Scholarly article on topic 'Microbial community dynamics and biogas production from manure fractions in sludge bed anaerobic digestion'

Microbial community dynamics and biogas production from manure fractions in sludge bed anaerobic digestion Academic research paper on "Biological sciences"

0
0
Share paper
Academic journal
J Appl Microbiol
OECD Field of science
Keywords
{""}

Academic research paper on topic "Microbial community dynamics and biogas production from manure fractions in sludge bed anaerobic digestion"

Applied Microbiology

Journal of Applied Microbiology ISSN 1364-5072

ORIGINAL ARTICLE

Microbial community dynamics and biogas production from manure fractions in sludge bed anaerobic digestion

A.S.R. Nordgard1, W.H. Bergland2, R. Bakke2, O. Vadstein1, K. 0stgaard1 and I. Bakke1

1 Department of Biotechnology, Norwegian University of Science and Technology (NTNU), Trondheim, Norway

2 Department of Process, Energy and Environmental Technology, Telemark University College (TUC), Porsgrunn, Norway

Keywords

biodegradation, denaturing gradient gel electrophoresis, granules, microbial structure, polymerase chain reaction.

Correspondence

Anna Synnove Rostad Nordgard, Department of Biotechnology, Norwegian University of Science and Technology (NTNU), Sem Stands vei 6/8, 7491 Trondheim, Norway. E-mail: anna.s.r.nordgard@ntnu.no

The reactors were operated at Telemark University College, Porsgrunn, Norway, while the microbial analyses were executed at Norwegian University of Science and Technology, Trondheim, Norway.

2015/1335: received 22 July 2015, revised 2 September 2015 and accepted 4 September 2015

doi:10.1111/jam.12952

Abstract

Aims: To elucidate how granular sludge inoculum and particle-rich organic loading affect the structure of the microbial communities and process performance in upflow anaerobic sludge bed (UASB) reactors. Methods and Results: We investigated four reactors run on dairy manure filtrate and four on pig manure supernatant for three months achieving similar methane yields. The reactors fed with less particle rich pig manure stabilized faster and had highest capacity. Microbial community dynamics analysed by a PCR/denaturing gradient gel electrophoresis approach showed that influent was a major determinant for the composition of the reactor communities. Comparisons of pre- and non-adapted inoculum in the reactors run on pig manure supernatant showed that the community structure of the nonadapted inoculum adapted in approximately two months. Microbiota variance partitioning analysis revealed that running time, organic loading rate and inoculum together explained 26 and 31% of the variance in bacterial and archaeal communities respectively.

Conclusions: The microbial communities of UASBs adapted to the reactor conditions in treatment of particle rich manure fractions, obtaining high capacity, especially on pig manure supernatant.

Significance and Impact of the Study: These findings provide relevant insight into the microbial community dynamics in startup and operation of sludge bed reactors for methane production from slurry fractions, a major potential source of biogas.

Introduction

Traditional completely mixed reactors are consistently used in agriculture and wastewater sludge treatment at the present. Upflow anaerobic sludge bed (UASB) reactors can be more than 50 times more efficient than these (Lettinga et al. 1997; Barber and Stuckey 1999; Tchobanoglous et al. 2003; von Sperling and Oliveira 2009). High rate UASB reactors may treat more waste in smaller and presumably much cheaper digesters, but the technology for small-scale biogas facilities is at an early stage. The manure storage may serve as the first step in a treatment line allowing disintegration and hydrolysis of particles and separation of influent supernatant with low content of suspended debris. We recently demonstrated

that high rate UASB reactors can be used for anaerobic digestion (AD) of supernatant from pig manure to obtain sustainable energy recovery despite the solids content being well above the levels considered appropriate as UASB reactor influent (Bergland et al. 2015). The microbial community structure and dynamics in such high rate processes for AD of manure supernatant has so far not been investigated.

The microbial community in an anaerobic digester is highly influenced by its influent (Ziganshin et al. 2013). This involves the influent chemical makeup, e.g. fats, proteins and inhibitors (Chen et al. 2008; Sousa et al. 2008; Kovacs et al. 2013) and its inherent microbial community as it is continuously introduced to the reactor. The microbial community of e.g. municipal solid waste is

different from those found in dairy cow manure (Nari-hiro and Sekiguchi 2007; Hagen et al. 2014). Fluctuations in operational variables may also influence the reactors biota and include feeding pattern, hydraulic retention time (HRT), organic loading rate (OLR), levels of volatile fatty acids (VFA), pH, ammonium content and temperature (Sun et al. 2014). Each biogas fermenter is therefore a unique system defined by its substrate and process conditions (Jenicek et al. 2010; Krakat et al. 2011; Ziganshin et al. 2013; Rosa et al. 2014). However, the interactions and the roles of the consortia of micro-organisms involved are still poorly understood. Knowledge about how the microbial communities are influenced by reactor design and operational variables is needed to improve design and proper operation of reactors.

