Evaluation of Parallel-Series Conﬁgurations of Two-Phase Partitioning Biotrickling Filtration and Biotrickling Filtration for Treating Styrene Gas-Phase Emissions

: The removal of styrene from industrial representative gaseous emissions was studied using two reactors connected in series: a two-phase partitioning biotrickling ﬁlter (TPPB-BTF) and a conventional biotrickling ﬁlter (BTF). The system was operated under industrial conditions, which included steady and transient conditions and intermittent spraying. Silicone oil was used in the TPPB-BTF with a quantity as low as 25 mL L − 1 , promoting a faster start-up compared to the BTF. By working at a styrene loading of 30 g m − 3 h − 1 , nearly complete removal e ﬃ ciency (RE) was obtained. In addition, the removal was not adversely impacted by using non-steady emission patterns such as overnight shutdowns (97% RE) and oscillating concentrations (95% RE), demonstrating its viability for industrial applications. After 2 months from inoculation, two additional conﬁgurations (reverse series BTF + TPPB-BTF and parallel) were tested, showing the series conﬁguration as the best approach to consistently achieve RE > 95%. After 51 days of operation, high throughput sequencing revealed a sharp decrease in the bacterial diversity. In both reactors, the microorganisms belonging to the Comamonadaceae family were predominant and other styrene degraders such as Pseudomonadaceae proliferated preferably in the ﬁrst reactor. being worse than the series conﬁgurations. In any case, BTF performed better than the TPPB-BTF in any of the tested conﬁgurations to handle ﬂuctuating These results indicate that the BTF can achieve similar or even better removals than the TPPB-BTF if enough time is given for the bioﬁlm development. Furthermore, it can be concluded than the series conﬁguration o ﬀ ers a clear safety margin for achieving very low emissions when compared to the parallel conﬁguration. The obtained herein using the series conﬁguration of two


Introduction
The treatment of derivative air emissions of styrene became a priority given its detrimental effects on human health and the environment and its massive use, especially in the industries of coating (styrene resins) and of polystyrene polymers [1]. Apart from its known potential for photochemical ozone creation [2], the styrene classification was recently upgraded to Group 2A, "probably carcinogenic to humans" by the International Agency for Research on Cancer [1]. Therefore, there is a need to develop cost-effective technologies for the control of styrene emissions. The diluted gas-phase emissions of styrene encountered in such industrial activities [3] make the biotechnologies a suitable alternative. Moreover, bioprocesses have some advantages, such as lower energy impact and lower operational and energetic cost when compared to physical-chemical technologies. Among biotechnologies, biotrickling filtration (BTF) stands out-its total direct cost excluding capital recovery could be up to 3 times lower than a regenerative catalytic oxidizer when treating emissions from a fiber reinforce plastic industry [4].
For the three kinds of biotechnologies (biofilters, biotrickling filters and bioscrubbers), the mass transfer step has been identified as the controlling step of the process rather than the biodegradation, especially for poor solubility compounds such styrene. Therefore, different strategies have been developed to enhance the mass transfer [5][6][7]. In this regard, the addition of a non-aqueous phase (NAP) with affinity for styrene seems to be a promising way to improve the process efficiency [5,8]. In addition, other authors have also proven the benefits of introducing an NAP in the removal of o-xylene by BTF [9]. Silicone oil is an inert and non-toxic substance with a low price that appears between the most suitable NAPs due to the high solubility of styrene in it, which is 236 times higher than in water [10]. Previous studies showed the potential of two-phase partitioning bioreactors operated as BTF (TPPB-BTF) for treating styrene air emissions [11,12]. However, the enhanced performance of the TPPB-BTF is most notable at extreme conditions (high loads or low empty bed residence times), while at moderate conditions both configurations exhibited similar performances [12]. Furthermore, the additional cost of the NAP, the clogging problems derived from its use and the extra waste management associated with oil emulsions can compromise its benefits, making BTF more cost-effective than TPPB-BTF, especially at low styrene concentrations. Therefore, the use of both bioreactors in series (TPPB-BTF followed by BTF) could encourage the industrial application of the technology. This combination would take advantage of the TPPB-BTF, while the NAP quantity could be reduced by subsequently finalizing the treatment in a BTF. In addition, laboratory research should consider the influence of the characteristics of industrial gas-phase emissions on the process performance. Typically, the industrial gas-phase emissions of styrene randomly fluctuate with the status of the manufacturing process in the range of 50-350 mg m −3 [4]. Moreover, industrial BTFs should be intermittently irrigated in order to reduce the energetic costs [13,14].
