Gaseous emissions during the solid state fermentation of different wastes for enzyme production at pilot scale

The emissions of Volatile Organic Compounds (VOC), H4, N2O and NH3 during the Solid State Fermentation process of some selected wastes to obtain different enzymes have been determined at pilot scale. Orange peel + compost (OP), hair wastes + raw sludge (HW) and winterization residue + raw s ludge (WR) have been processed in duplicate in 50L reactors to provide emission fa ctors and to identify the different VOC families present in exhaust gaseous emissions. Ammonia emission from HW fermentation (3.2 ± 0.5 kg Mg -1 dry matter) and VOC emission during OP processes ( 18 ± 6 kg Mg dry matter) should be considered in an industrial application of these processes. Terpenes have been the most emitted VOC family during all the processes although the emission of sulphide molecules during HW SSF is notable. The most emitted compound was dimethyl disulfide in HW and W R processes, and limonene in the SSF of OP.


Introduction
In biotechnology, solid-state fermentation (SSF) represents an interesting alternative to submerged fermentation (SmF) processes for the production of some industrially useful compounds (Singhania et al., 2009). Indeed, SSF presents some advantages compared to SmF such as higher concentrations of gaseous oxygen with higher overall mass transfer capacity, volume reduction of the fermentation vessel owing to a lower water volume and the possibility of carrying out the process using non-sterilized substrates (Subramaniyam and Vimala, 2012). Moreover, the use of organic wastes suppose a clear advantage towards a sustainable alternative process.
Enzymes are one of the valuable products that can be successfully produced through SSF. Lipases, proteases and cellulases have been obtained processing selected organic wastes such as residual hair from the tanning industry, soybean, pomace waste, orange peels or oil cakes (Abraham et al., 2014). These enzymes have an important role in biotechnological, pharmaceutical and chemical industries. They can also be used in food applications or wastewater treatment. There is a general interest in obtaining new, more specific and stable enzymes which can be produced through SSF processes (Thomas et al., 2013).
Many works in literature refer the use of specific microorganisms in SSF processes. Another alternative is to mix a common waste or amendment that will act as inoculum with a specific waste that will act as the substrate to produce the enzyme (Santís-Navarro et al., 2011). Compost or sludge from a wastewater treatment plant are materials that can be used as inoculum and also to adjust moisture or pH levels. The specific waste will depend on the final added-value product desired e.g.: oil cake (as a source of lipids to produce lipases), hair waste (as a source of proteins to produce proteases) or orange peel (as a source of cellulose to produce cellulases). It is technically and economically more unfavorable to use as inoculum a pure culture of a known microorganism that can degrade a specific waste (Dhillon et al., 2012).
Once the useful enzymes have been obtained, the biodegradation process should continue to produce a stabilized product that can be used as an organic amendment (Thomasa et al., 2013). Thus, when wastes are being used as substrates, SSF can be considered a waste valorization process where a value added product and a final organic amendment are obtained from the biological degradation of the waste under controlled conditions, closing the cycle for the organic matter.
The environmental impacts of the SSF process must be taken into account when considering SSF benefits. However, to our knowledge, this point has not been studied in the scientific literature. Gaseous emissions play an important role in the environmental impact of processes used to manage solid organic substrates (Colón et al., 2012).
Greenhouse gases (GHG), ammonia and volatile organic compounds (VOC) are the main gases emitted during SSF processes.
Regarding GHG, biogenic CO 2 is the main compound emitted although it is not normally accounted as GHG (Colón et al., 2012). Also CH 4 and N 2 O are often released, and present a global warming potential of 34 and 298 kg CO 2 eq., respectively (IPPC, 2013). On the other hand, the emission of VOC and ammonia has been associated to the generation of unpleasant odors and other environmental impacts such as acidification, eutrophication or the formation of photochemical oxidants. Terpenes, alcohols, ketones, sulfur-containing compounds and amines have been reported among the VOC emitted from waste handling and treatment processes described in literature (Komilis et al., 2004).
The objective of this work is to study the emissions of VOC, CH4, N2O and NH3 during the production of different enzymes through SSF using specific wastes. Emission factors (amount of compound emitted per amount of waste treated) have been determined for the studied compounds, providing data to compare the different processes. Special attention has been addressed to VOC emissions identifying and quantifying the main VOC present.

