Introduction

Cytarabine (CY) is an antimetabolite, used as an anticancer drug for lymphoma and acute leukemia. It is also used as an antiviral agent against herpes simplex virus and human cytomegalovirus (Liu et al. 2014). Similar to all drugs, CY is excreted through feces and urine, after administrated to patients, as mixture of unchanged parent compound and its metabolites. This mixture enters the aquatic environment through hospital and municipal wastewaters (Isidori et al. 2016a). Its urinary excretion rate is ~ 10% and its half-life degradation in the human body is 1–3 h (Zhang et al. 2013). As a result, CY has been found in concentrations up to 9.2, 14, and 13 ng L−1 in wastewater (influents and effluents) and surface water, respectively (Kosjek and Heath 2011, Martín et al. 2011). The presence of anti-cancer drugs, like CY, in effluents indicates limited removal in wastewater treatment plants (Isidori et al. 2016b). Additionally, potential toxic effects in organisms exposed to continuous or intermittent presence of parent anticancer drugs and their transformation products (TPs) cannot be ignored (Toolaram et al. 2014). Under this light, the development of effective alternative processes for wastewater treatment is necessary, in order to prevent the entrance of these compounds to the aquatic environment.

A promising approach for the removal of pharmaceuticals, such as anti-cancer drugs, that gained recognition is advanced oxidation processes (AOPs) (Koltsakidou et al. 2017c, Lin and Lin 2014, Lutterbeck et al. 2015a, b). The degradation of recalcitrant pollutants by AOPs is resulting from the formation of highly reactive hydroxyl radicals (HO·). HO· can lead to total mineralization of a variety of organic compounds (Ribeiro et al. 2015). CY degradation has been studied by several AOPs so far, including TiO2 photocatalysis, ozonation, UV/H2O2, UV/K2S2O8, and other radical promoter systems, gamma radiation, and TiO2/activated carbon under UV radiation (Koltsakidou et al. 2017a, Ocampo-Pérez et al. 2010, 2011, 2016). Although some of these treatments are very effective in CY degradation, they suffer from some major disadvantages. For example, ozonation may result in the formation of the potentially toxic oxidation by-product bromate in bromide-containing wastewaters (Soltermann et al. 2017). Additionally, ozone has an extremely short half-life for only 20 min and low solubility in water. As far as TiO2 heterogeneous photocatalysis is concerned, post-treatment recovery of the catalyst particles after water treatment is required while inefficient exploitation of visible light hinders its use (Dong et al. 2015). On the other hand, UV irradiation has limited treatment times, suffers from light-screening water constituents while leading to energy depletion and high operational costs for complete destruction (Katheresan et al. 2018). Furthermore, recent investigations are concentrated on methods that can exploit more efficiently solar light for reducing considerably operating costs. The photo-Fenton process has gained increasing attention due to its simplicity and the possibility of using sunlight for its application.

Since the generation of intermediates is inevitable during AOPs application, the determination of their chemical structure and toxicity is crucial. In silico methods (methods based on computer simulation) for the prediction of toxicity have been increasingly used by researchers during the last decade (Raies and Bajic 2016). ECOSAR is a computer program, developed by EPA, that estimates the toxicity of chemicals to aquatic organisms, such as fish (both in fresh and saltwater), invertebrates, and algae, without test data. ECOSAR’s calculations are based on structure–activity relationships (SARs) for the prediction of chemicals’ aquatic toxicity, depending on the similarity of their molecular structures to other compounds for which the aquatic toxicity is known.

So far, no efforts have been done to study the photocatalytic degradation of CY compound using photo-Fenton and photo-Fenton-like processes under simulated solar light (SSL), as well as to predict the ecotoxicity of CY and its TPs by applying in silico prediction tools using QSAR models, to the best of our knowledge. In addition, CY intermediates described in the literature have been determined using LC-MS techniques (Koltsakidou et al. 2017a), which do not allow MSn fragmentation and therefore their identification is still not fully explored.

Considering the above-mentioned issues and the scarcity of studies evaluating the degradation of CY by photo Fenton processes, the main objectives of this study were as follows: (1) to obtain new information on the treatment and transformation of CY by using photo-Fenton and photo-Fenton-like approaches under simulated solar irradiation, (2) to estimate the removal and mineralization effectiveness and suitability of different solar treatments, (3) to provide comprehensive insight concerning the formation and elucidation of TPs by using Orbitrap-LC/MS/MS, and (4) to apply in silico methods at various toxicological endpoints for the assessment of toxicity of TPs.