Another aspect of interest lies in the microbiology structure whose analysis would help to gain a better understanding of the process. Among the inoculation alternatives, TPBBs have been seeded with NAPs enriched with microorganisms previously adapted to the pollutant of interest [12]. However, the industrial BTFs are usually seeded with wastewater sludge without previous acclimation as a cost-effective strategy [4,15]. Therefore, the evaluation of TPPF-BTFs seeded with mixtures of industrial silicone-oil, water and wastewater sludge without prior acclimatization to styrene would be a further step for scaling up this technology. Recent high-throughput sequencing analysis identified Azoarcus and Pseudomonas as the dominant genera in BTFs treating styrene, showing that styrene loads modulated the community composition [16]. However, microbiological studies on the influence of the NAP are still scarce.
This work aims to investigate the styrene removal efficiency of a TPPB-BTF plus a BTF operated in series under typical industrial conditions, which included steady and transient conditions, intermittent spraying and inoculation with activated sludge without previous acclimatization. Silicone oil was selected as the NAP for the TPPB-BTF and its influence on the microbial community was analyzed by high throughput sequencing after two months of operation of the TPPB-BTF + BTF configuration. In a second stage, the influence of the NAP on the long-term performance was assessed by comparing the TPPB-BTF + BTF scheme with the reverse one (BTF + TPPB-BTF) and with both bioreactors (TPPB-BTF and BTF) operating in parallel.

Materials
All chemicals were purchased with analytical grade from VWR lnternational Eurolab. Industrial-grade silicone oil with a kinematic viscosity of 50 cSt and a density of 0.96 g mL −1 at 25 • C was used (XIAMETER PMX-200; Univar, Spain). The stock of the mineral salt medium contained (g L −1 ): 19

Experimental set-up
Two bioreactors of identical size were assembled in series, connecting the outlet gas from the first reactor to the inlet gas of the second reactor. The first reactor operated as a two-phase partitioning bioreactor (TPPB-BTF), while the second reactor operated as a conventional biotrickling filter (BTF) (Figure 1) (unless otherwise noted). The bioreactors were cylindrical methacrylate columns (inner diameter of 14.4 cm) filled with 20 L of polypropylene rings (Flexiring ® , Koch-Glitsch B.V.B.A., Belgium) of 25 mm nominal diameter and a surface area of 207 m 2 m −3 . The polypropylene rings had a void fraction and bulk density of 92% and 71 kg m −3 , respectively. Each reactor had its own trickling tank and in order to mimic industrial designs, the tanks' volumes were set up to 10% v/v relative to the packing volume. The air stream, whose flow-rate was adjusted by a mass flow controller (Bronkhorst Hi-Tec, The Netherlands), was polluted with styrene by a syringe pump (New Era, NE 1000 model, USA). The polluted air stream was introduced through the bottom of the column while the liquid stream flushed in counter current mode in a closed loop configuration. The liquid velocity was set to 3 m h −1 in both reactors. The pH of the TPPB-BTF was kept above 7.0 with a pH control system consisting of a pH probe, a digital control unit and a LabQuest ® data acquisition card (Vernier, Oregon, USA). The pH controller was connected to a peristaltic pump (Watson Marlow, USA) which dosed 1 M NaOH stock solution. As a cost-wise criterion, the percentage of silicone oil was set to 2.5% v/v relative to the volume of the packing. Therefore, the TPPB-BTF was seeded with a mixture of silicone oil (0.5 L) and activated sludge (0.5 L), while the BTF was seeded with 0.5 L of activated sludge. Both tanks were completed up to 2 L with water and mineral salt medium.

Experimental set-up
Two bioreactors of identical size were assembled in series, connecting the outlet gas from the first reactor to the inlet gas of the second reactor. The first reactor operated as a two-phase partitioning bioreactor (TPPB-BTF), while the second reactor operated as a conventional biotrickling filter (BTF) (Figure 1) (unless otherwise noted). The bioreactors were cylindrical methacrylate columns (inner diameter of 14.4 cm) filled with 20 L of polypropylene rings (Flexiring ® , Koch-Glitsch B.V.B.A., Belgium) of 25 mm nominal diameter and a surface area of 207 m 2 m −3 . The polypropylene rings had a void fraction and bulk density of 92% and 71 kg m −3 , respectively. Each reactor had its own trickling tank and in order to mimic industrial designs, the tanks' volumes were set up to 10% v/v relative to the packing volume. The air stream, whose flow-rate was adjusted by a mass flow controller (Bronkhorst Hi-Tec, The Netherlands), was polluted with styrene by a syringe pump (New Era, NE 1000 model, USA). The polluted air stream was introduced through the bottom of the column while the liquid stream flushed in counter current mode in a closed loop configuration. The liquid velocity was set to 3 m h −1 in both reactors. The pH of the TPPB-BTF was kept above 7.0 with a pH control system consisting of a pH probe, a digital control unit and a LabQuest ® data acquisition card (Vernier, Oregon, USA). The pH controller was connected to a peristaltic pump (Watson Marlow, USA) which dosed 1 M NaOH stock solution. As a cost-wise criterion, the percentage of silicone oil was set to 2.5% v/v relative to the volume of the packing. Therefore, the TPPB-BTF was seeded with a mixture of silicone oil (0.5 L) and activated sludge (0.5 L), while the BTF was seeded with 0.5 L of activated sludge. Both tanks were completed up to 2 L with water and mineral salt medium.   