Molecular biology tools enable the study of microbial diversity without a cultivation step and thus overcome the cultivation bias. An efficient approach to investigate microbial communities is to amplify either the 16S rRNA gene (Kim et al. 2011; Demirel 2014; Madden et al. 2014; Tuan et al. 2014) or the gene specific to the functional microbial group of interest (Lueders et al. 2001; Galand et al. 2002; Nettmann et al. 2008; Gagnon et al. 2011) by PCR, and analyse the PCR products by denaturing gradient gel electrophorese (DGGE). Culture-independent molecular techniques based on 16S rDNA and functional genes have helped linking microbial community structure and dynamics to process performance (Nakasaki et al. 2013).

The objective of this study was to examine UASB performance with manure supernatants with high particle contents as influent and to investigate the microbial community dynamics by PCR/DGGE using 16S rRNA gene amplicons. The aim was to elucidate how the influent, granular sludge inoculum and OLR affected the structure of the bacterial and archaeal communities in the biogas reactors.

Experimental procedures

Reactor influent and inoculum

A total of eight laboratory scale reactors were studied. The filtrate of sieved dairy cow manure (named dairy manure for short hereafter) collected from Foss Farm in Skien, Norway, was used as influent for a period of 96 days in four reactors (CA1, CA2, CB1 and CB2). These reactor names are abbreviated C (all reactors run on dairy manure), CA (parallel CA1 and CA2) or CB (parallel CB1 and CB2). The HRT of the CA reactors was decreased towards the end of the experiment while the CB reactors were kept at constant HRT. Dairy manure handling is described in Bergland et al. (2014) while the

influent properties are in Table 1. The other four reactors (PA1, PA2, PB1 and PB2) were fed pig manure slurry supernatant (named pig manure for short hereafter) for 106 days. These reactor names are abbreviated P (all reactors run on pig manure), PA (parallel PA1 and PA2) or PB (parallel PB1 and PB2). The manure was collected from a production farm in Porsgrunn, Norway. Pig manure handling is described in Bergland et al. (2015) while the influent properties are in Table 1. Both manure slurries were stored at 4°C until use.

The granules (70 g VSS l_1) used as inoculum originated from a UASB reactor treating pulp and paper process wastewater at 'Norske Skog Saugbrugs' in Halden, Norway. For six of the reactors, the granular inoculum had been stored for six months at 11°C prior to the experiment. The remaining two reactors (PA1 and PA2) were run on pig manure influent six months prior to the experiment to adapt the granular sludge inoculum. All the reactors were filled half way up with granules at the start of the experiment.

Reactor design and operation

Lab-scale process lines were set up utilizing identical pulse fed 370 ml UASB reactors with a liquid volume of 354 ml. The reactor design and measurements of COD, pH, VFA, NH^-N, gas composition and methane production are described in Bergland et al. (2015). The C reactors were all started at HRT 1-77 days together with two of the pig manure fed reactors, PB, while the PA reactors previously adapted to pig manure were started at 0-35 days HRT (Table 2). Effects of load increases were investigated by running periods of reduced HRT, after the biogas production had stabilized in the PA and PB reactors. The HRTs were reduced during days 35-68 by 5% per day. The corresponding OLRs are in Table 2. Production never stabilized in the C reactors, and hence the HRTs were decreased by 5% per day for the CA

Table 1 Properties of the influent for the anaerobic digestion reactors (Average and SD)

Pig manure slurry Dairy manure

Property supernatant filtrate

pH 7-3 ± 0-3 7-5 ± 0-2

CODt (g T1) 28-1 ± 2-7 48-3 ± 7-6

CODs (g T1) 16-0 ± 2-8 13-1 ± 2-2

CODvfa (g r1) 12-2 ± 1-1 5-8 ± 0-9

CODparticle 12-1 35-2

Acetate (g i_1) 5-5 ± 0-8 3-2 ± 0-5

Propionate (g i_1) 1-9 ± 0-4 0-8 ± 0-2

NH4 - N (g T1) 2-35 ± 0-04 0-90 ± 0-14

Table 2 HRT and OLR at the time of sampling for microbial analysis

HRT (day) OLR (g CODT l"1 day"1)

Dairy Pig Dairy Pig

Time (day) CA CB PA PB CA CB PA PB

4 1-77 1-77 0-35 1-8 27 27 79 24

35 1-77 1-77 0-35 1-77 27 27 79 16

61 1-77 1-77 0-12 0-5 27 27 245 56

68 1-77 1-77 0-07 0-35 27 27 397 79

96 0-45 1-77 0-1 0-35 107 27 283 79

103 - - 0-1 0-21 - - 283 135

105 - - - 0-17 - - - 163

HRT, hydraulic retention time; OLR, organic loading rate.

reactors during days 71-96 while HRT of the CB reactors were never changed.

Sampling, DNA extraction, PCR amplification and DGGE

Samples were collected from the effluent line of each reactor on day 35, 68 and 96 for the C reactors and on day 35, 61, 68, 96, 103 (PA only) and 105 (PB only) for the P reactors. The first three samples from the P reactors have been subject to microbial analyses previously (Bergland et al. 2015), but are included here for continuance. Samples of both influents were taken on day 35, 68 and 96. The nonadapted granular sludge inoculum was sampled and DNA extracted on two occasions prior to this experiment (six months and one year). Granular sludge inoculum that had been adapted to pig manure was not sampled prior to the experiment start-up.

