Bioresource Technology 101 (2010) 2243–2251

Contents lists available at ScienceDirect

Bioresource Technology

journal homepage: www.elsevier .com/locate /bior tech

Long-chain fatty acids inhibition and adaptation process in anaerobicthermophilic digestion: Batch tests, microbial community structureand mathematical modelling

J. Palatsi a,b, J. Illa c, F.X. Prenafeta-Boldú a,d, M. Laureni a,b, B. Fernandez a, I. Angelidaki b, X. Flotats a,e,*

a GIRO Technological Centre, Rambla Pompeu Fabra 1, E-08100 Mollet del Vallés, Barcelona, Spainb Department of Environmental Engineering, Technical University of Denmark, Building 113, DK-2800 Lyngby, Denmarkc Department of Computer Science and Industrial Engineering, University of Lleida, Jaume II 69, E-25001 Lleida, Spaind IRTA, Passeig de Gràcia 44, 3ª pl., E-08007 Barcelona, Spaine Department of Agrifood Engineering and Biotechnology, Universitat Politècnica de Catalunya, Parc Mediterrani de la Tecnologia Edifici D-4, E-08860 Castelldefels, Barcelona, Spain

a r t i c l e i n f o

Article history:Received 27 August 2009Received in revised form 11 November 2009Accepted 16 November 2009Available online 16 December 2009

Keywords:Thermophilic anaerobic digestionLCFA inhibition–adaptation16S rDNA profilingADM1 model

0960-8524/$ - see front matter � 2009 Elsevier Ltd. Adoi:10.1016/j.biortech.2009.11.069

* Corresponding author. Address: GIRO TechnologFabra 1, E-08100 Mollet del Vallés, Barcelona, Spain. T93 5796785.

E-mail address: xavier.flotats@giroct.irta.cat (X. Flo

a b s t r a c t

Biomass samples taken during the continuous operation of thermophilic anaerobic digestors fed withmanure and exposed to successive inhibitory pulses of long-chain fatty acids (LCFA) were characterizedin terms of specific metabolic activities and 16S rDNA DGGE profiling of the microbial community struc-ture. Improvement of hydrogenotrophic and acidogenic (b-oxidation) activity rates was detected uponsuccessive LCFA pulses, while different inhibition effects over specific anaerobic trophic groups wereobserved. Bioreactor recovery capacity and biomass adaptation to LCFA inhibition were verified. Popula-tion profiles of eubacterial and archaeal 16S rDNA genes revealed that no significant shift on microbialcommunity composition took place upon biomass exposure to LCFA. DNA sequencing of predominantDGGE bands showed close phylogenetic affinity to ribotypes characteristic from specific b-oxidation bac-terial genera (Syntrophomonas and Clostridium), while a single predominant syntrophic archaeae wasrelated with the genus Methanosarcina. The hypothesis that biomass adaptation was fundamentally ofphysiological nature was tested using mathematical modelling, taking the IWA ADM1 as general model.New kinetics considering the relation between LCFA inhibitory substrate concentration and specific bio-mass content, as an approximation to the adsorption process, improved the model fitting and provided abetter insight on the physical nature of the LCFA inhibition process.

� 2009 Elsevier Ltd. All rights reserved.

1. Introduction

Lipid containing wastes are interesting substrates for biogasproduction because of their high methane yield potential. Lipidsare initially hydrolyzed to glycerol and long-chain fatty acids(LCFA), which are further converted by syntrophic acetogenic bac-teria to hydrogen (H2) and acetate (Ac), and finally to methane(CH4) by methanogenic archaea. The degradation of LCFA takesplace through the b-oxidation pathway, which has been reportedas the rate-limiting step of the whole anaerobic digestion process(Lalman and Bagley, 2002). LCFA are known to inhibit the metha-nogenic activity. The inhibitory effect was initially attributed topermanent toxicity resulting from cell damage and it is known toaffect both syntrophic acetogens and methanogens (Hwu et al.,

ll rights reserved.

ical Centre, Rambla Pompeuel.: +34 93 5796780; fax: +34

tats).

1998). Further studies have demonstrated that LCFA inhibition isreversible and that microorganisms, after a lag-phase, are able toefficiently methanise the accumulated LCFA (Pereira et al., 2004).Adsorption of LCFA onto the microbial surface has been suggestedas the mechanism of inhibition, affecting the transport of nutrientsinto the cell (Pereira et al., 2005).

Recent advances in molecular microbial ecology have broughtnew insights on the specific microorganisms that are involved inthe b-oxidation process. LCFA degrading bacteria have been foundto be closely related to the Syntrophomonadaceae and Clostridiaceaefamilies (Hatamoto et al., 2007; Sousa et al., 2007). These microor-ganisms are commonly proton-reducing acetogenic bacteria thatrequire the syntrophic interaction with H2-utilizing methanogensand acetoclastic methanogens (Sousa et al., 2007). Biomass adapta-tion to inhibitory levels of LCFA has recently been reported in sev-eral studies (Nielsen and Ahring, 2006; Cavaleiro et al., 2009;Palatsi et al., 2009). Currently, it is not clear whether this adapta-tion process is the result of a microbial population shift towardsthe enrichment of specific and better adapted LCFA-degraders

2244 J. Palatsi et al. / Bioresource Technology 101 (2010) 2243–2251

(population adaptation), or to the phenotypic adaptation of theexisting microorganisms towards high LCFA concentrations (phys-iological acclimatation).