Table 1. The influence of the spraying frequency on the removal efficiency (RE) and elimination capacity (EC) of the system was evaluated at a constant and continuous styrene loading (phase A, days 0-44). Inlet concentration was fixed at 330 mg m −3 as a   Table 1. The influence of the spraying frequency on the removal efficiency (RE) and elimination capacity (EC) of the system was evaluated at a constant and continuous styrene loading (phase A, days 0-44). Inlet concentration was fixed at 330 mg m −3 as a representative of styrene gas-phase emissions. The empty bed residence time (EBRT) was fixed at 20 s for each reactor, resulting in an EBRT of 40 s for the total system (TPPB-BTF + BTF). Therefore, the inlet load (IL) was of 60 g m −3 h −1 referring to the first reactor (TPPB-BTF) or 30 g m −3 h −1 referring to the whole system. The bioreactors were operated without water purge during the first two weeks to avoid losses of silicone oil and biomass. Afterwards, each water tank was daily purged at a rate of 5% v/v, replacing the purged volume with water and without silicone oil. Nutrient addition was fixed to 40 mL d −1 for the first reactor (TPPB-BTF) and 15 mL d −1 for the second reactor (BTF). From phase B in Table 1, the best spraying frequency (3 min spraying on/9 min spraying off) was selected and styrene feeding was reduced to 16 h d −1 to simulate typical night closures at industry (phase B-I). Fresh air without styrene was supplied at the same flow rate during the other 8 h of the day. In phase B-II, the effect of discontinuous and oscillating inlet concentrations of styrene ranging from 200 to 600 mg m −3 was tested. The EBRT and the average IL were kept constant at 60 g m −3 h −1 and 20 s (data referred to the first reactor). From day 63 onwards, the experimental conditions of phase B-I were restored to evaluate the impact that the oscillating concentrations had on the resilience of the system.

Alternative Parallel-Series Configurations
The order of the reactors was reversed on day 72, and thus the polluted air stream was introduced to the conventional BTF and its outlet air stream was subsequently fed to the TPPB-BTF. Operational conditions of phase B-III were used (discontinuous styrene feeding during 16 h d −1 , inlet concentration of 300 mg m −3 ). Moreover, the same oscillating styrene pattern as that of phase B-II was applied for several days to evaluate the individual response of both reactors in the reverse configuration.
To compare the TPPB-BTF and BTF performances after a long term of operation, the experiment was ended by setting up both reactors in parallel. The air stream polluted with 330 mg m −3 of styrene was split into the two reactors. The IL of the whole system was kept constant at 30 g m −3 Sustainability 2020, 12, 6740 5 of 14 h −1 , while the single EBRT to each reactor was doubled (40 s). Styrene feeding was set to 16 h d −1 . Fluctuating transient conditions (those of phase B-II) were also tested in this configuration.

Microbial Community Analysis
Biomass samples from both bioreactors were taken at day 51 from the bottom and top ports which were located at 20 and 105 cm from the inlet gas. The inoculum sample was also analyzed. DNA was extracted using the PowerSoil DNA isolation kit (Mobio Laboratories, Qiagen, USA). DNA extraction was checked by electrophoresis in an agarose gel. The V4 hyper variable region of the 16S rRNA gene was amplified using the primers 515F (5 -GTG CCA GCM GCC GCG GTA A-3 ) and 806R (5 -GGA CTA CHV GGG TWT CTA AT-3 ). Afterwards, the amplicons were sequenced using the MiSeq system (Illumina, San Diego, USA). The sequences were screened and trimmed by using the Quantitative Insights Into Microbial Ecology (QIIME) software with a sequence length of 200 nt and a mean quality score cut-off of 25 nt. Alpha diversity (within the sample) was assessed by the Shannon diversity index and richness estimator (Chao1), which were calculated by normalizing the lowest number of the sequences for all five samples. Beta diversity (similarity between samples) was assessed using principal coordinates analysis (PCoA) based on weighted UniFrac distances.

Analytical methods
The styrene concentration was monitored by a total hydrocarbon analyzer (Nira Mercury 901, Spirax Sarco, Spain

TPPB-BTF + BTF Series Configuration: Influence of Spraying Frequency
The time course of the inlet and outlet styrene concentration of both reactors (TPPB-BTF and BTF) is shown in Figure 2a. After inoculation, both reactors were irrigated for 24 h per day for promoting the attachment of the biomass to the media and the silicone oil in the TPPB-BTF (phase A-I). The use of silicone oil enhanced the start-up of styrene biotrickling filtration. As it can be observed, TPPB-BTF had a faster start-up than BTF. At day 3, RE of the TPPB-BTF (reactor 1) was 26.3% (outlet concentration = 267 mg m −3 ) while the BTF (reactor 2) reached only 8.6% (outlet concentration = 244 mg m −3 ). This was accompanied by a faster development of biomass (visual inspection) in the TPPB-BTF than in the BTF.