Experimental

Reagents and materials

The studied anti-cancer drug CY (> 98%) was purchased from TCI (Tokyo Chemical Industry Co.). The HPLC–MS grade organic solvents (methanol and acetonitrile) used for chromatographic analysis were supplied by Merck (Darmstadt, Germany). The salts used for ion chromatography analyses (Na2CO3 and NaHCO3) were purchased from Sigma–Aldrich (Germany). Hydrogen peroxide (H2O2, 30% w/v) was obtained from Panreac (Barcelona, Spain). All the salts used for the oxidation techniques (FeCl3·6H2O, C2K2O4·1H2O, Na2S2O8 > 99 + %) were supplied from Chem-Lab NV (Zedelgem, Belgium). The water used for chromatographic analysis and photocatalytic experiments was ultrapure water (Millipore Waters Milli-Q water purification system).

Photodegradation experiments

The equipment used in the study as well as the experimental conditions for the photo-Fenton experiments has been reported in detail previously (Koltsakidou et al. 2017b). Briefly, the simulated solar irradiation was conducted by Atlas Suntest CPS+ (Germany), which was equipped with a xenon lamp (< 290 nm) with a constant irradiance of 500 Wm−2. CY solutions (10 mg L−1) were acidified (pH 2.9–3), before each experiment, using HCl, in order to prevent iron precipitation. Measurements of pH at various irradiation times and at the end of the treatment show no significant variation from the starting values. For the determination of mineralization’s evolution, as well as the identification of TPs, 30 mg L−1 of CY was used as initial concentration, in order to obtain slower kinetics and facilitate their detection. Ferrioxalate was also prepared as reported in previous studies (Doumic et al. 2015, Monteagudo et al. 2010).

Analytical procedures

CY determination

CY concentration, during the photodegradation experiments, was determined by an HPLC system with SPD 20A DAD detector coupled in series with the LC–MS 2010EV mass selective detector obtained from Shimadzu (Kyoto, Japan). The exact conditions of CY determination have been reported in a previous work (Koltsakidou et al. 2017a).

Mineralization studies

A Shimadzu TOC V-csh Analyzer was used for the determination of total organic carbon (TOC) concentration. NO3− ions were quantified by an ion chromatography (Metrohm) as reported in a previous study (Koltsakidou et al. 2017a).

By-product evaluation

The intermediates formed during CY photodegradation were characterized by an LTQ Orbitrap Discovery MS high resolution system that operated in positive ionization mode. The chromatographic analysis was conducted using an Acquity UPLC HSS T3 column (100 × 2.1 mm, 1.8 μm, Waters, Milford, MA, USA), at a constant temperature of 40 °C and a flow rate of 500 L min−1. The mobile phases used were water/formic acid (0.1% v/v) (phase A) and methanol/formic acid (0.1% v/v) (phase B). The chromatographic separation was obtained by the application of a linear gradient (90% A to 0% A in 15 min, 90% A in 20 min). ESI source parameters were described in a previous work (Antonopoulou et al. 2013).

Ecotoxicity assessment

Predictions of ecotoxicity (acute and chronic) of CY and its TPs on fish, daphnia, and green algae were conducted by ECOlogical Structure-Activity Relationship Model (ECOSAR), a program (version 1.11) developed by USEPA (https://www.epa.gov/tsca-screening-tools/ecological-structure-activity-relationships-ecosar-predictive-model). The detailed description of ECOSAR predictions has been already reported elsewhere (Ferri et al. 2017).

Results and discussion

Optimization of photo-Fenton treatment

Preliminary Fe3+/H2O2 and Fe3+/H2O2/SSL degradation experiments

Experiments in the dark and under SSL by employing the Fe3+/H2O2 system for the degradation of CY were conducted and compared as a preliminary study. Figure 1 summarizes the obtained results. Under no irradiation, only 10% reduction is achieved during the first 20 min. During the same period of time, the reduction achieved by Fe3+/H2O2/SSL treatment is approximately 99%. The effect of simulated solar light irradiation alone is also depicted in Fig. 1. Obviously, photolysis provokes no significant degradation (20% after 60 min of irradiation) of CY, compared to Fe3+/H2O2/SSL treatment. The above preliminary results indicate that photolysis as well as dark reaction (Fe3+/H2O2) has minus contribution to the degradation of CY, when Fe3+/H2O2/SSL treatment is applied.