Despite the fact that LCFA inhibition is well documented andhas a significant impact on the anaerobic digestion process, thisphenomenon has still not been included in IWA ADM1 referencemodel (Batstone et al., 2002). In other developed models, LCFAinhibition is mainly modelled as a non-competitive process onthe lipolytic, acetogenic or methanogenic activities (Angelidakiet al., 1999; Salminen et al., 2000; Lokshina et al., 2003). However,LCFA adsorption phenomena or the microbial aspects of the LCFAinhibition/adaptation process remain poorly characterized. Furthermodelling developments are required in order to relate the resultsfrom physiological activity tests and the characterization of micro-bial population dynamics throughout the whole LCFA inhibition/adaptation process.

The aim of the present study is to gain a deeper insight on theLCFA inhibition and adaptation process of the anaerobic consor-tium. Specific physiological activity rates and the microbial struc-ture composition in biomass samples obtained from reactorsexposed to LCFA pulses were compared. These samples were char-acterized by means of culture-independent molecular profiling ofdominant eubacterial and archaeobacterial populations, respec-tively. The obtained results were used in the implementation andtesting of a new LCFA inhibition kinetics expression, in the frame-work of the IWA ADM1 model (Batstone et al., 2002).

2. Methods

2.1. Analytical methods

Total solids (TS), volatile solids (VS), total Kjeldhal nitrogen(TKN), ammonia nitrogen ðNHþ4 -NÞ and pH were determinedaccording to Standard Methods (APHA, AWWA, WEF, 1995). Meth-ane content in the biogas (%CH4) and volatile fatty acids concentra-tion in the liquid media (VFA), corresponding to acetate (Ac),propionate (Pr), iso- and n-butyrate (Bu), iso- and n-valerate (Va)and hexanoate (Hex), were measured in a gas chromatograph fittedwith a flame ionization detection (GC-FID 20100, Shimatzu, Japan).Two different capillary columns, Porapak 60/80 Molsieve (6 ft3 mm) and ZEBRON Phase ZB-FFAP (30 m � 0.53 mm � 1.00 lm),were used for CH4 and VFA determination, respectively (Angelidakiet al., 2009).

2.2. Biomass and specific batch test

Samples from the outflow of semi-continuous thermophilic(55 �C) laboratory completely stirred reactors, fed with manureand exposed to two successive LCFA pulses (4 g L�1), were usedin subsequent anaerobic batch activity assays and in the molecularcharacterization of the microbial community structure. The LCFApulse was composed by a mixture of sodium oleate (C18:1), so-dium stearate (C18:0) and sodium palmitate (C16:0) in a ratio40:10:50 (w/w/w), respectively (analytical grade, BDH ChemicalsLtd., Poole England), since these are the main constituents in li-pid-rich wastewaters (Hwu et al., 1998). Manure was fed in theinfluent as the basic substrate at hydraulic retention time (HRT)of 20 days, and a corresponding organic loading rate (OLR) of1.0 g VS L�1 d�1. Fresh manure was diluted with distilled waterprior to its use, in order to decrease the ammonia level(1.41 ± 0.25 g TNK L�1; 0:92� 0:03 g NHþ4 -N L�1) and ensure thatthe pulse of LCFA was the only inhibitory cause throughout theexperiments. Samples were withdrawn from reactors at differentstages; before each LCFA pulse (samples I and III), when the pro-cess was clearly inhibited (samples II and IV), and when it recov-

ered and reached a new steady state (samples III and V). Thesampling program is shown in Table 1. The time between sampledbiomass I and III was 25 days, and between samples III and V was24 days. So, in all cases, more than one HRT had elapsed before itwas assumed that a new state was established. The concentrationof LCFA in the reactors at sampling times II and IV, were approxi-mately 4 g L�1, while LCFA were not detected at samples III andV. A detailed description on the experimental set-up and operationof the sampled reactors can be found in Palatsi et al. (2009).

Specific batch activity tests of non-inhibited (samples I, II and V)and LCFA-inhibited biomass (samples II and IV) were performed inanaerobic batch assays with specific substrates, according to Table1. Glass bottles (118 mL total volume) were inoculated with2.5 g VS L�1 from bioreactor sampled biomass, resuspended in basicanaerobic medium (Angelidaki et al., 2009), previously amendedwith 31 mM NaHCO3. A reducing solution of sodium sulfide(3.20 mM Na2SO3) was also added up to a final liquid total volumeof 50 mL and the pH was adjusted to neutrality. The flasks were stir-red and bubbled with N2 gas in order to remove O2 before sealingthem with rubber stoppers and aluminum crimps. In order to mea-sure the aceticlastic methanogenesis and acetogenetic activity rates,the bottles were supplemented with 20 and 10 mM of acetate (Ac)and butyrate (Bu), respectively, while the hydrogenotrophic metha-nogenesis was assayed by injecting 70 mL H2 and 40 mL CO2 in theheadspace (1 atm, 20 �C), as described by Angelidaki et al. (2009).Additional batches with inhibited and non-inhibited biomass wereincluded as controls, without the addition of any substrate, to deter-mine the methane production derived from the depletion of theLCFA adsorbed onto the biomass (for samples II and IV) and fromthe utilization of residual organic matter (for samples I, III and V).Activity tests were conducted in quadruplicate (3 vials for CH4 anal-ysis and 1 vial for VFA determination). Methane and VFA were mon-itored in the head space and in the liquid medium, respectively.Batch tests set-up and monitored variables are presented in Table 1.

The specific biomass activity rate was determined by linearregression on the initial slope of the accumulated methane produc-tion curve, and was expressed as mg CODCH4 g VS�1 d�1. For sub-strates that are not directly converted into methane, likebutyrate or LCFA, the methane production rate is only a valid mea-sure of syntrophic activity, when the aceticlastic and hydrogeno-trophic steps are not the rate limiting process (Dolfing andBloemen, 1985). Consequently, the maximum specific substrateutilization rate in the assays with butyrate was also calculatedfrom the steepest linear decline in substrate concentration(mg CODBu g VS�1 d�1), as described by Nielsen and Ahring(2006). In control vials with inhibited biomass (Control + LCFA inII and IV samples, according to Table 1), the LCFA maximum spe-cific utilization rate was estimated from the initial maximum slopeof Ac production (mg CODAc g VS�1 d�1), assuming that Ac was themain product from LCFA b-oxidation (Batstone et al., 2002).