From day 4 onwards, the system was intermittently irrigated for 15 min every hour (15 min spraying on/45 min spraying off, phase A-II). Under these conditions, the decrease in the outlet concentration of both reactors continued following the same tendency as during the start-up: higher improvement in RE for the TPPB-BTF than for the BTF. After~one week of operation under this trickling frequency (day 9), the TPPB-BTF and the BTF achieved styrene outlet concentrations of 133 and 57 mg m −3 respectively, thus corresponding to 60% RE for the TPPB-BTF (TPPB-BTF EC = 35 g m −3 h −1 ) and a RE of 83% for the whole system (total EC = 24 g m −3 h −1 ). The approach for using both bioreactors connected in series (TPPB-BTF plus conventional BTF) allowed a faster start-up of the process with significant savings in the use of silicon oil.
The reduction in the spraying frequency to 15 min every 2 h (phase A-III, days 10-14) caused a worsening of the RE of the TPPB-BTF by decreasing it to 37%, negatively affecting the global performance of the system (total outlet concentration = 166 mg m −3 , total EC = 16 g m −3 h −1 ). The stronger impact on the TPPB-BTF when compared to the BTF was due to its higher nutrient demand, indicating that enlarging the non-frequency periods could limit the biodegradation rate in biotrickling filtration. In fact, when the spraying frequency was increased to 15 min every 1 h (phase A-IV, days 15-20) the RE of the TPPB-BTF improved to 62 ± 3% (EC = 35 ± 2 g m −3 h −1 ), while the combined RE of the TPPB-BTF + BTF was 81 ± 3% (total EC = 23 ± 1 g m −3 h −1 ). From day Sustainability 2020, 12, 6740 6 of 14 21 onwards (phase A-V), the spraying strategy was modified by irrigating each reactor for 3 min every 12 min. The system was evaluated for long-term performance by keeping these conditions for 3 weeks. The RE of the TPPB-BTF remained at similar values, 63 ± 8% (TPPB-BTF EC = 34 ± 4 g m −3 h −1 ), indicating that this reactor already achieved its maximum removal capacity working at an IL of 60 g m −3 h −1 and 20 s of EBRT. Increasing the spraying frequency had a positive effect on the BTF, enhancing the RE of the TPPB-BTF + BTF series configuration up to 91 ± 4% (total EC = 24 ± 1 g m −3 h −1 ). This was accompanied by a higher development of biofilm in the BTF (visual observation), showing that a more frequent nutrient/water addition is advisable to promote biomass growth in the gas-phase styrene-degrading bioreactors, especially at low inlet concentrations. Furthermore, low pressure drops were achieved in both bioreactors (8 Pa m −1 ). Figure 2a. After inoculation, both reactors were irrigated for 24 h per day for promoting the attachment of the biomass to the media and the silicone oil in the TPPB-BTF (phase A-I). The use of silicone oil enhanced the start-up of styrene biotrickling filtration. As it can be observed, TPPB-BTF had a faster start-up than BTF. At day 3, RE of the TPPB-BTF (reactor 1) was 26.3% (outlet concentration = 267 mg m −3 ) while the BTF (reactor 2) reached only 8.6% (outlet concentration = 244 mg m −3 ). This was accompanied by a faster development of biomass (visual inspection) in the TPPB-BTF than in the BTF.

BTF) is shown in
From day 4 onwards, the system was intermittently irrigated for 15 minutes every hour (15 min spraying on/45 min spraying off, phase A-II). Under these conditions, the decrease in the outlet concentration of both reactors continued following the same tendency as during the start-up: higher improvement in RE for the TPPB-BTF than for the BTF. After ~one week of operation under this trickling frequency (day 9), the TPPB-BTF and the BTF achieved styrene outlet concentrations of 133 and 57 mg m −3 respectively, thus corresponding to 60% RE for the TPPB-BTF (TPPB-BTF EC = 35 g m −3 h −1 ) and a RE of 83% for the whole system (total EC = 24 g m −3 h −1 ). The approach for using both bioreactors connected in series (TPPB-BTF plus conventional BTF) allowed a faster start-up of the process with significant savings in the use of silicon oil.  Although there are no literature data on a combined TPPB-BTF + BTF system for styrene removal, several studies can be found regarding TPPB-BTF and BTF alone. Previous authors pointed out that a BTF required a minimum EBRT of 30 s to achieve high and stable removal. For example, San-Valero et al. [12] reported a decrease in the RE from 92% to 66% of a BTF treating a styrene concentration of 184 mg C m −3 when EBRT was decreased from 30 to 20 s (ILs varying from 22 to 33 g C m −3 h −1 ). Similarly, Rene et al. [11] also observed a decrease in the RE for an EBRT of 20 s, although their ILs were much higher than that of this study (211 g m −3 h −1 ). In this regard, the addition of a NAP such as silicone oil allowed the reduction in the EBRT without compromising the performance [11,12], although the main disadvantage derived from the greater amount of silicone oil required. Herein, we demonstrated that the silicone oil usage can be significantly reduced (up to three times) by dividing the required packing volume in two reactors connected in series and operating only the first reactor as a TPPB. By using this strategy, the complexity of the system increased but the silicone oil usage was reduced to the minimum value reported in the literature, facilitating the industrial application. Only a single dose of silicone oil of 0.83 m 3 per kg h −1 of styrene emitted is required to achieve REs >90%.