Materials
Raw sludge (RS) from a wastewater treatment plant located in Manresa (Barcelona, Spain) and final compost from a mechanical-biological treatment plant located in Montcada i Reixac (Barcelona, Spain) were used as inoculum. The wastes used were orange peel from Valencia (Spain), winterization residue provided by an oil refining facility in Montcada i Reixac and hair wastes from a local tanning industry in Igualada (Barcelona, Spain). Orange peel was mixed with compost (1:0.1, w:w), winterization residue was mixed with raw sludge (0.1:1, w:w) and hair wastes were mixed also with raw sludge (1:2, w:w). The ratios of each specific waste to inoculum were different for each waste since the composition of the raw materials is different (Table S1). These ratios were adjusted in previous experiments at smaller scale (Abraham et al., 2014;Santís-Navarro et al., 2011). All these mixtures were finally mixed (1:1, v:v) with wood chips to adjust moisture content and porosity to a value within the range recommended for the adequate development of the biodegradation process under full aerobic conditions (Ruggieri et al., 2009). A homogeneous sample of each mixture and raw material was separated and stored at -18 ºC to be used for waste characterization prior to introducing the mixture in the reactors. It is reported that freezing did not change the biological activity of the waste during this period.
The main characteristics of the initial mixture and the final products obtained from each experiment are presented in Table S1 (Supplementary information). Dry and organic matter and pH were determined in triplicate following the standard procedures (US Department of Agriculture and US Composting Council, 2001).

Pilot plant
The results presented in this study were obtained, in duplicate, in 50 L pilot plant

Sampling
Gas samples were collected in 1-L Tedlar® bags for VOC, N 2 O and CH 4 determination. Also a 250-mL glass gas collector was used for VOC composition determination and identification. In all cases, one sample was taken per day and per reactor.
Reactors were opened twice a week to monitor the degradation process. In these cases, the material was homogenized and a 0.5 kg sample was taken to analyze moisture, pH, organic matter and enzymatic activity. A total of 2.5 kg were withdrawn from each reactor. Emissions potentially released during sampling were quantified prior sampling by stripping all the gas in the reactor with airflow of 10 L/min for 5 minutes and collecting this gas in a 50 L Tedlar sampling bag.

Enzymatic activity
The procedure followed to determine lipase activity is detailed in Santis-Navarro et al. (2011). The determination of enzymatic activity from proteases is described in Abraham et al. (2014). Extract for cellulase activity determination was obtained by mixing 10 g of a wet sample, from the 0.5 kg taken from the reactor, with 150 mL of citrate buffer 4.8 (1:10 w/v ratio). The enzyme extraction was carried out by incubating the samples in a wrist action shaker for 30 min at 100 rpm. After incubation, the samples were centrifuged at 10000 g and 4°C for 15 min and the supernatant was analyzed for cellulase activity (Dhillon et al., 2012). To measure total cellulase activity (TC), the modified filter paper assay, was used (Ghose, 1987). Instead of using filter paper strips as substrate, microcrystalline and insoluble Sigmacell cellulose type 20 µm was used (high purity cellulose, Sigma-Aldrich). Modifications were mainly made due to several authors report filter paper assay as very laborious and time consuming method, requiring many manual manipulations that lead to non-reproducible results (Coward et al., 2003). Carboxymethylcellulase (CMCase) and β-glucosidase activity were also measured according to Ghose (1987). Released reducing sugars were measured by 3,5-dinitrosalicylic acid (DNS) method (Miller, 1959). For TC and CMCase, one activity unit (UA) was defined as the amount of enzyme that produces 1 µmol of reducing sugar (glucose equivalent) per minute. For β-glucosidase one activity unit (UA) was defined as the amount of enzyme which produces 1 µmol of glucose per minute.