2.3. Molecular analysis of microbial community

The effect of LCFA pulses on the anaerobic microbial communitycomposition of both eubacterial and archaeal domains was ana-lyzed at beginning and at the end of reactor operation (samples Iand V, according to Table 1). Reactor samples of 2 mL were fixatedin 1 mL of guanidine thyocyanate (4 M-Tris–Cl, pH 7.5: 0.1 M, auto-claved) and 0.5 mL of N-lauroyl sarcosine (10% N-LS autoclaved)and stored at �20 �C until further processing. The total DNA wasextracted by using the PowerSoil DNA isolation kit (MoBio Labora-tories Inc., USA), according to the instructions of the manufacturer.The V3–V5 variable regions of the eubacterial 16S rDNA gene wasamplified by the polymerase chain reaction (PCR) using the F341and R907 primers (Yu and Morrison, 2004). A nested approachwas applied to amplify archaeal 16S rDNA by using the primer

Table 1Summary of batch tests set-up and monitored variables in experimental assays.

Sample LCFA inhibition Days from LCFA pulse Added substrate (k) Initial substrate concentration in vials (kg COD m�3) Monitored variables (j)

I NO �1 H2/CO2a

AcBuControl

Sgh2(0) = 0.04/0.04/0.04Sac(0) = 1.49/1.50/1.31Sbu(0) = 1.76/1.67/1.54–

SgCH4

Sac, SgCH4

Sbu, Sac, SgCH4

Sac, SgCH4

III +24 (�1)V +48 (+23)

II YES +2 H2/CO2a (+LCFA)

Ac (+LCFA)Bu (+LCFA)Control (+LCFA)

Sgh2(0) = 0.04/0.04 (+Sfa(0) = 2.23/2.64)Sac(0) = 2.00/1.35 (+Sfa(0) = 2.23/2.64)Sbu(0) = 2.00/1.67 (+Sfa(0) = 2.23/2.64)(+Sfa(0) = 2.23/2.64)

Sac, SgCH4

Sac, SgCH4

Sbu, Sac, SgCH4

Sac, SgCH4

IV +27 (+3)

Note: Roman numbers indicate biomass samplings from reactors. LCFA pulses were introduced in reactors on day 0 and day 25. Days in parenthesis indicates time from thesecond LCFA pulse. Sfa(0) is LCFA remaining concentration from reactors pulse (4 g L�1) adsorbed onto biomass introduced in vials. Different values for initial substrateconcentration in vials, correspond to different sampled biomass or different batch test (I–V).

a Gas substrate units kmol m�3.

Table 2Process rates modifications used in different model approaches.

Batch Model approach Process rate (qj, kg COD m�3 d�1) Estimated parameters

I, III, V IWA ADM1 IWA ADM1 Xh2ð0Þ; Xacð0Þ; Xbu

II, IV (A1) IWA ADM1 IWA ADM1 Xfað0Þ; km;fa; K I

(A2) Inhibition Model Haldane kinetics for LCFA uptake rate q7 ¼ km;faSfa

KSþSfaþS2

faKI

XfaIpHIIN;limIh2 Xfað0Þ; km;fa; K I

Non-competitive term for Ac uptake rate q11 ¼ km;acSac

KSþSacXacIpHIIN;limINH3;Xac

KIKIþSfa

Non-competitive term for H2 uptake rate q12 ¼ km;h2Sh2

KSþSh2Xh2IpHIIN;lim

K IKIþSfa

(A3) Inhibition–Adsorption Model Replace KI by KIFA in q7, q11 and q12 with K IFA ¼ K 0I XfaSfa

Xfað0Þ; km;fa; K 0I

Nomenclature and units were maintained from IWA ADM1 (Batstone et al., 2002).

J. Palatsi et al. / Bioresource Technology 101 (2010) 2243–2251 2245

pairs ARCH0025F–RCH151R and F344–R915 for the first and thenested PCR reactions, respectively (Raskin et al., 1994). The for-ward primer used in the generation of the DGGE amplicons in-cluded a GC clamp at the 50 in order to stabilize the meltingbehaviours of the DNA fragments during DGGE. All PCR reactionswere performed in a Gradient Mastercycler (Eppendorff, Germany).

Approximately 300 ng of purified PCR product was loaded ontoa 8% (w/v) polyacrylamide gel (0.75 mm), with a denaturing chem-ical gradient ranging from 30% to 70% (100% denaturant stock solu-tion contained 7 M urea and 40% formamide). DGGE wasperformed in 1� TAE buffer (40 mM Tris, 20 mM sodium acetate,1 mM EDTA, pH 7.4) using a DGGE-4001 System (CBS Scientific,USA) at 100 V and 60 �C for 16 h. DGGE gels were stained for45 min in 1� TAE buffer containing SybrGold (Molecular Probes,USA) and then scanned under blue light by means of a blue con-verter plate and a transilluminator (GeneFlash Synoptics Ltd., USA).

Relevant DGGE bands were excised with a sterile filter tip,resuspended in 50 lL sterilized Milli-Q water, and stored at 4 �Covernight. These extracts were subsequently reamplified by PCRand sequenced. Sequencing was accomplished using the ABI prismBigDye Terminator v. 3.1 cycle sequencing kit (Perkin–ElmerApplied Biosystems, USA) and an ABI 3700 DNA sequencer(Perkin–Elmer Applied Biosystems, USA), according to instructionsof manufacturer. Sequences were edited using the BioEdit softwarepackage v. 7.0.9 (Ibis Biosciences, USA) and compared against theNCBI genomic database with the BLAST search alignment tool(NCBI, USA, http://www.blast.ncbi.nlm.nih.gov/Blast.cgi). Nucleo-tides sequences obtained in the present study have been depositedin the GenBank database under accession numbers GQ468297 toGQ468308.