TPPB-BTF + BTF Series Configuration: Influence of Transient Styrene Loading
Styrene feeding was reduced to 16 h d −1 to evaluate the impact of discontinuous emissions typically found at industrial sites on the RE (phase B-I). The spraying pattern from which the highest RE was achieved (3 min spraying on/9 min spraying off) was used. The results are shown in Figure 2b. Both reactors maintained the same performance as under continuous emissions (phase A-V, Figure 2a). The RE of the TPPB-BTF was 63 ± 6% (EC = 32 ± 4 g m −3 h −1 ) with a combined RE of the series configuration of 90 ± 3% (total EC = 22 ± 1 g m −3 h −1 ). The slight difference in EC with respect to the previous phase is due to a~10% decrease in the inlet styrene concentration associated to inherent variations of the experiment. Therefore, the robustness of this configuration for its application to factories with night closures was corroborated.
Emissions from industrial facilities are usually characterized by fluctuating conditions of loading associated with variations on the manufacturing processes which could hinder the performance of the process. In this study, fluctuating and discontinuous styrene feeding was tested over~2 weeks (days 52-66, phase B-II) by applying two identical 8h cycles with styrene inlet concentrations ranging from 200 to 575 mg m −3 . The average IL during the feeding period was 60 g m −3 h −1 referred to the TPPB-BTF (first reactor) and the EBRT of each reactor was 20 s (total EBRT = 40 s). The transient response for the first 8 h after styrene feeding resumption from night shutdown for a representative day is shown in Figure 3. As it can be observed, the outlet concentration for both reactors followed the inlet concentration pattern. Despite the outlet peaks occurring when inlet styrene rose up to 575 mg m −3 , the rest of the time, low styrene concentrations were emitted from the BTF (reactor 2). The average outlet emission of the system (TPPB-BTF + BTF) during the 8 h cycle was 33 mg m −3 . Although this value was slightly higher when compared to the previous stage (phase B-I, 27 mg m −3 ), it can be concluded that the two bioreactors operating in series is a robust configuration for the efficient management of industrial emission fluctuations. In fact, although the performance of the TPPB-BTF was slightly lower (RE = 60%, EC~37 g m −3 h −1 ) than that of phase B-I, the global RE increased to 95% due to the progressive improvement of the BTF performance (reactor 2). , it can be concluded that the two bioreactors operating in series is a robust configuration for the efficient management of industrial emission fluctuations. In fact, although the performance of the TPPB-BTF was slightly lower (RE = 60%, EC~37 g m −3 h −1 ) than that of phase B-I, the global RE increased to 95% due to the progressive improvement of the BTF performance (reactor 2). On day 64, a constant inlet concentration of 330 mg m −3 for 16 h d −1 was restored for a week to evaluate the resilience of the system after the transitory loading conditions (phase B-III). The RE of the TPPB-BTF was 66 ± 2% (EC = 38 ± 3 g m −3 h −1 ), slightly higher than the RE obtained on the previous days to starting the oscillating emissions (days 45-51). Interestingly, the total system achieved nearly complete removal (97 ± 1%; total EC = 27 ± 2 g m −3 h −1 ). This fact was associated with the clear improvement of the performance of BTF (reactor 2) during the disturbance period, corroborating the slower progress of the BTF on developing enough degrading biofilm. This phenomenon was also On day 64, a constant inlet concentration of 330 mg m −3 for 16 h d −1 was restored for a week to evaluate the resilience of the system after the transitory loading conditions (phase B-III). The RE of the TPPB-BTF was 66 ± 2% (EC = 38 ± 3 g m −3 h −1 ), slightly higher than the RE obtained on the Sustainability 2020, 12, 6740 8 of 14 previous days to starting the oscillating emissions (days 45-51). Interestingly, the total system achieved nearly complete removal (97 ± 1%; total EC = 27 ± 2 g m −3 h −1 ). This fact was associated with the clear improvement of the performance of BTF (reactor 2) during the disturbance period, corroborating the slower progress of the BTF on developing enough degrading biofilm. This phenomenon was also observed in pilot BTFs, where improvements in removal overtime despite styrene load increase were achieved [4,15]. These results show that although the use of silicone oil has a clear impact on the performance of biotrickling filtration during the first weeks after inoculation, its long-term effect remains unclear.