DNA was extracted using the Power Soil DNA isolation kit from MO BIO Laboratories as described by the manufacturer. PCR was performed using the primer pairs GC-338f (cgcccgccgcgcgcggcgggcggggcgggggcacgggggg actcctacgggaggcagcag) and 518r (attaccgcggctgctgg) amplifying the v3 region of the 16S rRNA gene in bacteria (Muyzer et al. 1993) and GC-624f (cgcccgccgcgcgcg-gcgggcggggcgggggcacgggggg caccdrtggcgaaggc) and 820r (gccrattcctttaagtttca) amplifying the v5 region in archaea (Bergland et al. 2015). The PCR products were analyzed by DGGE as described by Bakke et al. (2013) using 8% acrylamide gels with a denaturing gradient of 35-55% for the bacterial PCR products and 35-50% gradient for the methanogenic archaeal PCR products.

Statistical analysis

The gel pictures were analyzed using Gel2K (Svein Nordland, Department of Microbiology, University of Bergen,

Norway). This program converts the band profiles in DGGE images to histograms where the peaks correspond to the intensity in the DGGE bands. The peak area matrices were exported to Excel where the values were normalized and square root transformed to reduce the impact of strong bands. The matrices were then exported to PAST ver. 2.17 (Hammer et al. 2001) for statistical and multi-variate analysis. Principal coordinate analysis (PcoA) (Davis 1986) was based on ordination of Bray-Curtis similarities (Bray and Curtis 1957). One-way permanova (Anderson 2001) was employed for testing the differences in average Bray-Curtis similarities between different groups of samples. Variance partitioning (Borcard et al. 1992) was carried out with the varpart function in Vegan, a package in R (R Development Core Team 2014), which partitions the variation of the DGGE peak area matrices with respect to two, three, or four explanatory tables (Oksanen et al. 2013). Continuous explanatory variables in variance partitioning were transformed to standard normal distribution before analysis.

Results

Reactor performance

The start-up and 5% daily increase periods of the PA and PB reactors are described in Bergland et al. (2015). The end of the 5% increase period resulted in a maximum OLR of 397 g CODT l"1 day"1 for PA reactors (Fig. 1a) with HRT 0-07 days and a maximum biogas production of 34 NL methane l" reactor day"1 (Fig. 1b). The PB reactors which started at a lower load had a maximum OLR of 163 g CODT l"1 day"1 at day 106 with HRT 0-17 days and a biogas production of 16 NL methane l"1 reactor day"1. Reduced methane production and propionate removal (Fig. 2) at the highest loads of the PA reactors were signs of stress but the process did not fail and maintained production throughout the experiment, however, with relatively low yield (Fig. 1c). The PB reactors without preadapted inoculum, not exposed to such extreme loads, achieved higher methane yield than PA except for the first two weeks (Fig. 1c).

The C reactors, all inoculated with nonadapted granules, took a long time to reach stable biogas production (Fig. 1). The CB reactors were therefore kept at constant load of 29 g CODT l"1 day"1 (Fig. 1a). They slowly increased the methane production yield and showed increasing methane production rate (Fig. 1). From day 71, the CA reactors got 5% daily influent flow increase from HRT 1-77 to 0-45 days and OLR 29 to 107 g CODT l"1 day"1 resulting in increased methane production rates up to 3-1 NL methane l"1 reactor day"1 for CA1 and 4-5 for CA2, introducing the largest devia-

Time (d)

Time (d)

Time (d)

Figure 1 Organic loading rate (a), methane production rate (b) and methane yield (c) for pig manure fed reactors PA (■) and PB (▲), and dairy manure fed reactors CA (□) and CB (m) (average parallel 1 and 2). Error bars represent standard deviation between parallel reactors.

tion between two CA parallel reactors in this study. The standard deviations between effluent samples from parallel reactors are given as error bars in Figs 1 and 2, while often not visible due to similar reactor behaviour. The influent is without error bars, since the reactors receive the same feed.

The yield of the CB reactors increased during the whole test to a maximum of 4-2 l methane l-1 manure at 27 g CODT T1 day-1 OLR, never reaching the level of the stable PB reactors of 56 g CODT l-1 day-1 OLR with 5 NL methane l-1 manure. Methane yield for the CA reactors dropped with increased load to 1-7 NL methane l-1 influent at 107 g CODT l-1 day-1. Acetate concentrations for the C reactors (Fig. 2a) shows increasing removal with stabilized 83% acetate removal after day 57. Insignificant propionate removal was observed until day 80 after which 73-92% propionate removal were observed at day 96 for the CB reactors (Fig. 2b). Propi-onate removal did not occur in the two CA reactors exposed to load increase after day 71. Acetate (Fig. 2c) and propionate (Fig. 2d) removal was considerably higher

in the pig manure fed reactors where the propionate levels were low and stable in all cases except for the peak loads.