2.4. Mathematical modelling and parameter estimation

Processes related to monitored variables (Table 1) were mod-elled with IWA ADM1 as basic model implemented in MatLab(The Mathworks, USA), applying the same structure, nomenclature

and units (Batstone et al., 2002). Data obtained from activity batchtest and molecular microbiology analysis, were used to estimateseveral unknown parameters and the initial biomass concentra-tions. The default values for kinetic parameters and stoichiometriccoefficients suggested by Batstone et al. (2002) for thermophilicoperation were adopted, with the following exceptions: (a) the va-lue of LCFA inhibition constant on hydrogenotrophic methanogen-esis (KI,h2fa), which is not given for thermophilic range in the ADM1model, was assumed to be the same as for mesophilic,KI,h2fa = 5 � 10�6 kg COD m�3; (b) the adopted value for the li-quid–gas mass transfer coefficient was kLa = 45 d�1; and (c) thepH was assumed to be constant, since a buffering solution wasadded to each vial and no significant pH change was detected. Inall simulations, the initial value for inorganic nitrogen wasSin(0) = 10�2 kmol N m�3. The initial specific substrate concentra-tion in each vial, used as model initial vector, are summarized inTable 1.

The time course of the variables monitored in vials with non-inhibited biomass (samples I, III and V), with H2/CO2, Ac and Buas substrates (Table 1), were used to estimate the initial concentra-tion of H2, Ac and Bu degrading microbial populations, Xi(0)(kg COD_X m�3), by a sequential estimation procedure (step-by-step, where the found values were then used as fixed parametersin next step), using ADM1 and its default biochemical parametersvalues (Batstone et al., 2002), as indicated in Table 2.

Different approaches were considered concerning the modellingof the inhibition phenomena observed on the activity tests withinhibited biomass (samples II and IV, according to Table 2). The firstassumption (A1) consisted on a direct application of the IWA ADM1Model using the suggested biochemical parameters (Batstone et al.,2002) and the calculated initial biomass content (Xh2(0), Xac(0), andXbu(0)), for assays I and III. This initial biomass content was consid-ered to be equal to subsequent sampled inhibited biomass, samplesII and IV, respectively, since the time delay between the sampling ofnon-inhibited and inhibited biomass was only 2–3 days (Table 1).With those assumptions, the initial amount of LCFA degrading

Table 3Substrate utilization rates of non-inhibited biomass (I, III and V).

Substrate Unit I III V

H2/CO2 mg CODCH4 g VS�1 d�1 91.1 ± 5.9 a 131.7 ± 6.6 b 147.2 ± 3.7 cAc mg CODCH4 g VS�1 d�1 127.7 ± 6.5 a 122.9 ± 8.2 a 135.0 ± 10.7 a

Bu mg CODCH4 g VS�1 d�1 183.4 ± 18.8 a 181.8 ± 2.6 a 183.9 ± 37.4 amg CODBu g VS�1 d�1 �263.8 �285.8 �230.9

Note: Different letters in rows indicate significant differences between rates (a = 0.05).

2246 J. Palatsi et al. / Bioresource Technology 101 (2010) 2243–2251

microorganisms, Xfa(0) (kg COD_X m�3), and the maximum LCFAuptake rate, km,fa (kg COD_S kg COD_X�1 d�1), were estimated bya multiple parameter optimization procedure, using the time evolu-tion data of all the monitored variables during activity tests ofinhibited biomass (samples II and IV), according to Table 2.

The second approach (A2), named as Inhibition Model, consid-ered the uptake of LCFA to be described by the Haldane’s inhibitionkinetics and both methanogenic processes (uptake of acetate andhydrogen) to be affected by a non-competitive term with acommon LCFA inhibition constant, KI (kg COD_SI m�3), as shownin Table 2. Such inhibition kinetics has already been proposed byother authors. Angelidaki et al. (1999), studying manure codiges-tion with glycerol trioleate or bentonite bound oil degradation,considered a non-competitive LCFA inhibition on the lipolitic, ace-togenic and methanogenic steps, and the Haldane inhibition kinet-ics on the b-oxidation process. Salminen et al. (2000) and Lokshinaet al. (2003), using solid slaughterhouse waste, considered a non-competitive inhibition kinetics due to LCFA, affecting acetogenesisand methanogenesis. With those assumptions, new initial valuesfor Xfa(0), km,fa and KI, were estimated by multiple parameter opti-mization (Table 2).

The last approach (A3), was named as Inhibition–AdsorptionModel, and included a simple mathematical expression for thedescription of the physical adsorption process of LCFA onto the bio-mass, as an inhibition mechanism. Adsorption is considered as arapid physico-chemical phenomenon, while desorption (degrada-tion) is a biologically mediated process by LCFA-degraders (Hwuet al., 1998). Pereira et al. (2004) proposed a modification of theHaldane equation for the LCFA inhibition process, which considersthe adsorbed substrate per VS unit, Sba ðMsubstrate M�1

biomassÞ, insteadof total substrate concentration (Sfa). Consequently, by adoptingthis concept, the proposed Inhibition–Adsorption Model assumesthe following hypothesis: (a) the inhibition of LCFA uptake processcan be expressed by the Haldane kinetics; (b) a non-competitivereversible inhibition term can be used on acetogenesis and meth-anogenesis; (c) in the previous inhibition processes, the inhibitoryconstant (KI) is replaced by a new inhibitory term, K IFA ¼ K 0I� Xfa=Sfa,proportional to the specific ratio between the LCFA degrading pop-ulation and the substrate (Xfa/Sfa), being higher (less inhibition)when this ratio value increases (Table 2).