Alternative Series and Parallel Configurations
In order to elucidate the long-term influence of the use of silicone oil during the start-up of BTF treating styrene, two alternative configurations were evaluated after more than two months of operation of the TPPB-BTF + BTF series configuration. Each configuration was tested over a week by using the same styrene feeding than in phase B-III (inlet concentration = 330 mg m −3 during 16 h d −1 ). Firstly, a series configuration reversing the order of the two bioreactors was used, so that the EBRT was maintained to 20 s for each reactor. The BTF was moved to the first position while the second reactor was the TPPB-BTF (BTF + TPPB-BTF series configuration). The BTF reached 74 ± 5% RE (EC = 40 ± 2 g m −3 h −1 ) with a total RE of the two reactors in series of 98 ± 1% (total EC = 26 ±1 g m −3 h −1 ). By comparing the efficiency of the first reactor in the two series configuration, the original and the reverse ones, the BTF showed a slightly higher RE than the TPPB-BTF. Indeed, both configurations achieved a nearly complete removal of styrene, with outlet concentrations <6 mg m −3 , demonstrating that biotrickling filtration is a very competitive alternative to thermal oxidizers for the control of styrene air emissions. These results also reveal that after more than two months of start-up, there were no differences between styrene degradation regarding the presence or absence of a NAP. These results are of high interest for the industrial scalability of the process, as the use of silicone oil increases the operational cost while showing a positive impact on performance only in the short-term.
In the next experiment, both bioreactors were set-up in parallel configuration by splitting the polluted air stream into two parts. Consequently, the EBRT increased by two-fold (40 s each one) and the IL applied to each reactor was~30 g m −3 h −1 . Similar REs were obtained for both reactors (88% ± 1% for the TPPB-BTF and 91% ± 1% for the BTF) with an average outlet styrene concentration of 30.0 ± 8.1 mg m −3 . The results corroborate that after more than 2 months after silicone oil addition, its effect on the process performance could be considered negligible. Despite the greater operational cost due to greater pressure drop, the series configuration is the preferred option versus the parallel one when tight and very low emissions limits are required.
Both configurations (reverse serial and parallel modes) were tested under transient styrene loading (Figure 4). The reverse series configuration (BTF + TPPB-BTF, Figure 4a) showed a similar outlet emission pattern to that obtained with the TPPB-BTF + BTF series configuration (Figure 3). A slightly lower emission was achieved (average outlet emission during 8 h cycle: 18 mg m 3 ) due to the progressive improvement of the BTF overtime. Therefore, the silicone oil does not provide an advantage for buffering peaks two months after its application. It could be hypothesized that the progressive growth and thickening of the biofilm results in outer layers with similar physical and chemical properties than those of a conventional biofilm. For the parallel configuration, both reactors showed similar results (TPPB-BTF: 54 g m −3 , BTF: 40 mg m 3 ), being worse than the series configurations. In any case, BTF performed better than the TPPB-BTF in any of the tested configurations to handle fluctuating emissions. These results indicate that the BTF can achieve similar or even better removals than the TPPB-BTF if enough time is given for the biofilm development. Furthermore, it can be concluded than the series configuration offers a clear safety margin for achieving very low emissions when compared to the parallel configuration. The removal obtained herein using the series configuration of two reactors Sustainability 2020, 12, 6740 9 of 14 (RE >98% for stationary emissions and >95% for oscillating ones) is the best reported for treating 30 g m −3 h −1 of styrene at an EBRT as low as 40 s.
showed similar results (TPPB-BTF: 54 g m −3 , BTF: 40 mg m 3 ), being worse than the series configurations. In any case, BTF performed better than the TPPB-BTF in any of the tested configurations to handle fluctuating emissions. These results indicate that the BTF can achieve similar or even better removals than the TPPB-BTF if enough time is given for the biofilm development. Furthermore, it can be concluded than the series configuration offers a clear safety margin for achieving very low emissions when compared to the parallel configuration. The removal obtained herein using the series configuration of two reactors (RE >98% for stationary emissions and >95% for oscillating ones) is the best reported for treating 30 g m −3 h −1 of styrene at an EBRT as low as 40 s.