Stability degree
The dynamic respiration index (DRI) of initial mixtures and final products was measured in triplicate using a dynamic respirometer (Ponsá et al., 2010) on the basis of the methodology proposed by Adani et al. (2006) to assess the degree of biological stability.

Determination of gaseous emissions
VOC, CH 4 and N 2 O analysis was performed by means of gas chromatography (Agilent Technologies 6890N Network GC system, Madrid, Spain) as explained in Colón et al. (2012). All samples were analyzed in triplicate for each compound. The deviation found per each triplicate was lower than 10% for all the compounds.
Ammonia concentration was measured in situ at the gas outlet of the composting reactor by means of an ammonia sensor (Industrial Scientific sensor iTX-T82, Oakdale, PA, USA) with a measurement range of 0 to 1200 ppmv. The sensor was placed inside a hermetic recipient with inlet and outlet holes that allowed gas circulation.

GC-MS detection
VOC characterization was performed using gas samples analyzed by SPME

Process evolution
Temperature, airflow, oxygen concentration and OUR were monitored on line, while enzymatic activity was determined twice a week, in the six SSF trials. Figure S1 (Supplementary information) shows the profiles for all these parameters obtained in one of the two process replicates undertaken for each of the three mixtures studied: orange peel + compost (OP), winterization residue + raw sludge (WR) and hair wastes + raw sludge (HW). Similar OUR, oxygen, temperature and airflow profiles were observed in the duplicates for the same waste.
The process evolution for the SSF of orange peel with compost ( Figure S1a) was as follows: thermophilic conditions were achieved in the third day of process, although maximum temperature was not achieved until the twelfth process day (OP-I: 67.43ºC; OP-II: 61.21ºC). The delay in temperature peaking may be due to different factors: initial acidic pH of the mixture (5.88, Table S1) rose during the process reaching a final value of 8.45; the lack of homogeneity of the initial mixture was overcome through successive mixing when sampling, contributing to mass transfer between the different materials inside the reactor; finally, a potential inhibition by the presence of limonene is further discussed in this paper. As expected, the lowest oxygen values (OP-I: 11.8 % OP-II: 9.6 %) and the highest OUR values (OP-I: 10 with lowest oxygen content (7.6 %) and highest airflow value in HW-I (1.7 ml min -1 ). In the HW-II experiment, minimum oxygen value (5.23 %) and maximum airflow supplied (4.7 ml min -1 ) were recorded slightly later, on the third and fourth day respectively. However, on both replicates the maximum protease activity was found on the fourth day (HW-I: 27135 UA g -1 DM; HW-II: 22400 UA g -1 DM).
According to the temperature profile and the values of DRI of the final material (Table S1), the process evolution was correct, and the final product was stabilized  Figure 1 presents the evolution of the emissions of the four gaseous pollutants studied through the degradation process. Table 1 Figure S1a). However a peak was also observed on the fifth day due to the low airflow supplied at that moment and the oxygen decrease ( Figure S1a).  (Fukumoto et al., 2003). In Figure S1 it can be observed that thermophilic temperatures were achieved during the first week of process (for HW and WR) pointing that N 2 O was emitted even at thermophilic temperatures. There were no considerable differences between the emission factors obtained for N 2 O in the different processes studied (Table 1) Table 1 for raw sludge and HW, it can be deduced that ammonia was emitted mainly from hair wastes in HW trials although raw sludge also contributes to these emissions. A similar trend was also observed during WR processes, in this case ammonia emissions were clearly related to the presence of raw sludge, as the winterization residue did not contribute to the nitrogen content of the mixture. Values in Table S1 show considerable higher nitrogen content in HW compared with the other mixtures. Also, nitrogen contain in WR is higher than in OP (Table S1), where no ammonia emissions were detected (Table 1). However, WR emissions were lower than RS emissions due to the higher bioavailable C/N ratio (Puyuelo et al., 2011).

