7 ± 0.7 10 −4 ) significantly different (Fig. 2b). At pH 7, the overall degradation kinetics appeared as comparable to the 254 nm ones, with a pseudo-first-order kinetic constant calculated as 12.8 ± 0.9 10 −5 s −1 (n = 5) and a quantum yield not significantly different from 254 nm pH 7 (1.05 ± 0.1 10 −3 ) ( Fig. 2a and b). However, the regression coefficients were lower than in the other conditions (0.86 to 0.92) (r 2 ) because of an important initial lag phase more pronounced than at pH 3 (Fig. 1b). Chatterji and Gallelli (1977) observed the same lag phase and explained it by the accumulation of free radical produced by bicarbonates introduced in the solution, but in this study, such important free radical production could not be expected in Milli-Q water. At high initial concentration of MTX and with an irradiation wavelength comparable to ours, Hsu et al. (2019) also observed the initial lag phase. They demonstrated that it was due to a self-sensitized photodegradation mechanism: MTX enhanced its own direct photolysis, thus increasing the degradation rate through the production of the triplet-excited state of methotrexate ( 3 MTX*) and the triplet-excited state of pteridine derivatives, whereas the produced singlet oxygen quenched degradation. This self-sensitized degradation mechanism probably explains why we observed a higher degradation rate (and thus quantum yield) at pH 7 than at pH 3 and two pteridine derivatives issuing for cleavage were identified, TP206 and TP208 (see below). Even if their structure is not strictly identical to the compound evidenced by Hsu et al. (2019), these TPs could also be hypothesized to self-sensitize the reaction. This assumption is furthermore supported by the protonation site located on the diamino-pteridine and corresponding to the dissociation constant (pK a ) of 5.71 (Poe 1977;Rubino 2001).

Structural identification of photodegradation products
The elemental composition of the various compounds and their main fragment ions is summarized in Table 1. Due to the lack of standards corresponding to most of the hypothesized photoproducts, (1) a comprehensive study of the MTX fragmentation was conducted and (2) most of the identifications corresponded to a level 2b (otherwise specified) according to the Schymanski et al. (2014) classification.
In the positive mode, the MTX fragmentation pathways have been found to be consistent with the literature (Calza et al. 2014;Lutterbeck et al. 2015;Lai et al. 2017;Hsu et al. 2019;Kosjek et al. 2015;Kanjal et al. 2020). The higherenergy collisional dissociation (HCD) of [M + H] + produced the ion m/z 308 by the cleavage of the amide bond between the glutamic and p-methylaminobenzoic acid residues. From this m/z 308 ion, the cleavage of the C-N bond between pmethylaminobenzoyl and diaminopteridinemethylene groups produced the ions m/z 175 and 134. In the negative mode, unexplored to our knowledge, three fragmentation pathways were distinguished (Fig. 3 and Fig. S4). The first fragmentation may correspond to the elimination of H 2 O to form m/z 435 or to the elimination of CO 2 from the glutamic acid group, giving ion at m/z 409. This m/z 409 ion further evolved through elimination of H 2 N-CN to form m/z 367, through N-C cleavage to form m/z 235 and then through CO 2 elimination to form m/z 191. An alternative and major MTX fragmentation pathway was interpreted through a rearrangement initiated by the hydroxyl group from the carboxylic acid linked to the benzamide derivative, forming a five-centered transition state before the elimination of 3-oxoaziridin-2-yl-propanoic acid as proposed by Secrétan et al. (2018) for pemetrexed. An alternative mechanism leading to the same m/z 324 ion was also proposed in the case of pralatrexate, including an N-C cleavage followed by rearrangement (Sastry et al. 2016). The obtained ion m/z 324, which presents the same structure as [M-H] − of DAMPA, may further eliminate CO 2 (to form m/z 280) or produce the radical ions m/z 149, 174 and 189 (through m/z 280 rearrangement) by heterolytic cleavage.
In the positive mode, most of the TPs (TP324, TP424a, TP424b, TP408a) presented the fragment ions m/z 308, 175, and 134, or at least for the lower mass TPs the ion m/z 134 (TP280, TP250). Being respectively characteristics of the diamino-pteridine and N-methylaminobenzoic acid parts of the original MTX, the ions m/z 175 and 134 suggested that those parts were unchanged, and m/z 308 comforted both were unchanged. In contrast, TP440 HCD spectrum presented m/z 175 ion (thus intact diamino-pteridine group), but m/z 294 and 120 instead of m/z 308 and 134. The −14 Da shift was proved to be due to a CH 2 difference from MTX according to the elemental composition. This was thus due to the demethylation of the N-methyl bridge. Accordingly, TP266, which presents a 14 Da mass lower than TP280 and an m/z 120 ion, should be 2-(4-aminobenzamido)pentanedioic acid and results from N-demethylation of TP280 (Calza et al. 2014). The TP408b also presented the fragments m/z 294 and 120 showing its demethylation of the N-methyl bridge; its identification will be presented below. The lower mass TPs corresponding to pteridine derivatives, TP206 and TP208, were identified according to their positive mass spectra as corresponding to already known structures (Lutterbeck et al. 2015;Calza et al. 2014;Lai et al. 2017;Kosjek et al. 2015;Kanjal et al. 2020). Both spectra contained the same m/z 163 ion (C 6 H 7 N 6 + ) resulting either from loss of CO 2 for TP206 or successive losses of H 2 O (m/z 191) and CO for TP208. These two pteridine derivatives could explain the lag phase observed in the kinetic studies and explained by a self-sensitized degradation mechanism (Hsu et al. 2019). Indeed, in the first hour of MTX degradation, the abundance of these two TPs was higher under 300-450 nm pH 7 than under 254 nm pH 7.
Only ESI-positive mode is generally studied, which appeared to be more sensitive and suitable for both the parent compound and for most of the photogenerated products (Calza et al. 2014). However, the fragmentation in negative mode gives useful information about the glutamic acid part. For instance, TP280, C 13 H 16 N 2 O 5 , could eliminate consecutively two CO 2 molecules, forming m/z 235 and then m/z 191 ions. As observed for MTX, a rearrangement can lead to the formation of ion m/z 150. These two fragmentation pathways indicate that the glutamic part of TP280 is unchanged, leading to its identification as 2-(4-(methylamino)benzamido)pentanedioic acid. TP250, whose chemical formula indicates that it differs from TP280 with a loss of CH 2 O, can also be identified on the basis of its fragments in negative mode. The fragment m/z 149 indicates that the methylamino part was conserved, as this ion is 4-(methylamino)benzoylamide. The loss concerned the glutamic acid part, as proven by the ion 4-oxobut-2-enoate at m/z 99. Such ions m/z 99 and 149 were also observed in TP424a and TP424b spectra, indicating that both result from MTX oxidation, formally by loss of formaldehyde from the glutamic  acid in α and γ position. The identification of TP250, TP424a, and TP424b are of Schymanski level 3, as no indication is available to attribute α or γ position. Two isomers TP408a and TP408b were also identified at level 3. The compound TP408a resulted from MTX loss of HCOOH. Its isomer TP408b, N-demethylated according to the positive mode, presented the same m/z 363, 321, 189, and 174 ions as TP408a, but also as major ions m/z 309 and 267 (absent in the TP408a spectrum). Those ions indicate that the glutamic acid part of MTX was in TP408b transformed into amino-pentenoic acid (SI, Fig. S5 and S6). Finally, TP324, which presented the fragments m/z 149 and 174 previously mentioned, was identified (at level 2 b ) a s t h e 4 -( ( ( 2 , 4 -d i a m i n o p t e r i d i n -6yl)methyl)(methyl)amino)benzamide. Among these eleven identified TPs presented in Fig. 4, only six have been evidenced previously (Calza et al. 2014;Kosjek et al. 2015;Lutterbeck et al. 2015;or Lai et al. 2017;Kanjal et al. 2020): TP206, TP208, TP280, TP266, TP324, and TP440. In summary, three degradation pathways can be distinguished: N-demethylation of the bridge and C-N c l e a v a g e b e t w e e n p -m e t h y l a m i n o b e n z o y l a n d diaminopteridinemethylene groups, previously known, and glutamic acid oxidation evidenced in this study.