Microbial Community Analysis
Pyrosequencing yielded a range from 29,125 to 63,295 raw reads for the inoculum (sample 0, day 0) and for the four biofilm sections (samples 1 to 4, day 51) of the series configuration TPPB-BTF + BTF (Table 2). Therefore, samples 1 to 4 corresponded to biomass exposed to progressive decreases in styrene concentration. Inlet and outlet styrene concentrations were (day 51): 302 and 91 mg m −3 for TPPB-BTF, and 91 and 30 mg m −3 for BTF. From the lowest number of normalized sequences for the five samples (29,125), between 1565 and 565 operational taxonomic units (OTUs) were detected, the biggest number corresponding to the inoculum. The richness estimator (Chao1) and Shannon diversity index are summarized in Table 2. Both parameters achieved much lower values in the samples from the two reactors (samples 1 to 4) than from the inoculum (sample 0), indicating a sharp shift in the bacterial diversity after 51 days of operation. Therefore, a specialization of the microbial community occurred, probably associated with the change in the carbon source from a complex wastewater to styrene only. In addition, the lowest value was obtained for the biomass exposed to the highest styrene concentration (sample 1 nearest to inlet gas to the system containing 302 mg m −3 of styrene). Moreover, the highest value corresponded to the biomass exposed to the lowest styrene concentration (sample 4 nearest to the outlet gas from the system with 30 mg m −3 of styrene). Therefore, this study corroborates that styrene concentration is a critical factor in decreasing the microbial diversity of BTFs, as was previously reported [16]. To evaluate the similarity between samples, beta diversity analysis was carried out through PCoA analysis based on weighted Unifrac distances ( Figure 5). PCoA revealed that the samples from each reactor clustered together, indicating that the recirculation of the trickling water contributes to the homogenization, to some extent, of the microbial population along the reactor. Despite being exposed to similar styrene concentration, samples 2 and 3 showed a greater dissimilarity between themselves than with the other sample of its own reactor due to the use of independent water tanks.  6.58 (6.55-6.60) 1 Operational taxonomic units observed using the normalized sequences (29,125). 2 The 95% confidence intervals of respective estimations are shown in parenthesis. 3 Sample taken at 20 cm bed height from the gas inlet to the reactor. 4 Sample taken at 105 cm bed height from the gas inlet to the reactor.
At the phylum level, three of the five classes of Proteobacteria concentrate the biggest part of the described catabolic pathways to degrade aromatic compounds. In our study, most of the sequences identified belong to the Proteobacteria phylum, with a relative abundance ranging from 57.0% to 60.1% in every sample. Previous studies on BTFs degrading styrene also found that Proteobacteria was the predominant phylum [16][17][18]. The relative abundance of the dominant families (with a relative abundance >1% at least in one sample) is presented in Figure 6. As it can be seen, there was a shift in the predominant ones from the inoculum (sample 0) to the four samples of day 51 due to the adaptation of the microbial population to styrene as a unique exogenous carbon source. Among the top 22 classified families, those predominant in the inoculum (sample 0) were identified as Rhodocyclaceae, Saprospiraceae and Comamonadaceae, with a relative abundance of 13.0%, 11.3% and 9.8% of total sequences, respectively. Furthermore, the dominant families observed on day 51 (samples 1-4) were dependent on the reactor, since the water trickling of each reactor homogenized the bacterial population through its bed packing. In the TPPB-BTF (samples 1-2), Comamonadaceae, Pseudomonadaceae, Weeksellaceae, and Nocardiaceae were detected in high relative abundances (average values: 23.9%, 10.1%, 8.5%, and 6.5% of total sequences, respectively). In the BTF (samples 3-4), Comamonadaceae, Xanthomonadaceae, Sphingobacteriaceae and Nocardiaceae made up the dominant members (average values: 20.4%, 17.6%, 10.4% and 6.7% of total sequences, respectively). At the phylum level, three of the five classes of Proteobacteria concentrate the biggest part of the described catabolic pathways to degrade aromatic compounds. In our study, most of the sequences identified belong to the Proteobacteria phylum, with a relative abundance ranging from 57.0% to 60.1%  Biofilm samples from both reactors had two predominant families in common, suggesting that styrene was the critical factor, despite its concentration, to shape the bacterial population, which was greater in the TPPB-BTF (samples 1-2) than in the BTF (samples 3-4). The highest percentage of sequences that were classified to known families across all bioreactor samples belonged to the Comamonadaceae, with relative percentages greater than 20% of the total sequences. Indeed, its relative abundance was more than doubled from the inoculum (9.8%). In this regard, genes for styrene monooxygenase, which encodes the initial enzyme in the metabolic pathway for degradation of styrene, have been identified in some species of the Comamonadaceae family. Thus, they have been reported as one with the highest metabolic potential to degrade aromatic compounds [19,20]. The other common dominant family detected in both reactors at very similar percentages in all samples (6.6% ± 1.3%) was Nocardiaceae. Interestingly, this family was not detected as an abundant family in the inoculum. Some species belonging to Nocardiaceae family have been described as capable to mineralize styrene as Nocardia spp. and Rhodococcus spp. [16,[21][22][23]. On the other side, the styrene concentration rather than the presence or absence of silicone oil seems to play a role in the proliferation of the other dominant families, such as Pseudomonadaceae, Sphingobacteriaceae