The objective function was minimized, in the sequential param-eter estimation procedure, for each step or specific substrate k,according to the following equation,

fobjk ¼X

j

wkj

Xnkj

i¼1

ðy�kji � ykjiÞ2

!; ð1Þ

where y�kji represents the measured value of variable j, in vial k, attime i, and ykji is the corresponding simulated value. Variable j fromvials k has nkj measured values at successive different times i. Theweight factors, wkj, used in the objective function were defined asin the following equation,

wkj ¼ nkj max y�kji

� ��min y�kji

� �� �2� ��1

; ð2Þ

with max y�kji

� �and min y�kji

� �being the maximum and minimum

measured value of variable j in vial (step) k, respectively. The objec-tive function used in the multiparameter estimation procedure,with data-sets II and IV, was calculated according to Eq. (3), andthe optimization routine followed the downhill simplex methodas implemented in the MatLab package,

fobj ¼X

k

fobjk: ð3Þ

Model data fitting accuracy was measured by the coefficients ofdetermination R2 defined in the following equation:

R2kj ¼ 1�

Pnkji¼1 y�kji � ykji

� �2

Pnkji¼1 y�kji � �y�kji

� �2 ; ð4Þ

where �y�kji is the mean of nkj measured values of variable j fromvial k.

3. Results and discussion

3.1. Specific batch tests

The first set of analyzed batch tests were those with biomass ta-ken from the reactors, just before the application of LCFA pulses(samples I and III, in Table 1), and when the system had recoveredfrom a previous inhibition stage (sample V, in Table 1). Results ofactivity batch tests on specific substrates; H2/CO2, Ac and Bu,respectively, as model substrates for the main trophic groups, aresummarized in Table 3. Mean separation was performed on thecalculated rates by Multiple Range Test (MRT) with a significancelevel a = 0.05 (Sheskin, 2000).

A significant increase on the hydrogenotrophic methanogenicactivity rate was observed after subsequent inhibitory stages (Table3), in samples I–V (from 91.1 to 147.2 mg CODCH4 g VS�1 d�1), whilethe net acetoclastic methanogenic activity remained at a relativelysimilar level along time (127.7, 122.9 and 135.0 mg CODCH4 gVS�1 d�1, for samples I, III and V). These results are in agreementwith previous findings on suspended sludge and fixed bed reactorssubjected to LCFA inhibition, which concluded that hydrogeno-trophic methanogens appeared to be more resistant to LCFA inhibi-tion than acetoclastic methanogens (Templer et al., 2006).

Concerning the acetogenic activity, the n-butyrate (Bu) uptakerate remained fairly constant (263.8, 285.8 and 230.9 mg CODBu

g VS�1 d�1, respectively, for samples I, III and V) and no significantstatistical differences were found in terms of methane productionrate (in CODCH4 units, according to Table 3). Similarly, Nielsen andAhring (2006) found that the maximum substrate utilization ratefor Ac and Bu by biomass from thermophilic anaerobic reactors,fed with a mixture of cattle and pig manure and subjected to oleatepulses (2 g L�1), decreased or remained constant, while the metha-nogenic activity rate from H2/CO2, but also from formate and Ac,experienced an increase.

To analyze the inhibitory effect of LCFA pulses on specific activ-ities of representative trophic groups, a second set of batch tests

Table 4Substrate utilization rates of LCFA-inhibited biomass (II and IV).

Substrate Units II IV

H2/CO2 (+LCFA) mg CODCH4 g VS�1 d�1 67.6 ± 7.9 a 90.8 ± 2.7 bAc (+LCFA) mg CODCH4 g VS�1 d�1 44.6 ± 1.3 a 56.7 ± 4.4 a

Bu (+LCFA) mg CODCH4 g VS�1 d�1 183.9 ± 3.8 a 174.0 ± 15.4 amg CODBu g VS�1 d�1 �183.2 �161.8

Control (+LCFA) mg CODCH4 g VS�1 d�1 163.3 ± 8.7 a 218.8 ± 16.1 bmg CODAc g VS�1 d�1 104.9 153.6

Note: Different letters in rows indicate significant differences between rates(a = 0.05).

Fig. 1. DGGE profiles on eubacterial and archaeal 16S rDNA amplified from samplesI and V. A standard ladder (L) has been used at both gel ends in order to check theDNA migration homogeneity. Successfully excised and sequenced bands have beennamed with lower-case letters.

J. Palatsi et al. / Bioresource Technology 101 (2010) 2243–2251 2247

were run with biomass, sampled 2–3 days after each LCFA pulse,when biogas production in the reactor evidenced a clear inhibition(biomass samples II and IV, according to Table 1). Tests were per-formed with H2/CO2, Ac, and Bu as methanogenic and acetogenicmodel substrates, respectively. Samples II and IV had remainingLCFA adsorbed onto the biomass. Additionally, one set of vials wereincluded as controls, Control (+LCFA), incubated without any sub-strate supplementation in order to monitor the b-oxidation pro-cess. The specific activities of inhibited biomass, II and IV, aresummarized in Table 4.