Evolution of by-products and cytotoxicity at pH 7
The IC 50 of MTX against A549 cancer cells was determined as 120 ± 20.0 nM. Consistently, at 100 nm and before photodegradation, the viability of cancer cells was close to 50% (Fig. 5). Then, for both irradiation wavelengths, the viability increased as the percentage of MTX decreased, indicating that TPs formed during irradiation at pH 7 were less cytotoxic than the parent compound. For the 300-450 nm treatment, the two pteridine derivatives, TP208 and TP206, were the main photoproducts reaching up to 31 and 92 % of residual MTX, respectively. While MTX fully disappeared after 3 h of irradiation, both TPs persisted in solution, even if TP208 decreased. Meanwhile, the viability reached 100%, showing the TPs innocuity on A549 cancer cells (Fig. 5a). At 254 nm, in addition to these same TP208 and TP206 appearing in a lower abundance than at 300-450 nm (both around 30%), TP424 also accumulated after 5 h of irradiation (with an equivalent of 18% of residual MTX). The slower degradation of MTX and the presence of TP424 could explain that complete viability was not retrieved (90.61 ± 4.18%) (Fig. 5b). TP280, TP266, and TP250 were not detected with the irradiation at pH 7, but this monitoring was done without preconcentration step (by SPE); therefore, their presence at trace level cannot be excluded. Calza et al. (2014) demonstrated that TPs formed during photodegradation in pure water were less toxic than MTX and identified TP208 as a major photoproduct. Lutterbeck et al. (2015) concluded that the intermediates formed during the reaction were less toxic than the parent compound, which included TP206 and TP208. However, these results were obtained in Vibrio Fischeri bioassays and did not focus on the anticancer activity of MTX as proposed herein.

Conclusion
Under different irradiation conditions, MTX photodegradation can produce a large variety of TPs. The kinetics showed that photodegradation was the fastest under environmental conditions (300-450 nm, pH 7) with only 7% of residual MTX after 4 h of irradiation. Nevertheless, not only the residual MTX is of environmental concern and several TPs were monitored and characterized. LC-HRMS allowed identifying eleven photoproducts, including five compounds, which were not mentioned in previous studies. The identification of these uncommon TPs was allowed by the use of mass spectrometry in the negative mode and evidenced the important oxidation degradation pathway of the glutamic acid part. The decrease in cytotoxicity during MTX degradation demonstrated that TPs formed in pure water at pH 7 are less cytotoxic than the parent compound. Nevertheless, the well-known TPs of MTX, DAMPA and 7-OH-MTX, were not observed in our irradiation conditions. Even though our experiments attempt to mimic the photodegradation that occurs in natural water or wastewater treatment, one needs to be aware that the formation of TPs highly depends on the composition of water, as well as the concentration of the parent compound. It is therefore well possible that MTX phototransformation products identified in our study appeared only as transient species in other aqueous matrices than pure water. Therefore, further investigations are necessary to estimate the occurrence of the identified TPs and their effects in natural and wastewaters. This implies to further develop analytical methods to identify and quantify these TPs in natural matrices.