Fig. 1
figure 1

Photo-Fenton treatment of CY in comparison to Fenton treatment and direct photolysis (experimental conditions: Co(CY) = 10 mg L−1, [Fe3+] = 1 mg L−1, [H2O2] = 30 mg L−1)

Full size image

Degradation kinetics and effect of initial CY concentration

Since the reaction between HO· and CY is the rate-determining step, as reported to previous homogeneous photocatalytic studies for pollutants (Evgenidou et al. 2007, Martínez-Costa et al. 2018), the studied photo-Fenton process can be expressed by Eq. (1):

$$ r=-\frac{\mathrm{dC}}{\mathrm{dt}}={\mathrm{k}}_{\mathrm{HO}}\left[\mathrm{H}{\mathrm{O}}^{\cdot}\right]\mathrm{C} $$
(1)

where C is the concentration of CY and kHO is the reaction rate constant.

Because HO· steady-state concentration is reached quite fast, the above equation can also be written as Eq. (2):

$$ r=-\frac{\mathrm{dC}}{\mathrm{dt}}={\mathrm{k}}_{\mathrm{HO}}\left[{\mathrm{HO}}^{\cdot}\right]\mathrm{C}={k}_{app}C $$
(2)

where kapp is the apparent pseudo-first-order constant.

In this study, the above pseudo-first-order model appeared to fit well the experimental data (R2 > 0.995 for ln(C0/C) vs. time (t)), and the apparent kapp values were obtained by linear fitting from the degradation curves. For minimization of variations resulting from competitive effects of intermediates, only the data exported from the first minutes of irradiation were taken into consideration for the calculation of kapp.

For the evaluation of CY initial concentration influence to the reaction rate (kapp), experiments were carried out at various concentrations of CY, varying from 10 to 50 mg L−1, while the concentration of Fe3+ and H2O2 were kept constant at 1 mg L−1 and 60 mg L−1, respectively. As indicated in Fig. 2a, an increase in the initial concentration of CY till 50 mg L−1 leads to a decrease of the apparent rate constant, which later reaches a plateau. The effectiveness of photo-Fenton process depends highly on the formation of hydroxyl radicals and their ratio to the initial concentration of the substrate. Consequently, since the concentration of the target compound is increased and the initial concentration of the hydroxyl radicals remains the same, the degradation of the target compound will need longer irradiation times under identical experimental conditions (Michael et al. 2010). Moreover, higher substrate concentration may absorb a fraction of light decreasing the activation of the catalyst thus decreasing the formation of hydroxyl radicals (Lambropoulou et al. 2017, Tamimi et al. 2008).

Fig. 2
figure 2

The effect of operational parameters on CY photo-Fenton degradation; effect of CY loading ([Fe3+] = 1 mg L−1, [H2O2] = 60 mg L−1) (a); Effect of Fe3+ loading (Co(CY) = 10 mg L−1, [H2O2] = 60 mg L−1) (b); Effect of H2O2 loading (Co(CY) = 10 mg L−1, [Fe3+] = 1 mg L−1) (c); Effect of S2O82− loading (Co(CY) = 10 mg L−1, [Fe3+] = 1 mg L−1) (d)

Full size image

Effect of iron concentration

Since iron residues in effluents and sludge are considered harmful to the environment, low concentrations (lower than 5 mg L−1) that result in short reaction times, avoiding the need of further treatment for iron removal, are preferable (Gar Alalm et al. 2015). In this study, the effect of iron concentration was evaluated by the addition of 0–2.5 mg L−1 Fe3+, 60 mg L−1 H2O2, and 10 mg L−1 initial concentration of CY. Figure 2b represents the evolution of degradation rate within the studied concentration levels of Fe3+. The addition of Fe3+ increases the reaction rate in a linear trend which is attributed to the proportional amount of HO· formation, as indicated by photo-Fenton reactions (Eq. 3):

$$ {\mathrm{Fe}}^{3+}+{\mathrm{H}}_2\mathrm{O}\overset{hv}{\to }{\mathrm{H}\mathrm{O}}^{\cdotp }+{\mathrm{Fe}}^{2+}+{\mathrm{H}}^{+} $$
(3)