The COD removal varied between 9-50 and 24-68% for reactors fed with dairy and pig manure respectively. The pH was stable for all the reactor effluents with 7-78-3 for pig reactors also at very high loads and 7-5-8-0 for dairy reactors. The methane content was 71-82% for pig manure fed reactors and 75-82% for dairy manure fed reactors.

Microbial community dynamics in reactors fed dairy or pig manure

A DGGE gel with samples from the CA1, CB1, PA1 and PB1 reactors were run to examine the influence of the influent and granular sludge inoculum on microbial communities in the reactors (Fig. S1). Ordination by PcoA based on Bray-Curtis similarities for the microbial communities associated with the influent, the granular inoculum and the reactor slurries (Fig. 3), indicates that the reactor sludge bacterial and archaeal communities

Figure 2 Concentrations and standard deviations in influent (O) and effluent from the dairy manure fed reactors CA (□) and CB (m) and pig manure fed reactors PA (□) and PB (m). Dairy manure acetate (a) and propionate (b), pig manure acetate (c) and propionate (d). Error bars represent standard deviation between parallel reactors.

Time (d)

Time (d)

Time (d)

Time (d)

differed between the reactors run on dairy and pig manure. A permanova test confirmed this (P < 0-0002).

The microbial communities in the reactor sludge seemed to be more similar to the communities in the influent than those associated with the granular sludge inoculum (Fig. 3). Average Bray-Curtis similarities support this (Table 3). Interestingly, for both bacterial and

archaeal communities, the average Bray-Curtis similarities for comparisons of reactor sludge and influent communities were approximately twice those calculated for comparisons of reactor sludge and granular sludge inoculum communities. A t-test confirmed that these differences were significant (P < 0-0001). This suggests that the influent had a higher impact on microbial community

or 0-1

-03 -02 -0 1

0 01 Coordinate 1

or o 0-1

-03 -02 -01 0 01 Coordinate 1

Figure 3 Principal coordinate analysis ordination based on Bray-Curtis similarities for the (a) bacterial and (b) archaeal community profiles of the granular sludge inoculum (x) for the pig manure influent (•) and reactors PA1 (■) and PB1 (▲), and dairy manure influent (O) and reactors CA1 (□) and CB1 (m). Samples from granular inoculum that had been pre-adapted to pig manure (used for the PA reactors only) were not available. Samples were collected from the effluent line of each reactor on day 35, 68 and 96 for the C reactors and on day 35, 61, 68, 96, 103 (PA only) and 105 (PB only) for the P reactors. Samples of both influents were taken on day 35, 68 and 96. The non-adapted granular sludge inoculum was sampled six months and one year prior to the experiment.

composition in the reactors than the granular sludge inoculum, and that the influent is a strong determinant for the reactor microbial communities.

Moreover, the archaeal community profiles were more similar between sample types (influent, reactor sludge and granular inoculum) than the bacterial communities (t-test, P < 0-0001; Table 3). This was confirmed by determining the average Bray-Curtis similarity for comparisons among all samples, which was found to be significantly higher (t-test, P = 4"28) for archaea (0-69 ± 0-11) than for bacteria (0-56 ± 0-13). The Shannon's diversity was also more variable across sample types

for bacterial than for archaeal communities (data not shown), corroborating the finding that bacterial community structures were more variable than the archaeal ones.

Temporal developments and effects of operational parameters and granular sludge inoculum on the community structure in the reactors running on pig manure supernatant

The PA and PB reactors were run with the same influent under stable and similar conditions except for different granular sludge inoculum (preadapted for PA and non-adapted for PB), and varying OLRs and HRTs (Table 2). To investigate the effect of different granular sludge inoculum and varying OLR on the reactor microbial communities, a DGGE gel (Fig. S2) was run with samples from all the P reactors and all sampling time points. The PcoA ordination based on Bray-Curtis similarities indicate a difference in the microbial community profiles between the PA and PB reactors for both bacteria and archaea (Fig. 4). This was confirmed by a permanova test (P < 0-01) and indicates an influence of the granular sludge inoculum on the reactor sludge communities. The PcoA plot further suggests that the bacterial communities (Fig. 4a) of the four reactors were more divergent at the first two sampling dates, but became more similar from day 68 onwards. This trend seemed to be more pronounced for the PB reactors inoculated with nonadapted granules. Average Bray-Curtis similarities for comparisons of bacterial communities between PA and PB reactors were found to be lowest at day 61 (0-52 ± 0-06) and highest at day 96 (0-78 ± 0-2). A similar tendency was seen for the archaeal communities (Fig. 4b), with average Bray-Curtis similarities for comparisons of PA and PB communities being lowest at day 61 (0-52 ± 0-01) and highest at day 96 (0-81 ± 0-1).