In general, a clear reduction in all monitored metabolic activitieswas observed upon the application of each LCFA pulse (Table 4compared to Table 3). During batch activity tests on LCFA-inhibitedbiomass, the remaining LCFA content (from the reactor pulse andadsorbed onto the biomass) was completely consumed and themethane production reached a maximum plateau close to the ex-pected theoretical value. These results confirm that LCFA inhibitionis a reversible phenomenon, since neither syntrophic acetogenicnor methanogenic activities were irreversibly damaged, which isin accordance to what has previously been reported (Pereira et al.,2004). Yet, acetoclastic methanogenesis was the most affectedactivity by LCFA (44.6–56.7 mg CODCH4 g VS�1 d�1, compared to127.7–122.9 mg CODCH4 g VS�1 d�1 for the LCFA-inhibited anduninhibited biomass, as shown in Tables 4 and 3, respectively).Those vials exhibited not only lower methane production ratesbut also a longer lag-phase, compared to the activities before theLCFA pulse. The hydrogenotrophic methanogenesis was the meta-bolic activity affected the least by LCFA inhibitory pulses, with ratevalues up to 90.8 mg CODCH4 g VS�1 d�1 (Table 4), very similar tothe system hydrogenotrophic activity prior to the LCFA inhibitorypulse (91.1 mg CODCH4 g VS�1 d�1, Table 3).

The results of the present study are in agreement with thehypothesis of LCFA-induced transport limitation (Pereira et al.,2005). Those authors found that hydrogen, the smallest methano-genic substrate molecule, was the first to be transformed intomethane in LCFA-inhibited systems, in relation to other substratesof higher molecular weight, due to its higher diffusivity throughthe LCFA adsorbed layer.

It has also been described in the literature that methanogensare more susceptible to LCFA inhibition than acidogens (Lalmanand Bagley, 2002; Mykhaylovin et al., 2005), which is also in agree-ment with the lower differences in acetogenic activities detectedon Bu vials, before and after LCFA inhibition (I–II on Table 3 andIII–IV on Table 4).

In relation to the control vials, LCFA batch Control (+LCFA), aclear improvement on the b-oxidation process along time was ob-served (from 163.3 to 218.8 mg CODCH4 g VS�1 d�1 or from 104.9 to153.6 mg CODAc g VS�1 d�1 in terms of substrate production rate,for the tests II and IV, respectively). Mladenovska et al. (2003) de-scribed that the biomass of digested manure and lipids was moreactive and had higher initial rates of methane production thanthe biomass of only digested manure (not exposed to lipids). Theseresults were related to the importance of the interaction microor-

ganism–substrate–particle size and, in particular, to the effect oflipids on cell density and aggregation. Pereira et al. (2004) reportedan enhancement on the microbial activity upon depletion of ad-sorbed LCFA, by favouring specific degrading populations, whileNielsen and Ahring (2006) also reported an increasing oleate toler-ance (from 0.3 to 0.7 g L�1) in manure thermophilic systems ex-posed to oleate pulses. Different explanations for this behaviorwere hypothesized, like the induction of higher hydrolysis rates,an increase on biomass concentration or changes in the microbialcomposition.

The observed differential LCFA inhibition effect on distinct tro-phic groups might, in principle, be related to an enrichment of spe-cific populations involved on LCFA degradation process. Therefore,a shift in bacterial and archaeal communities cannot be excludedand was studied further by means of molecular biology techniquesand mathematical modelling tools, as described in the subsequentparagraphs.

3.2. Microbial community structure

DGGE molecular profiling of PCR amplified eubacterial andarchaeal 16S rDNA ribotypes was performed on biomass taken atthe beginning (sample I) and at the end (sample V) of reactor oper-ation (Table 1). Despite the fact that both sampling events wereseparated in time by more than 40 days (equivalent to two HRTintervals), and that the biomass suffered two inhibitory LCFApulses and subsequent recoveries stages during this period, nosignificant differences were observed in the microbial communitystructure of eubacterial and archaeal populations (Fig. 1). Up to 12DGGE bands were successfully excised, reamplified and sequenced.BLAST sequence comparison against NCBI genomic database

2248 J. Palatsi et al. / Bioresource Technology 101 (2010) 2243–2251

resulted in close matches with several uncultured ribotypes fromthe Clostridiaceae, Syntrophomonadaceaae, Bacillaceae and Synerg-ites, all families that belong to the Firmicutes eubacterial phylum(Fig. 2).

The Clostridiaceae appears to be one of the most representedbacterial families in the microbial community of anaerobic digest-ers. In our study, the DGGE bands f and g are related to unculturedbacteria previously found in different solid waste-thermophilicanaerobic bioreactors (95–97% of sequence homology), and to thetype strains of Clostridium thermocellum (93%) and Clostridiumstercolarium (95%), respectively, as the closest phylogenetically de-fined matches. The sequence from band a also clustered with theClostridiaceae family, but its poor homology (88%) with databasesequences indicates that it might belong to a yet undescribedtaxon.

The sequence from band b was relatively homologous (95%) toan uncultured bacterium from an anaerobic digester and, more

[GQ468297] DGGE band f[FJ599513] Clostridium thermocellum CTL-6

[AB428531] Uncultured bacterium from a thermophilic

[AM947536] Uncultured bacterium from a thermophilic

[GQ468298] DGGE band g[AJ310082] Clostridium stercorarium DSM8532T

[DQ887964] Uncultured bacterium from a therm

[AB428533] Uncultured bacterium from a thermop

[GQ468303] DGGE band a

[U20385] Bacillus infernus TH-23[GQ468301] DGGE band j

[AB438007] Uncultured bacterium from

[DQ661718] Uncultured bacterium fr

rfmuiretcabderutlucnU]043039MA[

[GQ468299] DGGE band h[AB428538] Uncultured bacterium from a thermophilic a

[FJ205855] Uncultured bacterium from a mesophilic ana

[AB114321] Uncultured bacterium from a thermophilic a

[FJ205846] Uncultured bacterium from a mesophilic anae[GQ468300] DGGE band i

[EF586053] Uncultured bacterium from a solid waste ana

[AM947554] Uncultured bacterium from a thermophilic an

[AB428539] Uncultured bacterium from a thermophilic an

[GQ468306] DGGE band d[AJ243189] Anaerobaculum mobile D

[EF559055] Uncultured bacterium from a

[EU639163] Uncultured bacterium from a

[AB274507] Uncultured bacterium from a [GQ468305] DGGE band c[EF586051] Uncultured bacterium from a solid waste anaero