Moreover, according to Eq. (3), ferric ions are photoreduced to ferrous ions which are the main species that catalyze hydrogen peroxide (Fenton reactions (Eqs. 4–6)) thus producing additional radicals such as HO· and HO2·.

$$ {\mathrm{Fe}}^{2+}+{\mathrm{H}}_2{\mathrm{O}}_2\to {\mathrm{Fe}}^{3+}+{\mathrm{O}\mathrm{H}}^{-}+{\mathrm{H}\mathrm{O}}^{\cdotp } $$
(4)
$$ {\mathrm{Fe}}^{3+}+{\mathrm{H}}_2{\mathrm{O}}_2\to \mathrm{Fe}\hbox{--} {\mathrm{O}\mathrm{OH}}^{2+}+{\mathrm{H}}^{+} $$
(5)
$$ \mathrm{Fe}\hbox{--} {\mathrm{OOH}}^{2+}\to {{\mathrm{HO}}_2}^{\cdotp }+{\mathrm{Fe}}^{2+} $$
(6)

Additionally, when no Fe3+ are added, a low degradation of CY is still observed due to the photolysis of H2O2 under UV radiation (Irmak et al. 2004), as descripted by Eq. (7):

$$ {\mathrm{H}}_2{\mathrm{O}}_2\overset{hv}{\to }2{\mathrm{H}\mathrm{O}}^{\cdot}\kern0.50em \left(\lambda <320\ \mathrm{nm}\right) $$
(7)

However, the overlap of the adsorption spectrum of hydrogen peroxide with the emission spectrum of xenon lamp of the solar simulator is very small; hence, a minor degradation rate was observed in the absence of iron.

All in all, for further experiments, the concentration of 1 mg L−1 was selected, since higher concentrations of Fe3+ would result in much faster degradation of CY, that disable the proper monitoring of the degradation kinetics.

Effect of H2O2 concentration

During photo-Fenton process, hydrogen peroxide’s initial concentration is crucial for achieving effective oxidation of the organic substrate. The efficiency of the applied procedure may be decreased both by low and excess of H2O2 (Bautitz and Nogueira 2010). In this study, concentrations varying from 0 to 120 mg L−1 of H2O2 were examined. As indicated in Fig. 2c, the kapp increases till the concentration of 30 mg L−1, reaching a plateau or even decrease a little thereafter. This decrease can be explained by the enhancement of parallel reactions, such as scavenging of generated HO· by H2O2 as described below (Eqs. (8–10)) (Abdessalem et al. 2008, Dal Bosco et al. 2011).

$$ {\mathrm{H}}_2{\mathrm{O}}_2+{\mathrm{H}\mathrm{O}}^{\cdotp}\to {{\mathrm{H}\mathrm{O}}_2}^{\cdotp }+{\mathrm{H}}_2\mathrm{O} $$
(8)
$$ {2\mathrm{H}}_2{\mathrm{O}}_2\to {2\mathrm{H}}_2\mathrm{O}+{\mathrm{O}}_2 $$
(9)
$$ {2\mathrm{HO}}^{\cdotp}\to {\mathrm{H}}_2\mathrm{O}+1/{2\mathrm{O}}_2 $$
(10)

Therefore, the concentration of 30 mg L−1 has been chosen for further experiments.

In addition, when no peroxide was added to the treated solution, photodegradation of the target compounds still occurs. This may be attributed to the excitation of the [FeIIIOH(H2O)5]2+ complex which is the dominant specie in the aquatic solution of the target compound at pH ~ 3. This aqua-complex can produce also hydroxyl radicals under irradiation following Eq. (11):

$$ {\left[{\mathrm{Fe}}^{\mathrm{II}\mathrm{I}}\mathrm{O}\mathrm{H}{\left({\mathrm{H}}_2\mathrm{O}\right)}_5\right]}^{2+}+{\mathrm{H}}_2\mathrm{O}+\mathrm{hv}\to {\left[{\mathrm{Fe}}^{\mathrm{II}}{\left({\mathrm{H}}_2\mathrm{O}\right)}_6\right]}^{2+}+{\mathrm{H}\mathrm{O}}^{\cdotp } $$
(11)

Moreover, photodegradation of CY in the absence of peroxide can also occur through the formation of Fe-CY complex since cytarabine disposes functional groups that can be acted as ligands to Fe(III). As reported elsewhere, Fe(III) complexes may exhibit higher absorbance and higher quantum yields than simple aquated Fe(III) species, and therefore may be more important in photo-Fenton systems (Pignatello et al. 2006). Méndez-Arriaga et al. (2010) (Méndez-Arriaga et al. 2010) showed that Fe(III) bounds with the organic target compound (ibuprofen) thus forming a complex that is photoactive. This complex promotes in a strong manner the decarboxylation of the organic substrate.