Variance partitioning was used to evaluate the influence of running time, OLR and type of granular sludge inoculum on the variation in microbial community composition in the PA and PB reactors. Only 26 and 31% of the variation in microbial community structure for bacteria and archaea, respectively, was explained by these three variables. However, most of this variance was explained by interaction effects, and they accounted for 58 and 88% of the explained variance in bacterial and archaeal community composition respectively. Running time alone accounted for 7%, the granular sludge inoculum for 4% and the OLR for 2% of the variance of bacterial community structure. The same variables accounted for only 0, 0 and 5%, respectively, of the variance in the archaeal community structure. These three variables thus had small impacts separately. Running time and OLR, for example, explained together 9% of the variance in the bacterial

Table 3 Average Bray-Curtis similarities and standard deviation for comparison of microbial communities in reactor sludge (average of parallel 1 and 2) to those in the granular sludge inoculum and in the influent

Bacteria Archaea

Dairy (CA, CB) Pig (PB)* Dairy (CA, CB) Pig (PB)*

Reactor sludge vs gr. sl. inoculum 0-29 ± 0-04 0-35 ± 0-03 0-44 ± 0-05 0-46 ± 0-03

Reactor sludge vs influent 0-66 ± 0-07 0-62 ± 0-07 0-81 ± 0-05 0-79 ± 0-04

*PA was left out due to lack of samples of the pre-adapted inoculum used for the PA reactors.

Figure 4 Principal coordinate analysis ordination based on Bray-Curtis similarities for (a) bacterial and (b) archaeal communities in pig manure fed reactors PA1 (■), PA2 (□), PB1 (▲) and PB2 (m).

community while the highest explanatory effect found was on the archaeal community where 21% of the variance was due to the impact of granular sludge inoculum and OLR.

Discussion

The CA and CB reactors needed longer time to stabilize the methane yield, acetate removal and propionate removal than the PA and PB reactors, which may be a

consequence of the higher fraction of slowly degradable particles in dairy manure (Table 1). The methane yields obtained at high pig manure OLRs compared to the more particle rich dairy manure also indicates that pig manure supernatant is more suitable than dairy manure filtrate as UASB influent. The yield as NL methane l-1 influent is, however, quite similar for dairy and pig manure at the end. The generally low VFA content in the effluents, especially the low propionate concentration, are signs of a well-functioning AD process on dairy manure, even at the relatively high OLR of 27 g COD l-1 day-1 compared to recommended wastewater UASB loadings of 12-18 g COD l-1 day-1 (Tchobanoglous et al. 2003). Reduced propionate removal, as observed at the highest pig manure loads, can be explained by low growth rate and inhibition due to high levels of acetate and/or hydrogen (Bergland et al. 2015). High concentrations of these propionate removal products are thermodynamically unfavourable for propionate reduction (Batstone et al. 2002) and can occur during load increase as a stress symptom. The CA performance is comparable to that observed by Rico et al. (2011) with 3-1 NL methane l-1 dairy manure influent at load 72-5 g CODT l-1 day-1. The CB reactors had the best performance in term of methane yield and propionate removal (Figs 1c and 2b) towards the end of the experiment. This gradual process performance improvement was not reflected as a significant shift in the microbial community, suggesting that there is an alternative explanation. Improved mass transfer in the granular sludge due to morphological changes as an adaptation to particle rich feed is a possibility that cannot be verified the methods applied in this study.

Our findings imply that the influent is a strong determinant for the structure of both bacterial and archaeal reactor communities: First, we found that microbial communities differed significantly between reactors run with dairy and pig manure. Secondly, we saw that the micro-bial community structure in the reactors were more similar to the community structure in the influent than in the granular sludge inoculum for both bacteria and archaea (Fig. 3, Table 3). The influent can affect the reactor microbial communities both by selection due to its physical and chemical makeup and by its inherent

microbial community that is constantly introduced to the reactors. An experimental strategy comparing the effects of sterile and nonsterile influents on the reactor communities might be helpful for disentangling these two mechanisms. Ziganshin et al. (2013) stated that even though the microbial community dynamics may be strongly influenced by the type of inoculum, the influent is still the largest contributor to the microbiota in the reactor sludge, thus corroborating our results.

The bacterial communities in the reactors, influent and granular sludge inoculum were significantly more variable among bacterial samples than archaeal samples (Table 3). In addition, the variations in Shannon's diversity among the microbial communities of the granular sludge inoculum, influent and reactors indicate a more stable community structure for archaea than bacteria. Previous studies of anaerobic digesters report greater diversity within bacterial than archaeal communities in addition to a higher abundance of bacteria (Li et al. 2013; Yang et al. 2014). This probably reflects that the bacterial communities are characterized by a larger functional redundancy than the archaeal communities. This would allow for more variation in bacterial community structure among samples, as observed in our study. Whole bacterial genera may be substituted when process variables vary, but the niches are not left vacant due to the high diversity of Bacteria (Krakat et al. 2011; De Schryver and Vadstein 2014). This could explain the higher variability among samples that was observed for bacterial communities (Table 3).

