In neutral solutions, fluoroquinolones occur as zwitterions, as can be recognized from their pKa values. This property exercises influence on the photoinduced degradation through directing the chemical transformation and through the UV absorption efficiency, which depends on the pH as well, see
Figure 2a. The spectral range of the irradiation also affects the quantum efficiencies. Hence, UVC or UVA lamps used in this study may lead to different transformation pathways and products. In order to predict conditions for efficient degradation of LOM, absorption spectra were recorded at different pH values and compared with the emission spectra of the lamps, see
Figure 2b.
3.1. Occurrence of Lomefloxacin in Surface Water
The presence of LOM is a worldwide problem, and thus, also in Germany. A sample from the River Erft showed the occurrence of LOM, whereby no LOM was observed in the sample from the River Rhine in summer 2018. The findings were confirmed through comparison of chromatographic retention times and the MS/MS fragmentation pattern with a LOM reference sample, see
Figure 3. The accurate mass observed agreed also perfectly with the theoretical monoisotopic mass, cf.
Table 1. Further substances were found in the river, such as the b
1 receptor blocker metoprolol. The remaining substances found in both river water samples were not considered in this study.
Samples containing LOM were taken from the River Erft immediately after a sewage treatment plant, while in samples collected upstream of the plant showed no presence of LOM. In agreement with previous studies, WWTPs seem to concentrate LOM and act as entry path into the aquatic environment [
43,
44]. Findings as this underline the need for a fourth purification stage based on AOPs or at least advanced filtering systems. In this respect, the photo-induced degradation of LOM is investigated.
3.2. Photoinduced Degradation of Lomefloxacin—Kinetic of Degradation
The influence of UVA and UVC radiation, pH values, and model and surface water matrices on the photoinduced degradation were analyzed. Examples are given in
Figure 4. Photoinduced transformation and degradation were monitored using HPLC-MS techniques and plotted as concentration or area-under-the-curve mass signals versus time (
c-t).
The
c-t curves obtained could be best described by first or pseudo-first order kinetics in agreement with previous studies [
25,
26,
36]. The results are presented in
Table 2 as kinetic rate constants and quantum efficiencies.
Radiation range, pH, and matrix were found to influence the degradation rate. While the degradation of LOM in acidic milieu was slowest, it could be quadrupled by increasing the pH. The water matrix also affects the degradation rate. In MilliQ water, the degradation is faster compared to effluent and river water. In environmental water matrices, a large variety of compounds are present and decrease the amount of light reaching LOM which induces its transformation or degradation. The effect of pH on the rate constants may be twofold. Firstly, LOM itself may be degraded faster in its zwitterionic or anionic state, like in neutral and alkaline milieu. Secondly, the photoinduced formation of hydroxyl radicals contributes to the degradation and is pH dependent. The presence of hydroxide hampers hydroxyl radical induced degradation, hence would decrease the part of the rate constant originating from this mechanism. Yet, hydroxide can act as a reagent itself and would lead to an increase of the rate constant. As can be seen from
Table 2, quantum efficiencies exceeding 1 suggest more than one mechanism in action. In conclusion, the obviously best conditions for the removal of LOM were UVC irradiation and a pH between neutral and slightly alkaline.
Among AOPs, photocatalysts, such as titanium dioxide in combination with UVA irradiation or hydrogen peroxide in combination with UVC irradiation, were often used to accelerate the degradation of many anthropogenic micropollutants [
7,
45,
46,
47]. Yet, no significant differences were observed between presence and absence of these photocatalysts in this study, see
Table 3. The sole exception occurred when the degradation was accelerated in the presence of hydrogen peroxide under UVC irradiation at pH 3–4. The lack of efficiency of hydrogen peroxide was traced back to saturation effects [
7,
48], while the presence of titanium dioxide was assumed to cause light scattering hence reducing the amount of light for direct degradation [
49]. Since effluents most often show pH values in the neutral or weak alkaline range, no need for the application of hydrogen peroxide or titanium dioxide was indicated.
To this point, only the induced degradation of LOM was considered. Since organic decomposition products were formed, it was also interesting to monitor TOC with irradiation time.
Figure 5 shows the irradiation time dependent TOC of LOM in river Erft water and MilliQ water at an initial concentration of 20 mg L
−1 and under UVC irradiation.
The initial TOC in river water was about 6.5 times higher than that of MilliQ water containing 20 mg L−1 of LOM, since the river water contained a variety of organic substances. After 7 h of UVC irradiation, the TOC in both samples was reduced by 10 mg L−1. As the initial TOC of the MilliQ water sample amounted to 10 mg L−1, complete mineralization was achieved. In contrast, the irradiation time of the river Erft water sample would have had to be significantly extended for complete mineralization. It is obvious that TOC is no suitable measure for the removal of LOM or other pharmaceuticals as surface waters or effluents carry innumerable organic compounds that react differently to UV exposure. Nevertheless, it is reasonable to assume that LOM degradation proceeded via transformation products that were analyzed and identified using HPLC-ESI-Q-TOF-MS and HPLC- ESI-IT-MS as described in the following section.
3.3. Photoinduced Degradation Products of Lomefloxacin
During the degradation experiments, both known and unknown compounds were identified by HPLC-ESI-IT-MS and/or HPLC-ESI-Q-TOF-MS. An overview of the observed substances is shown in
Table 4.