Comparison of photo-Fenton with persulfate and Fenton-like treatments of CY

The efficiency of Fe3+/H2O2/SSL photocatalytic system was compared with Fe3+/S2O82−/SSL and [Fe(C2O4)3]3−/H2O2/SSL systems, regarding the reaction rate and the mineralization percentage achieved in each case.

Sulfate radicals (SO4·−, E0 = 2.60 V), which can be produced by activation of peroxodisulphate, are an alternative oxidant that has been lately studied as degradation agent for several organic pollutants (Ding et al. 2017, Ismail et al. 2017, Soltani and Lee 2017). Also, sulfate radicals present a longer life-time than hydroxyl radicals (Gao et al. 2016) and are more selective towards organic substrates reacting via direct electron transfer mechanism (Kamińska et al. 2018, Matta et al. 2011). In this study, SO4·− is produced by the reaction of peroxodisulphate and Fe(III) as described in Eqs. (12) and (13) (Avetta et al. 2015, Hu et al. 2016):

$$ \mathrm{Fe}\left(\mathrm{II}\mathrm{I}\right)+{\mathrm{S}}_2{{\mathrm{O}}_8}^{2-}\to \mathrm{Fe}\left(\mathrm{II}\right)+{\mathrm{S}}_2{{\mathrm{O}}_8}^{\cdotp -} $$
(12)
$$ \mathrm{Fe}\left(\mathrm{II}\right)+{\mathrm{S}}_2{{\mathrm{O}}_8}^{2-}\to \mathrm{Fe}\left(\mathrm{II}\mathrm{I}\right)+{{\mathrm{S}\mathrm{O}}_4}^{2-}+{{\mathrm{S}\mathrm{O}}_4}^{\cdotp -} $$
(13)

The studied concentrations of added peroxodisulphate in this study were 0–120 mg L−1, while the concentration of Fe (III) was 1 mg L−1 for comparison purposes with the Fe3+/H2O2 process. As indicated in Fig. 2d, the degradation rate of CY is almost fivefold increased by the addition of peroxodisulphate up to 30 mg L−1, while a near plateau is observed for higher concentration levels. As a result, the concentration level of 30 mg L−1 is considered the most suitable under the current experimental conditions.

Ferrioxalate system has been also considered as an efficient alternative to Fe3+/H2O2 system for the degradation of pollutants, since it can expand the usage of solar spectrum range up to 450 nm, improving the solar-Fenton process (Li et al. 2015). Furthermore, the photolysis of ferrioxalate complex leads to additional oxidants according to the following equations (Lucas and Peres 2007):

$$ {\left[\mathrm{Fe}{\left({\mathrm{C}}_2{\mathrm{O}}_4\right)}_3\right]}^{3-}\overset{hv}{\to}\mathrm{Fe}\left(\mathrm{II}\right)+2{\mathrm{C}}_2{{\mathrm{O}}_4}^{2-}+{\mathrm{C}}_2{{\mathrm{O}}_4}^{\cdot -} $$
(14)
$$ {\mathrm{C}}_2{{\mathrm{O}}_4}^{\cdotp -}+{\left[\mathrm{Fe}{\left({\mathrm{C}}_2{\mathrm{O}}_4\right)}_3\right]}^{3-}\to \mathrm{Fe}\left(\mathrm{II}\right)+{3\mathrm{C}}_2{{\mathrm{O}}_4}^{2-}+{2\mathrm{CO}}_2 $$
(15)
$$ {\mathrm{C}}_2{{\mathrm{O}}_4}^{\cdotp -}+{\mathrm{O}}_2\to {{\mathrm{O}}_2}^{\cdotp -}+{2\mathrm{CO}}_2 $$
(16)