Analysis of the communities in the PA and PB reactors demonstrated that using distinct granular inoculum (pre-adapted or non-adapted) also resulted in significantly different archaeal and bacterial communities. However, the communities developed to become more similar with time (Fig. 4). This development probably reflects adaptation of the microbial communities as a response to similar selective pressure exerted by the influent on the communities in the PA and PB reactors, and indicates a decreasing influence of the original granular sludge inoculum during the first two months of operation. The granular inoculum came from a UASB reactor pulp and paper process wastewater treatment with almost no particles, high C/N ratio and high fraction of easily degradable carbohydrates, which is different from manure. Hence, our results extend the findings in Bergland et al. (2015), namely that granular sludge from easily digestible, particle free wastewater treatment can be used for complex, particle rich UASB dairy and pig manure treatment. A longer adaptation period is, however, required for dairy manure than for pig manure.

Even though we found that the inoculum affected the PA and PB reactor microbial communities, particularly at

the beginning of the experiment, only a minor part of the variance in community structure among samples throughout the experiment was explained by the inoculum in the variance partitioning analysis. In addition, this analysis showed that only 2 and 5% of the variance in the bacterial and archaeal communities variations observed between samples are explained by the OLR respectively. However, considerable interaction effects were observed, and for the archaeal communities 21% of the variance was explained by granular sludge inoculum and OLR. The OLR affected reactor performance: high OLRs resulted in decreased yield (Fig. 1c) and acetate/propi-onate removal rates (Fig. 2c, d). The variance partitioning analysis suggests that this is not due to changes in community structure. The relatively small effect of OLR on the microbial communities may be caused by the low HRTs employed in this study. When HRT is lower than the maximum growth rate of the micro-organisms, the planktonic microbiota may be a reflection of the communities associated with the influent more so than the active part of the microbiota that reside in the solid granular sludge phase. Microbes in the granules are expected to be less influenced by the OLR and HRT, although a potential effect on community structure might be mediated by changes in diffusion to the microbes in the granules imposed by the OLR.

Due to the low HRT in our system, most of the biomass in the reactors is present in the granules since planktonic organisms are washed out. The archaea, mainly situated in the deeper layers of granules in UASB reactors (Sekiguchi et al. 1999), was expected to be less influenced by OLR than the bacteria which mainly reside in the outer layers, but the opposite was observed. The porosity of biomass aggregates tend to change with OLR (van Loosdrecht et al. 2002) so sludge morphology maybe as important as community structure in the adaptation to load changes. This is an interesting topic for further research.

A four years long experiment conducted by Krakat et al. (2011) shows that HRT, OLR and the substrate influence the bacterial community but even using beet silage as sole substrate did not give rise to unique community structures linked to process parameters. Fluctuating bacterial communities are rather characteristic to such habitats. Furthermore, Rosa et al. (2014) observed differences in the microbial communities in two UASB reactors utilizing a mixture of cassava processing wastew-ater and glucose that was due to the effect of seed sludge and HRT. The highest explanatory effect found in the present study was the impact of granular sludge inoculum and OLR on the archaeal community while the running time and OLR had the highest impact on the bacterial communities.

HRT was not included in the variance partitioning since the COD concentration in the influent was fairly constant, implying that OLR and HRT are closely coupled variables. The variance partitioning analysis could not explain the majority of the variation in community structure observed throughout the experiment. Temporal changes in the chemical and physical properties of the influent, resulting in altered selective forces in the reactors, could have caused changes in the structure of the microbial communities. Further, stochastic processes, like drift, are assumed to contribute to the structuring of microbial communities (Leibold and McPeek 2006; Hanson et al. 2012). Deep sequencing providing information on taxonomy and abundance would contribute to a better understanding of the microbial community dynamics of UASB biogas reactors.

In summary, the UASB fed the more particle rich dairy manure filtrate needed longer time to stabilize than those fed pig manure slurry supernatant. Higher peak production and ability to handle varying loads are additional indications that pig manure slurry supernatant is a more suitable feed for UASB reactors than dairy manure. The yield, measured as methane per litre influent is, however, quite similar for dairy and pig manure at the end, and the generally low vFA content in all reactors at stable loads, imply that UASB treatment is a promising approach for both influents.

The results from the microbial analysis of this study illustrated that the influent is a strong determinant of the microbial community composition in UASB reactors running on dairy manure filtrate or pig manure slurry supernatant. Archaeal communities were found to be less variable over time and between samples than the bacterial communities. The granular inoculum adapted to the manure influents in approximately two months. Non-adapted inoculum may therefore be utilized in the startup of new UASB biogas reactors for dairy or pig manure treatment.

The running time, OLR and granular sludge inoculum explained 26 and 31% of the total variance in the bacterial and archaeal communities respectively. The three parameters had a low impact on the communities separately, but there were considerable interaction effects.

Acknowledgements

The project was supported by the Norwegian Agricultural Authority, Innovation Norway, The Research Council of Norway and the Biogas for Norwegian Agriculture project.

The authors thank Eivind Fjelddalen and Associate Professor Finn Haugen for the reactors automatic process monitoring and control. We also thank Randi Utgard

and Blanca Magdalena Gonzalez Silva for helping with

DNA extraction.