and Xanthomonadaceae. The Pseudomonas genus has been frequently studied in the aerobic degradation of styrene [24][25][26][27][28]. This genus belongs to the Pseudomonadaceae family and it was the second in relative abundance in the TPPB-BTF reactor. Furthermore, it is remarkable that there was a positive tendency between the relative abundance of the Pseudomonadaceae and the styrene concentration. The relative abundance of the Pseudomonadaceae family decreased as styrene concentration decreased through the system TPPB-BTF + BTF (11.9%, 8.2%, 3.5% and 1.7% for samples 1, 2, 3, and 4 respectively). It could suggest that species from this family are favored by high styrene concentrations versus other families such as Sphingobacteriaceae and Xanthomonadaceae. These two families show an opposite tendency to Pseudomonadaceae, more abundant in the second reactor (BTF) than in the first reactor (TPPB-BTF) of the series configuration. Therefore, their abundances were enhanced at low styrene concentrations. Recently, microorganisms belonging to the Sphingobacteriaceae family have been associated with the biodegradation of expanded polystyrene [29]. Arnold et al. [30] identified several styrene degraders belonging to families Tsukamurellaceae, Sphingomonadaceae, Xanthomonadaceae and Pseudomonadaceae in biofilms from a peat biofilter and Portune et al. [16,31] also observed some styrene degraders such Biofilm samples from both reactors had two predominant families in common, suggesting that styrene was the critical factor, despite its concentration, to shape the bacterial population, which was greater in the TPPB-BTF (samples 1-2) than in the BTF (samples 3-4). The highest percentage of sequences that were classified to known families across all bioreactor samples belonged to the Comamonadaceae, with relative percentages greater than 20% of the total sequences. Indeed, its relative abundance was more than doubled from the inoculum (9.8%). In this regard, genes for styrene monooxygenase, which encodes the initial enzyme in the metabolic pathway for degradation of styrene, have been identified in some species of the Comamonadaceae family. Thus, they have been reported as one with the highest metabolic potential to degrade aromatic compounds [19,20]. The other common dominant family detected in both reactors at very similar percentages in all samples (6.6% ± 1.3%) was Nocardiaceae. Interestingly, this family was not detected as an abundant family in the inoculum. Some species belonging to Nocardiaceae family have been described as capable to mineralize styrene as Nocardia spp. and Rhodococcus spp. [16,[21][22][23]. On the other side, the styrene concentration rather than the presence or absence of silicone oil seems to play a role in the proliferation of the other dominant families, such as Pseudomonadaceae, Sphingobacteriaceae and Xanthomonadaceae. The Pseudomonas genus has been frequently studied in the aerobic degradation of styrene [24][25][26][27][28]. This genus belongs to the Pseudomonadaceae family and it was the second in relative abundance in the TPPB-BTF reactor. Furthermore, it is remarkable that there was a positive tendency between the relative abundance of the Pseudomonadaceae and the styrene concentration. The relative abundance of the Pseudomonadaceae family decreased as styrene concentration decreased through the system TPPB-BTF + BTF (11.9%, 8.2%, 3.5% and 1.7% for samples 1, 2, 3, and 4 respectively). It could suggest that species from this family are favored by high styrene concentrations versus other families such as Sphingobacteriaceae and Xanthomonadaceae. These two families show an opposite tendency to Pseudomonadaceae, more abundant in the second reactor (BTF) than in the first reactor (TPPB-BTF) of the series configuration. Therefore, their abundances were enhanced at low styrene concentrations. Recently, microorganisms belonging to the Sphingobacteriaceae family have been associated with the biodegradation of expanded polystyrene [29]. Arnold et al. [30] identified several styrene degraders belonging to families Tsukamurellaceae, Sphingomonadaceae, Xanthomonadaceae and Pseudomonadaceae in biofilms from a peat biofilter and Portune et al. [16,31] also observed some styrene degraders such as Dokdonella inmobilis and Brevundimonas from families Xanthomonadaceae and Caulobacteraceae in biofilms from BTF. In addition, Liao et al. [18] showed bacterial genera Achromobacter belonging to Alcaligenaceae family as potential degraders of styrene, and Zhong et al. [32] reported the families Pseudomonadaceae, Caulobacteraceae and Sphingomonadaceae as dominant bacterial families in styrene degrading biofilters.

Conclusions
The operation of two biotrickling filters connected in series, one of them working as a TPPB, has been proven to be robust for the efficient treatment of styrene emissions under typical industrial conditions. The series configuration TPPB-BTF + BTF has been demonstrated as the best alternative since nearly complete removal of 30 g m −3 h −1 both in stationary and oscillating emissions was obtained. In addition, this approach allowed for reducing the silicone oil quantity to a minimum single dose of 0.83 m 3 per kg h −1 of styrene emitted. However, the effect of silicone oil on the process performance could only be corroborated during the first weeks, and it was mainly associated with a faster start-up and biofilm development of the TPPB-BTF when compared to the BTF. In fact, the conventional BTF achieved slightly better performance than the TPPB-BTF after two months of operation. Therefore, from an industrial point of view, the use of silicone oil should be carefully evaluated due its high cost. Microbial community analysis showed enrichment in identified styrene degraders belonging to Comamonadaceae, Nocardiaceae and Pseudomonadaceae families with their relative abundances through the system modulated by the styrene concentration.