In order to compare the three homogeneous photocatalytic treatments of CY, by means of mineralization as well as to identify and compare the TPs generated by each procedure, three times higher initial concentration of CY was selected (30 mg L−1), which enables simpler and less laborious procedures for the monitoring of their evolution. The concentration of ferric ions and the concentration of the oxidant were also increased three times (3 mg L−1 and 90 mg L−1 in each case), so that the reaction kinetics are close to those obtained by the previous experiments. Under the selected experimental conditions, the reaction rates were 0.353, 0.162, and 0.379 min−1 for Fe3+/H2O2/SSL, Fe3+/S2O82−/SSL, and [Fe(C2O4)3]3−/H2O2/SSL respectively. The addition of oxalate enhances TOC reduction of CY solution, as indicated in Fig. 3. The TOC reduction achieved by Fe3+/H2O2/SSL and Fe3+/S2O82−/SSL treatments is approximately 73%, while TOC reduction reaches 82% by [Fe(C2O4)3]3−/H2O2/SSL treatment, after 360 min of SSL irradiation. On the other hand, no significant difference is observed at all three treatments, as far as NO3− release is concerned, which was quite low (3–4% in all cases), indicating the formation of TPs that preserves N atoms. Consequently, it appears that [Fe(C2O4)3]3−/H2O2/SSL treatment is the most efficient among the studied processes. This result is in agreement with previous studies since it has been proven that counter ion of iron can play important role in the photodegradation kinetics (Evgenidou et al. 2007). More specifically, anions in general, that may originate either from background electrolytes or Fe3+ counter ion or even H+ counter ion, can affect the photocatalysed degradation of the target compound. Oxalates appear to give higher reaction rates since chloride ions may create complexes with iron thus inhibiting the activation of the oxidant or they can act as scavengers of hydroxyl radicals (Kiwi et al. 2000, Pignatello 1992). Moreover, Bautitz and Nogueira (2010) (Bautitz and Nogueira 2010) attributed the higher efficiencies of the ferrioxalate to the higher quantum yield of Fe2+ production. Comparing the Fe3+/H2O2/SSL system with Fe3+/S2O82−/SSL, no great differences appear. However, the lower reaction rates achieved in the latter may be attributed to the slightly lower oxidation potential of sulfate radical and the different mechanisms taking place during the degradation (Wang et al. 2015). Moreover, previous studies proved that in transition metal–based activation reactions like Eqs. (4) and (13), iron is a better activator for peroxide than for persulfate (Anipsitakis and Dionysiou 2004).

Fig. 3
figure 3

Mineralization process of CY, SSL/Fe3+/H2O2 (a); SSL/Fe3+/S2O82− (b); SSL/[Fe(C2O4)3]3−/H2O2, (Co(CY) = 30 mg L−1, [Fe3+] = 3 mg L−1, [oxidant] = 90 mg L−1) (c)

Full size image

Evaluation of TPs formed during photo-Fenton and photo-Fenton-like processes

The TPs generated during the photo-Fenton, photo-Fenton-like, and persulfate processes were identified through the analysis of mass spectra m/z ions, based on accurate mass measurements, as given by Table 1. Totally, twelve TPs were identified, four of which (TP 273, TP 277, TP 274 (I), and TP 274 (II)) were identified for the first time, to the best of our knowledge. Their proposed structures are depicted in Fig. 4. All of them were detected in low concentrations because of the very fast degradation kinetics that took place at all studied processes, and therefore their evolution profiles could not be obtained. The given generated TPs were identified during all the three studied processes. The MS2 product ion spectra of major TPs of CY are given by Fig. S1. As reported also in a previous study of ours (Koltsakidou et al. 2017a), HO· attack on CY molecule, resulting in the formation of hydroxylated TPs, TP 259 (I), and TP 259 (II). The presence of the fragment ion m/z 112.0505, that correspond to the pyrimidine ring of CY molecule, suggests that hydroxylation occurs at tetrahydrofuran ring (Fig. S1). Consecutive hydroxylation of CY molecule generated di- and tri-hydroxylated TPs (TP 275 and TP 291 (I, II)). The fragment ion at m/z 112.0505 was also observed for TP 275 fragmentation pattern, indicating again no hydroxylation at the pyrimidine ring of CY molecule (Fig. S1). The fragmentation of TP 291 indicates an initial loss of –H2O (m/z 274.0670) group, usual for hydroxylated TPs, and further loss of –CHNO group (m/z 231.0612), which corresponds to the cleavage of the pyrimidine ring (Fig. S1). TP 273 probably results from further oxidation of TP 275, which forms a keto-group. Furthermore, although that uracil arabinoside, formed from the breaking of C–NH2 bond of CY molecule and subsequent oxidation of pyrimidine ring (Koltsakidou et al. 2017a), has not been detected in this study, hydroxylated uracil arabinoside TPs were detected (TP 276 I,II) during the applied processes. Their fragmentation data suggest loss of –H2O (m/z 259.0561). Further oxidation of TP 276 I, II, lead to the formation of keto-products, TP 274 I, II, where loss of –H2O (m/z 257.0404) was also observed during fragmentation. Additionally, one TP formed from the cleavage of tetrahydrofuran ring after subsequent oxidation-hydroxylation of CY molecule was identified (TP 277) and its fragmentation data (fragment ion m/z 112.0505) suggest that hydroxylation happened only at the tetrahydrofuran ring of CY molecule. Finally, TP 111 (cytosine) that corresponds to the pyrimidine ring of CY molecule was also detected as a CY breakdown product. As it has already been reported, TP 111 can lead to hydroxyl and de-aminated products (as also suggested by its fragmentation data, m/z 95.0240 corresponds to –NH3 loss). These hydroxyl and de-aminated products may be finally transformed in tartronic, oxalic, mesoxalic, and lactic acid (Elsellami et al. 2014, Koltsakidou et al. 2017a).