[AM947543] Uncultured bacterium from a thermophilic ana

[AB274499] Uncultured bacterium from a mesophilic anaerobic s

[AB428536] Uncultured bacterium from a thermophilic anaerobic

[GQ468306] DGGE band b[FJ825442] Uncultured bacterium from a mesophilic ana

[DQ666176] Syntrophomonas wolfei subsp. saponavi

detsawdilosciboreanacilihpomrehtamorfmuiretcabderutlucnU]935749MA[

[GQ468307] DGGE band e

[EF559035] Uncultured bacterium from a thermophilic detsawdilosciboreana

retsegidciboreanacilihposemamorfmuiretcabderutlucnU]628502JF[

elpmasgnitsopmocamorfmuiretcabderutlucnU]000834BA[

[GQ468302] DGGE band k[AM947530] Uncultured bacterium from a

osamorfmuiretcabderutlucnU]130685FE[

[EU639311] Uncultured bacterium from a th

[AB428524] Uncultured bacterium from a the

64100

96

89

100

3435

100

100

100

271823

100

100

69100

97

9665

82

82

39

42

46

44

50

45

26

37

89

73

100

3129

0.02

[GQ468297] DGGE band f[FJ599513] Clostridium thermocellum CTL-6

[AB428531] Uncultured bacterium from a thermophilic

[AM947536] Uncultured bacterium from a thermophilic

[GQ468298] DGGE band g[AJ310082] Clostridium stercorarium DSM8532T

[DQ887964] Uncultured bacterium from a therm

[AB428533] Uncultured bacterium from a thermop

[GQ468303] DGGE band a

[U20385] Bacillus infernus TH-23[GQ468301] DGGE band j

[AB438007] Uncultured bacterium from

[DQ661718] Uncultured bacterium fr

rfmuiretcabderutlucnU]043039MA[

[GQ468299] DGGE band h[AB428538] Uncultured bacterium from a thermophilic a

[FJ205855] Uncultured bacterium from a mesophilic ana

[AB114321] Uncultured bacterium from a thermophilic a

[FJ205846] Uncultured bacterium from a mesophilic anae[GQ468300] DGGE band i

[EF586053] Uncultured bacterium from a solid waste ana

[AM947554] Uncultured bacterium from a thermophilic an

[AB428539] Uncultured bacterium from a thermophilic an

[GQ468306] DGGE band d[AJ243189] Anaerobaculum mobile D

[EF559055] Uncultured bacterium from a

[EU639163] Uncultured bacterium from a

[AB274507] Uncultured bacterium from a [GQ468305] DGGE band c[EF586051] Uncultured bacterium from a solid waste anaero

[AM947543] Uncultured bacterium from a thermophilic ana

[AB274499] Uncultured bacterium from a mesophilic anaerobic s

[AB428536] Uncultured bacterium from a thermophilic anaerobic

[GQ468306] DGGE band b[FJ825442] Uncultured bacterium from a mesophilic ana

[DQ666176] Syntrophomonas wolfei subsp. saponavi

detsawdilosciboreanacilihpomrehtamorfmuiretcabderutlucnU]935749MA[

[GQ468307] DGGE band e

[EF559035] Uncultured bacterium from a thermophilic detsawdilosciboreana

retsegidciboreanacilihposemamorfmuiretcabderutlucnU]628502JF[

elpmasgnitsopmocamorfmuiretcabderutlucnU]000834BA[

[GQ468302] DGGE band k[AM947530] Uncultured bacterium from a

osamorfmuiretcabderutlucnU]130685FE[

[EU639311] Uncultured bacterium from a th

[AB428524] Uncultured bacterium from a the

64100

96

89

100

3435

100

100

100

271823

100

100

69100

97

9665

82

82

39

42

46

44

50

45

26

37

89

73

100

3129

0.02

Fig. 2. Phylogenic tree on eubacterial16S rDNA from DGGE excised bands (Fig. 1) and fromgiven between box brackets). The tree was generated using the neigbor-joining algorithmbeside the nodes represent the percentage of branch support given by bootstrap analys

distantly (93%), to the type strain of Syntrophomonas wolfei subsp.saponavida. The Syntrophomonas genus has been describedpreviously as specific syntrophic LCFA degrading bacteria (Sousaet al., 2008). Band j sequence was identical to that of the type strainof Bacillus infernus, the only strictly anaerobic species in the genusBacillus (Boone et al., 1995). This halotolerant and thermophilicbacterium is characteristic from deep terrestrial subsurface areas.Yet, very similar uncultured ribotypes (98–99% sequencehomology) were obtained during the composting of hyperthermo-philically pre-treated cow dung and from a thermophilic anaerobicdigester of solid waste (Leven et al., 2007).

The sequence from band d was identical to a number of uncul-tured ribotypes obtained from solid waste anaerobic digesters, andclosely related to that of the species Anaerobacterium mobile (98%sequence homology). This is a novel anaerobic, thermophilic, andslightly halotolerant bacterium able to ferment organic acids andsome carbohydrates into acetate, hydrogen, and CO2 (Menes and

anaerobic solid waste digester (Unpublished; Sasaki et al.)

anaerobic solid waste digester (Goberna et al ., 2009)

ophilic solid waste anaerobic digester (Unpublished; Li et al.)

hilic anaerobic solid waste digester (Unpublished; Sasaki et al.)

a composting sample of cow dung (Unpublished; Yamada et al.)

om a thermophilic solid waste anaerobic digester (Leven et al., 2007)

).lateouG;dehsilbupnU(elpmasgnitsopmocamo

naerobic solid waste digester (Unpublished; Sasaki et al.)

erobic digester (Kröber et al., 2009)

naerobic municipal solid waste digester (Tang et al., 2004)

robic digester (Kröber et al., 2009)

erobic digester (Unpublished; Li, Bouchez, and Mazeas)

aerobic solid waste digester (Goberna et al ., 2009)

aerobic solid waste digester (Unpublished; Sasaki et al.)