Some of the photodegradates shown in
Table 4 were preferably formed on UVA irradiation, others exclusively on UVC irradiation. The latter gave rise to many more transformation products. In particular, the effect of hydroxyl group substitution was recognized when using different radiation sources. Alkaline pH and UVC irradiation yielded relatively many photodegradates with hydroxyl substituents, whereas these products could not be identified on UVA irradiation, see green marks in
Figure 6. Photoinduced deconstruction of the piperazine moiety seemed independent of pH, see orange marks in
Figure 6. The same observation was made for reactions of the ethyl substituent on the quinolone nitrogen, cf. purple marks in
Figure 6. Although it cannot be ruled out that different mechanisms might lead to identical transformation products, it could be concluded that pH independent UVA and UVC photoinduced chemistry led to the fragmentation of the piperazine ring. The ethyl substituent reactions might stem from hydroxyl radical and hydroxide chemistry depending on pH, as hydroxide acts as radical scavenger for hydroxyl radicals. At last, fluorine substitution by and addition of hydroxyl groups might also be due to both mechanisms, depending on pH. An overview of the observed transformation products is given in
Figure 6. As indicated in
Table 4, seven not previously reported products were proposed based on MS
2 and MS
3 experiments.
Following kinetic analysis of the
c-t diagrams, see
Figure 4, most of the transformation products could be described as intermediate products. While only the intermediate with m/z = 332 due to UVA irradiation possessed a life-time well above 30 min, all photoproducts due to UVC irradiation were eliminated within 15 min, cf.
Figure 4. Under suitable neutral and slightly alkaline conditions, UVC irradiation led to faster elimination of LOM and its photoproducts than UVA irradiation.
3.4. Assessment of Ecotoxicity
To assess the ecotoxicity of the resulting products and intermediates of the photoinduced elimination of LOM, IC
50, and MIC values were determined. The MIC value of LOM against
B. subtilis was found 0.17 µg·mL
−1, and 0.92 µg·mL
−1 against
P. fluorescens. The IC
50 value of LOM against
B. subtilis was 0.09 µg·mL
−1 and 0.22 µg·mL
−1 against
P. fluorescens. As discussed in previous studies, IC
50 values can be determined more accurately than MIC values due to their graphical representation [
9]. Yet, an error of a factor of 2 is negligible due to the dilution assay format. A factor of 4 between IC
50 and MIC in the case of
P. fluorescens is rather large. Example curves for IC
50 determination are given in
Figure 7.
As expected, the longer the LOM solution was irradiated with UVA light, the higher the IC
50 grew and the least activity remained. The comparison between MIC and IC
50 values depending on time of irradiation by UVA and UVC is shown in
Table 5.
In general, activity decreased with irradiation times where UVC irradiation appeared more effective than UVA, as can be seen from the higher MIC values at equal irradiation times. Since both bacteria represent Gram-positive and Gram-negative species ubiquitous in the aquatic environment, the lack of activity of a compound or compound mixture has been proposed as means to assess the ecotoxicity [
11]. The advantage of the assay is its ease of use. While QSAR analyses often rely on the knowledge of the exact structures of the transformation products after AOP application, MIC or IC
50 values represent sum parameters and therefore reflect the activity in total, hence answering indirectly to an assumption that transformation products might be more ecotoxic than the initial drug itself.
To address the issue whether transformation products are less or more toxic than their parent drug, QSAR analyses represent a common approach, since the synthesis of the suggested products is usually expensive, and no corresponding standards were available in this case. QSAR analysis was carried out on the basis of the chemical structures of the known and newly identified products of LOM, see
Table 4. The results for selected species are presented in
Table 6.
Most of the photoproducts observed in this work were predicted less toxic by the QSAR methods. The result was expected since the groups known to be responsible for the efficacy of fluoroquinolones were altered or deconstructed through irradiation or photoinduced chemistry. Differences in predictions against the same organisms were due to different programs or databases contained therein. Values for
Daphnia magna were not obtained from the QSAR toolbox. A direct comparison could be achieved for fathead minnow. Here, the values differed by several orders of magnitude, and the toxicity was predicted differently for several product structures, cf.
Table 6. This can be traced back to different databases and calibration models contained in the two programs. The prediction and computation of t.e.s.t. software is based on the database of EUCAST, while the QSAR toolbox calculation is based on the data of the European Chemicals Bureau. A literature search for reference data remained unsuccessful. Despite the different absolute values, both predictions were consistent with reference to the parent drug. Photoproducts were mostly predicted potentially less toxic. An exception was the product with
m/z = 324, which strongly resembled the educt after elimination of the ethyl substituent. Yet, QSAR analyses do not take concentrations into account, hence do not reflect the ecotoxicological activity of a mixture. In this respect, MIC and IC
50 values provide total activity information, although the extent of relevance of the model bacteria to the aquatic environment has not been proven yet.
When irradiation is considered for removal of pharmaceuticals before entering the environment, degradation rate constants in combination with MIC- or IC
50- values allow to determine the irradiation time according to Equation (4). The equation relates activity against bacteria with compound transformation or degradation. Hence, the time to remove potential ecotoxicological hazard from a sample containing LOM could be estimated. The values obtained for LOM samples in MilliQ water and effluent exposed to UVC irradiation are shown in
Table 7.
An irradiation of 10 min was calculated sufficient to remove efficacy against the investigated microorganisms P. fluorescens and B. subtilis regardless of pH and water matrix. In acid milieu, tact is highest. The comparison between different water matrices showed that a longer irradiation time was necessary to compensate for the presence of other compounds and light absorbing substances and to remove the risk of ecotoxicity in effluent water having pH 8.
The use of UV radiation for large-scale compound removal in WWTPs would lead to rather high energy consumption. Cost of goods for sufficiently large UVA and UVC lamps may also significantly increase treatment costs. Nonetheless, UV treatment is applied for microbiological treatment of drinking water. As UV irradiation proved a rather efficient measure for pharmaceutics elimination, UV treatment could be performed on demand. This might be achieved through effluent monitoring and switching on irradiation at need.