Table 1 High resolution mass spectra data for CY and identified TPs for the studied AOPs processes
Full size table
Fig. 4
figure 4

Transformation of CY under SSL/Fe3+/H2O2 (a), SSL/Fe3+/S2O82− (b), and SSL/[Fe(C2O4)3]3−/H2O2 (c) processes; transformation pathway 1 (1), transformation pathway 2 (2) and transformation pathway 3 (3)

Full size image

Based on the structures found in this study and previously reported results in literature (Elsellami et al. 2014, Koltsakidou et al. 2017a), Fig. 4 summarizes the proposed transformation pathways by HO· radical attack. Three different routes have been considered including hydroxylation that occurred preferably in the tetrahydrofuran ring of CY molecule with the subsequent oxidation (pathway 1), HO· radical attack to aromatic ring and the substitution of NH2 by HO group (pathway 2), and the cleavage of C–N bond (pathway 3).

Toxicity evaluation

Acute (LC50 and EC50) and chronic (ChV) toxicity of CY and its TPs were predicted using ECOSAR software. These kinds of predictions can be extremely helpful for water treatments, since the formation of potentially toxic TPs may be identified, enabling possible control (Tay and Madehi 2015). Acute and ChV toxicity levels were determined according to the Globally Harmonized System of Classification and Labelling of Chemicals, which is given by Table S1 (GHS 2011). The predicted ecotoxicity values of CY and its TPs on daphnia, fish, and green algae are categorized by the above classification and presented at Table 2. Each compound, depending on its chemical structure, may be classified by more than one toxicity class by ECOSAR program. Also, the predictions were contacted for more than one structure for some TPs (given at Electronic Supplementary Information (ESI)), since the exact hydroxylation and/or oxidation position cannot be determined accurately. The given structures are selected as the most representative for each compound, since ECOSAR calculations for possible other structures that are not given presented no difference, compared to those selected (Figs. S2–8).

Table 2 Predicted acute and chronic toxicity of CY and its TPs
Full size table

According to the results, CY itself appears to present neither acute nor ChV toxicity on daphnia, fish, and green algae. As far as the CY-TPs is concerned, TP 274 and TP 276 may pose a threat, since they appear to have very toxic acute and chronic effect on green algae. Additionally, TP 274, TP 291, TP 276, TP 273 may present acute toxic effect and harmful or toxic chronic effect on daphnia. Furthermore, TP 274, TP 291, TP 276, and TP 273 are possibly a threat to fish, since according to the results, they present harmful both acute and chronic effect on them. Conclusively, the TPs generated during the studied AOP treatments appear to present aquatic toxicity, in contrast to the parent compound, that presents no toxicity at all. However, TP 273 (TP 273D) may be completely harmless, if the hydroxylation is conducted at the tetrahydrofuran ring and the pyrimidine amino group (Fig. S4). Also, TP 277 presents neither acute nor chronic toxicity. Additionally, there were no significant differences between the toxicity values obtained for TPs (TP 291, TP 273, TP 276), when the hydroxylation was conducted at different positions of pyrimidine or tetrahydrofuran ring of CY molecule (that belong at the same chemical class).