Conflict of Interest

No conflict of interest is declared.

References

Anderson, M.J. (2001) A new method for non-parametric multivariate analysis of variance. Austral Ecol 26, 32-46.

Bakke, I., Skjermo, J., Vo, T.A. and Vadstein, O. (2013) Live feed is not a major determinant of the microbiota associated with cod larvae (Gadus morhua). Environ Microbiol Rep 5, 537-548.

Barber, W.P. and Stuckey, D.C. (1999) The use of the

anaerobic baffled reactor (ABR) for wastewater treatment: a review. Water Res 33, 1559-1578.

Batstone, D.J., Keller, J., Angelidaki, I., Kalyuzhnyi, S.V.,

Pavlostathis, S.G., Rozzi, A., Sanders, W.T.M., Siegrist, H. et al. (2002) The IWA Anaerobic Digestion Model No 1 (ADM1). Water Sci Technol 45, 65-73.

Bergland, W., Dinamarca, C. and Bakke, R. (2014) Efficient biogas production from the liquid fraction of dairy manure. RE&POJ 12, 519-521.

Bergland, W.H., Dinamarca, C., Toradzadegan, M., Nordgard, A.S.R., Bakke, I. and Bakke, R. (2015) High rate manure supernatant digestion. Water Res 76, 1-9.

Borcard, D., Legendre, P. and Drapeau, P. (1992) Partialling out the spatial component of ecological variation. Ecology 73, 1045-1055.

Bray, J.R. and Curtis, J.T. (1957) An ordination of the upland forest communities of southern Wisconsin. Ecol Monogr 27, 326-349.

Chen, Y., Cheng, J.J. and Creamer, K.S. (2008) Inhibition of anaerobic digestion process: a review. Bioresour Technol 99, 4044-4064.

Davis, J.C. (1986) Statistics and Data Analysis in Geology. New York: John Wiley & Sons.

De Schryver, P. and Vadstein, O. (2014) Ecological theory as a foundation to control pathogenic invasion in aquaculture. ISME J 8, 2360-2368.

Demirel, B. (2014) Major pathway of methane formation from energy crops in agricultural biogas digesters. Crit Rev Environ Sci Technol 44, 199-222.

Gagnon, N., Barret, M., Topp, E., Kalmokoff, M., Masse, D., Masse, L. and Talbot, G. (2011) A novel fingerprint method to assess the diversity of methanogens in microbial systems. FEMS Microbiol Lett 325, 115-122.

Galand, P.E., Saarnio, S., Fritze, H. and Yrjala, K. (2002) Depth related diversity of methanogen Archaea in finnish oligotrophic fen. FEMS Microbiol Ecol 42, 441-449.

Hagen, L.H., Vivekanand, V., Linjordet, R., Pope, P.B., Eijsink, V.G.H. and Horn, S.J. (2014) Microbial community

structure and dynamics during co-digestion of whey permeate and cow manure in continuous stirred tank reactor systems. Bioresour Technol 171, 350-359.

Hammer, 0., Harper, D.A.T. and Ryan, P.D. (2001) PAST: Paleontological Statistics software package for education and data analysis. Palaeontol Electronica 4, 9.

Hanson, C.A., Fuhrman, J.A., Horner-Devine, M.C. and Martiny, J.B.H. (2012) Beyond biogeographic patterns: processes shaping the microbial landscape. Nat Rev Microbiol 10, 497-506.

Jenicek, P., Koubova, J., Bindzar, J. and Zabranska, J. (2010) Advantages of anaerobic digestion of sludge in microaerobic conditions. Water Sci Technol 62, 427-434.

Kim, J., Shin, S.G., Han, G., O'Flaherty, V., Lee, C. and

Hwang, S. (2011) Common key acidogen populations in anaerobic reactors treating different wastewaters: molecular identification and quantitative monitoring. Water Res 45, 2539-2549.

Kovacs, E., Wirth, R., Maroti, G., Bagi, Z., Rakhely, G. and Kovacs, K.L. (2013) Biogas production from protein-rich biomass: fed-batch anaerobic fermentation of casein and of pig blood and associated changes in microbial community composition. PLoS ONE 8, 18.

Krakat, N., Schmidt, S. and Scherer, P. (2011) Potential

impact of process parameters upon the bacterial diversity in the mesophilic anaerobic digestion of beet silage. Bioresour Technol 102, 5692-5701.

Leibold, M.A. and McPeek, M.A. (2006) Coexistence of the niche and neutral perspectives in community ecology. Ecology 87, 1399-1410.

Lettinga, G., Field, J., vanLier, J., Zeeman, G. and Pol, L.W.H. (1997) Advanced anaerobic wastewater treatment in the near future. Water Sci Technol 35, 5-12.

Li, A., Chu, Y.N., Wang, X.M., Ren, L.F., Yu, J., Liu, X.L., Yan, J.B., Zhang, L. et al. (2013) A pyrosequencing-based metagenomic study of methane-producing microbial community in solid-state biogas reactor. Biotechnol Biofuels 6, 17.

van Loosdrecht, M.C.M., Heijnen, J.J., Eberl, H., Kreft, J. and Picioreanu, C. (2002) Mathematical modelling of biofilm structures. Antonie Van Leeuwenhoek 81, 245-256.