N 2 O emissions
To enhance the properties of the final product to be used as an organic amendment, it would be beneficial to reduce ammonia emissions retaining nitrogen in the solid phase. Further investigation on the strategies to reduce these significant nitrogen losses is necessary to assume the feasibility of a complete valorization process of the hair wastes, as it has been highlighted in other studies (Puyuelo et al., 2011). Figure 1d presents the evolution of VOC emissions. It should be noted than different scales on y axis have been used for OP than for WR and HW due to the large amount of VOC emitted during OP processes. Once again, the evolution of VOC emission is related to temperature. The VOC emission peak (Figure 1d, right axis) matches with OUR and temperature peak (Figure S1b and S1c) in WR and HW processes. This was not observed in the OP processes where VOC were mainly emitted from day 2 to day 6 (37-50ºC) but higher thermophilic temperatures were reached on days 11 to 16. On the other hand, VOC emission in first OP processing day was more than 10-fold the emission from HW or WR even if these residues were under thermophilic conditions. Differences in temperature, OUR and VOC emissions during OP processes, show that VOC emission was not only a consequence of the organic matter degradation, as has been previously reported (Pagans et al., 2006b). OP processes emitted important quantities of VOC from the beginning of the process, as a consequence of the stripping occurred when the airflow passes through the solid matrix.

VOC emissions
VOC emissions increased when temperature and biological activity (OUR) started to rise up. Moreover, VOC emissions in OP were one order of magnitude larger than in HW and WR processes until day 15. The difference in VOC emission could be also observed in Table 1, being the OP emission factor 100-fold the emission factor of the other processes.
To summarize, the emissions of the studied compounds were mainly produced when maximum biological activity (maximum temperature and maximum OUR values) was achieved. An appropriate treatment to avoid the release of important amounts of VOC and NH 3 to the atmosphere must be designed if these SSF process are implemented at industrial scale, as it has been previously reported in studies on related topics such as composting (Pagans et al., 2006b). The emission factors calculated in this study could be useful for this design.