In our previous study, Vibrio fischeri bioassay was conducted, for CY and its degradation TPs by AOP treatment (Koltsakidou et al. 2017a). The initial CY solution presented very low toxicity (inhibition ~ 16%), while the TPs generated during the first minutes of the applied treatment presented increased toxicity (45% inhibition at 30 min of irradiation), which decreased later (5% inhibition after 180 min). All in all, the results obtained from V. fischeri bioassay are in accordance with the predictions conducted in silico in the present study, for CY and its TPs, since for most of the identified TPs, the predicted toxicity is higher than CY itself.

Application of the studied treatments for CY removal in wastewater matrix

The majority of studies related to the photocatalytic degradation of various organic pollutants are carried out in solutions of deionized or ultrapure water or simulated waste. In the present study, experiments of homogeneous photocatalytic degradation of CY were carried out in wastewater obtained from an effluent of a conventional wastewater treatment plant (WWTP). The wastewater effluent composition and physical characteristics are given at ESI (Table S2).

The experiments were carried out under the optimal conditions (Fig. S9). Since the resulting correlation coefficients were R2 > 0.95, the first-order simplified model can be applied also to adequately describe the degradation of CY at complex wastewater matrix. The CY degradation rate constant, as well as the half-life (Table S3), is reduced in the sample from WWTP, compared to that obtained for ultrapure water, for all three studied treatments. This reduction can be attributed to the large amount of organic matter dissolved in the sample from WWTP (~ 12 mg L−1), which consumes the oxidant species generated during the applied photocatalytic treatment (Jallouli et al. 2018). Also, the inorganic components, such as bicarbonates, chlorides, and sulfates, which are present in wastewaters act like scavengers of hydroxyl radicals, affecting the efficiency of the applied photocatalytic treatment (Shao et al. 2018).

Moreover, a higher consumption of the oxidant during the treatment of an effluent is expected since higher amount of TOC is treated and because chloride ions react also with peroxide through the reactions:

$$ {\mathrm{Cl}}^{\cdotp }+{\mathrm{H}}_2{\mathrm{O}}_2\to {{\mathrm{H}\mathrm{O}}_2}^{\cdotp }+{\mathrm{Cl}}^{-}+{\mathrm{H}}^{+} $$
(17)
$$ {{\mathrm{Cl}}_2}^{\cdot -}+{\mathrm{H}}_2{\mathrm{O}}_2\to {{\mathrm{H}\mathrm{O}}_2}^{\cdot }+2{\mathrm{Cl}}^{-}+\mathrm{H} $$
(18)

However, all process can be considered effective since more than 60% degradation is achieved within 60 min of irradiation time while with the ferrioxalate system, 85% degradation is observed. Prolonged irradiation times and higher initial doses of iron and hydrogen peroxide that are commonly used in the treatment of effluents of WWTPs would give a higher removal outcome (Bautitz and Nogueira 2010, Gómez et al. 2008).

Conclusions

The investigation of CY degradation by photo-Fenton (Fe3+/H2O2/SSL) and photo-Fenton-like treatments (Fe3+/S2O82−/SSL and [Fe(C2O4)3]3−/H2O2/SSL) was conducted in this study. Under optimized conditions, CY’s complete removal was achieved in less than 45 min, by all treatments applied. The conducted mineralization studies indicated that [Fe(C2O4)3]3−/H2O2/SSL treatment was the most efficient, since TOC reduction reached 82% after 360 min of SSL irradiation, compared to 73% TOC reduction achieved by Fe3+/H2O2/SSL and Fe3+/S2O82−/SSL treatments, at the same irradiation time. In addition, twelve TPs were identified during the applied processes, four of which were identified for the first time. Their identification was conducted by liquid chromatography coupled to high resolution mass spectrometry. According to suggested structures of the generated TPs, CY’s transformation during the applied treatments occurred mainly by hydroxylation and subsequent oxidation. Last but not least, the predicted values of acute and chronic toxicity for CY and its TPs on fish, daphnia, and green algae suggested that the TPs generated during the studied AOP treatments may pose a threat to aquatic environment, in contrast to the parent compound that presents negligible toxicity. Finally, all the applied techniques presented lower degradation efficiencies when used in wastewater matrix, compared to those obtained for ultrapure water due to the presence of organic matter and inorganic ions acting as radical scavengers.