Lueders, T., Chin, K.J., Conrad, R. and Friedrich, M. (2001) Molecular analyses of methyl-coenzyme M reductase alpha-subunit (mcrA) genes in rice field soil and enrichment cultures reveal the methanogenic phenotype of a novel archaeal lineage. Environ Microbiol 3, 194204.

Madden, P., Al-Raei, A.M., Enright, A.M., Chinalia, F.A., de Beer, D., O'Flaherty, V. and Collins, G. (2014) Effect of sulfate on low-temperature anaerobic digestion. Front Microbiol 5, 15.

Muyzer, G., Dewaal, E.C. and Uitterlinden, A.G. (1993)

Profiling of complex microbial-populations by denaturing gradient gel-electrophoresis analysis of polymerase chain

reaction-amplified genes-coding for 16S ribosomal-RNA. Appl Environ Microbiol 59, 695-700.

Nakasaki, K., Kwon, S.H. and Ikeda, H. (2013)

Identification of microorganisms in the granules generated during methane fermentation of the syrup wastewater produced while canning fruit. Process Biochem 48, 912-919.

Narihiro, T. and Sekiguchi, Y. (2007) Microbial communities in anaerobic digestion processes for waste and wastewater treatment: a microbiological update. Curr Opin Biotechnol 18, 273-278.

Nettmann, E., Bergmann, I., Mundt, K., Linke, B. and

Klocke, M. (2008) Archaea diversity within a commercial biogas plant utilizing herbal biomass determined by 16S rDNA and mcrA analysis. J Appl Microbiol 105, 1835-1850.

Oksanen, J., Blanchet, F.G., Kindt, R., Legendre, P., Minchin, P.R., O'Hara, R.B., Simpson, G.L., Solymos, P. et al. (2013) Vegan: Community Ecology Package. R package version 2.0-10. http://CRAN.R-project.org/package=vegan

R Core Team (2014) R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing. http://www.R-project.org.

Rico, C., Garcia, H. and Rico, J.L. (2011) Physical-anaerobic-chemical process for treatment of dairy cattle manure. Bioresour Technol 102, 2143-2150.

Rosa, P.R.F., Santos, S.C., Sakamoto, I.K., Varesche, M.B.A. and Silva, E.L. (2014) The effects of seed sludge and hydraulic retention time on the production of hydrogen from a cassava processing wastewater and glucose mixture in an anaerobic fluidized bed reactor. Int J Hydrogen Energy 39, 13118-13127.

Sekiguchi, Y., Kamagata, Y., Nakamura, K., Ohashi, A. and Harada, H. (1999) Fluorescence in situ hybridization using 16S rRNA-targeted oligonucleotides reveals localization of methanogens and selected uncultured bacteria in mesophilic and thermophilic sludge granules. Appl Environ Microbiol 65, 1280-1288.

Sousa, D.Z., Pereira, M.A., Alves, J.I., Smidt, H., Stams, A.J.M. and Alves, M.M. (2008) Anaerobic microbial LCFA degradation in bioreactors. Water Sci Technol 57, 439444.

von Sperling, M. and Oliveira, S.C. (2009) Comparative

performance evaluation of full-scale anaerobic and aerobic wastewater treatment processes in Brazil. Water Sci Technol 59, 15-22.

Sun, L., Muller, B., Westerholm, M. and Schnurer, A. (2014) Syntrophic acetate oxidation in industrial CSTR biogas digesters. J Biotechnol 171, 39-44.

Tchobanoglous, G., Burton, F.L. and Stensel, H. (2003)

Wastewater Engineering: Treatment and Reuse. New York, NY: McGraw-Hill.

Tuan, N.N., Chang, Y.C., Yu, C.P. and Huang, S.L. (2014) Multiple approaches to characterize the microbial community in a thermophilic anaerobic digester running

on swine manure: a case study. Microbiol Res 169, 717-724.

Yang, Y., Yu, K., Xia, Y., Lau, F.T.K., Tang, D.T.W., Fung, W.C., Fang, H.H.P. and Zhang, T. (2014) Metagenomic analysis of sludge from full-scale anaerobic digesters operated in municipal wastewater treatment plants. Appl Microbiol Biotechnol 98, 5709-5718.

Ziganshin, A.M., Liebetrau, J., Proter, J. and Kleinsteuber, S. (2013) Microbial community structure and dynamics during anaerobic digestion of various agricultural waste materials. Appl Microbiol Biotechnol 97, 5161-5174.

Supporting Information

Additional Supporting Information may be found in the online version of this article:

Figure S1 DGGE-gel with PCR products obtained with the primer set a) GC-338f/518r targeting bacteria and b) GC-624f/820r targeting archaea.

Figure S2 DGGE-gel with PCR products obtained with the primer set a) GC-338f/518r targeting bacteria and b) GC-624f/820r targeting archaea.