VOC characterization by SPME/GC-MS
A large number of different volatile organic compounds was identified in the emissions from the six processes. These compounds were classified into the following chemical families: alcohols, esters, furans, ketones, aliphatic hydrocarbons, aromatic hydrocarbons, aldehydes, halogenated compounds, nitrogen-containing compounds, sulfur-containing compounds, volatile fatty acids and terpenes. Table 2 summarizes the percentage of each VOC family for each process. A quantitative analysis was also carried out for fifteen specific and typical VOC emitted during similar biodegradation processes (Scaglia et al., 2011). Eucalyptol and decane were not found and the 13 characterized compounds and their amounts are summarized in Table 3.
Broadly, similar distribution percentages were found for furans, esters, volatile fatty acids, aliphatic hydrocarbons, aldehydes and nitride molecules among the different processed wastes, being the less abundant families in the gaseous emissions. The differences between the maximum and the minimum percentage found in the emission of these compounds in the six SSF processes were below 2.9 points. Alcohols, ketones and aromatic hydrocarbons presented some differences depending on the treated waste, with differences between the maximum and the minimum percentage over 12.9 points.
The emission of sulphide molecules presented a higher dispersion, with a difference of Aldehydes also presented low percentages in OP and HW processes and were not detected in WR. As can be seen in Table 3, decanal and nonanal were the aldehydes found in HW processes and decanal and hexanal the ones found in OP processes.
Probably, in the case of HW, these compounds come from hair wastes, because no aldehydes were found either in RS or in WR.
Ketones were found in higher percentages in the processes using raw sludge. 2pentanone was not present in OP processes, while it was found in all the other trials.
Similarly happens for nitrogen-containing compounds. This family is represented by pyridine, a product related to a putrid odor (Suffet et al., 2009). Pyridine was found and quantified mainly in process containing raw sludge. Pyridine was emitted in HW in one order of magnitude higher than in all the other processes. This fact could be also a consequence of the high nitrogen content of hair wastes (Table S1).
Aromatic hydrocarbons were found in all the experiments. However, toluene, styrene and xylene were not found in HW trials (Table 3) where phenol, benzene and naphthalene were the main aromatic hydrocarbons emitted.
Terpenes are the family presenting higher emissions in all the experiments (Table 2). This dominance was also observed in other studies where organic wastes were biodegraded in similar conditions (Maulini-Duran et al., 2014, 2013a, 2013b).
Terpenes have been described as the main compounds responsible for odorous pollution in similar biodegradation processes dealing with organic wastes such as composting (Staley et al., 2006). Terpenes could be produced as an intermediate in the aerobic metabolism. During OP trials, the percentage of terpenes in gaseous emissions was above 80%. Terpenes dominance in OP processes and the higher percentage compared to the other processes is mainly due to limonene. Limonene is related to the fruity or citrusy sensorial experience (Suffet et al., 2009). Limonene is a harmless compound widely used in fragrances, but some studies revealed that reacting in ambient with ozone could form dangerous secondary pollutants, including formaldehyde (Walser et al., 2007). Table 3 shows that limonene emission factor is, at least, 10-fold higher in OP than in all the other processes studied. In fact, there are some studies of biological processes involving orange peels that report the need to remove limonene as a pre-treatment because it is considered an anti-microbial agent (Martin et al., 2010). In the light of these studies, the lag phase observed at the beginning of the OP process can also be attributed to the high concentration of limonene in the material. The increase in airflow observed during the first days of process, which coincides with an increment of total VOC emissions ( Figure S1a and Figure 1d), enhanced the release of limonene by stripping, as can be observed in Figure S2, which shows the daily evolution of limonene emission factor during OP processes. The lower limonene concentration inside the reactor could favor the microbial proliferation from day 6 th on. This fact could explain the late appearance of the thermophilic peak and its mismatch with the total VOC emission peak, both discussed before in this study.
Alpha and beta pinene have been specifically related to emissions coming from woody bulking agents (Maulini-Duran et al., 2014). In the present study the amount of bulking agent was the same in all the processes and, as can be observed in Table 3, the emission factor for these two compounds is quite similar among the different SSF trials. This confirms the study by Büyüksönmez and Evans (2007), working with green waste composting, who reported that terpenes emission is closely related to the material used as bulking agent.
Percentages of sulfur-containing compounds during HW processes were higher than those from OP, WR and RS and also higher than those observed in previous works (Maulini-Duran et al., 2014, 2013a, 2013b. It can be seen in Table 2 that the percentage of sulfur-containing compounds was similar in RS and WR but much higher for HW. The same raw sludge was used in RS, WR and HW trials. In consequence, the differences on sulfur-containing compounds between HW and RS or WR can be attributed to hair wastes. As discussed before, the hair structure contains large quantities of cysteine, a sulfur-containing compound. The degradation of cysteine provokes a higher emission of volatile sulfur compounds (López del Castillo-Lozano et al., 2008).
In fact, as can be observed in Table S1 sulfur content in HW is higher than WR.
Dimethyl disulfide was detected as the main sulfur-containing compound and it was quantified (Table 3). Dimethyl disulphide is a strong odorant and moderately toxic volatile compound that was detected in most of the samples analyzed during the different experiments, often over its odor threshold (EPA, 2010), 0.007 mg m -3 .
According to the percentage of sulfur-containing compounds the process with the highest emission factor of dimethyl disulfide was HW.
Most of the VOC emitted in the studied processes are harmless, but some of them are or could be transformed into hazardous compounds for the environment and for human health as it has been particularly explained for limonene (Walser et al., 2007).
Summarizing, terpenes is the most abundant emitted family for all the treated wastes, but the emission of sulfur-containing compounds during HW processes is also remarkable. Therefore, the most emitted compounds were dimethyl disulfide in HW and WR processes and limonene in OP SSF.

Conclusions
A systematic study on SSF gaseous emissions has been presented using pilot easily scalable reactors. Emission factors are strongly dependent on the waste processed under SSF. For instance, for most of the wastes, emission factors for CH 4  Hair waste SSF process is characterized by high emission of ammonia.
Winterization and hair waste SSF processes were also characterized by the presence of sulfur-containing compounds; while in orange peel SSF process limonene was emitted in higher amount.