SM13181

thermophilic anaerobic digester (Unpublished; Li, Bouchez, and Mazeas)

thermophilic microbial fuel cell (Wrighton et al., 2008)

mesophilic anaerobic solid waste digester (Sasaki et al., 2007)

bic digester (Unpublished; Li, Bouchez, and Mazeas)

erobic solid waste digester (Goberna et al., 2009)

olid waste digester (Sasaki et al., 2007)

solid waste digester (Unpublished; Sasaki et al.)

erobic digester (Unpublished; Podmirseg et al.)

da DSM4212

lateanreboG(retsegi ., 2009))saezaMdna,zehcuoB,iL;dehsilbupnU(retsegi

,.laterebörK( 2009)

).lateadamaY;dehsilbupnU(

thermophilic anaerobic solid waste digester (Goberna et al., 2009))saezaMdna,zehcuoB,iL;dehsilbupnU(retsegidetsawdil

ermophilic microbial fuel cell (Wrighton et al., 2008)

rmophilic anaerobic solid waste digester (Unpublished; Sasaki et al.)

References

anaerobic solid waste digester (Unpublished; Sasaki et al.)

anaerobic solid waste digester (Goberna et al ., 2009)

ophilic solid waste anaerobic digester (Unpublished; Li et al.)

hilic anaerobic solid waste digester (Unpublished; Sasaki et al.)

a composting sample of cow dung (Unpublished; Yamada et al.)

om a thermophilic solid waste anaerobic digester (Leven et al., 2007)

).lateouG;dehsilbupnU(elpmasgnitsopmocamo

naerobic solid waste digester (Unpublished; Sasaki et al.)

erobic digester (Kröber et al., 2009)

naerobic municipal solid waste digester (Tang et al., 2004)

robic digester (Kröber et al., 2009)

erobic digester (Unpublished; Li, Bouchez, and Mazeas)

aerobic solid waste digester (Goberna et al ., 2009)

aerobic solid waste digester (Unpublished; Sasaki et al.)

SM13181

thermophilic anaerobic digester (Unpublished; Li, Bouchez, and Mazeas)

thermophilic microbial fuel cell (Wrighton et al., 2008)

mesophilic anaerobic solid waste digester (Sasaki et al., 2007)

bic digester (Unpublished; Li, Bouchez, and Mazeas)

erobic solid waste digester (Goberna et al., 2009)

olid waste digester (Sasaki et al., 2007)

solid waste digester (Unpublished; Sasaki et al.)

erobic digester (Unpublished; Podmirseg et al.)

da DSM4212

lateanreboG(retsegi ., 2009))saezaMdna,zehcuoB,iL;dehsilbupnU(retsegi

,.laterebörK( 2009)

).lateadamaY;dehsilbupnU(

thermophilic anaerobic solid waste digester (Goberna et al., 2009))saezaMdna,zehcuoB,iL;dehsilbupnU(retsegidetsawdil

ermophilic microbial fuel cell (Wrighton et al., 2008)

rmophilic anaerobic solid waste digester (Unpublished; Sasaki et al.)

References

homologous sequences deposited at the GenBank database (accession numbers areand the Kimura 2-parameter correction, and was bootstrapped 500 times. Values

is.

Table 5aEstimated parameter values for non-inhibited batch tests data sets I, III and V.

Model approach Estimatedparameter

Results

I III V

IWA ADM1 Xh2(0) 5.89 � 10�4 5.08 � 10�4 2.33 � 10�3

Xac(0) 1.30 � 10�2 1.26 � 10�2 1.70 � 10�2

Xbu(0) 5.53 � 10�4 1.52 � 10�3 1.68 � 10�3

Units: Xi (kg COD m�3).

0.00

0.50

1.00

1.50

2.00

Sbu

(kg

CO

D/m

3)

Bu vials for samples I, III and V Data I ADM1_I R2=0.99Data III ADM1_III R2=0.98Data V ADM1_V R2=0.99

0.00

0.50

1.00

1.50

2.00

Sac

(kg

CO

D/m

3)

Data I ADM1_I R2=0.83Data III ADM1_III R2=0.90Data V ADM1_V R2=0.80

0.0E+00

1.0E-02

2.0E-02

3.0E-02

0 2 4 6 8 10 12 14

Sg C

H4

(km

ol/m

3)

days

Data I ADM1_I R2=0.96Data III ADM1_III R2=0.96Data V ADM1_V R2=0.95

a

c

b

Fig. 3. Experimental data (point markers) and IWA ADM1Model results (lines) foractivities of Bu for non-inhibited biomass I, III and V, in terms of Bu degradation (a),Ac production/degradation (b) and CH4 cumulative production (c). Coefficients ofdetermination (R2) for model fitting are indicated.

J. Palatsi et al. / Bioresource Technology 101 (2010) 2243–2251 2249

Muxi, 2002). No reference strains were found to be sufficiently re-lated to the sequences from bands h, i, k and e for its phylogeneticassignment, but they were highly homologous, or ever identical, toa number of uncultured ribotypes obtained predominantly fromthermophilic anaerobic reactors degrading organic solid wastes(Goberna et al., 2009; Kröber et al., 2009; Tang et al., 2004; Wrigh-ton et al., 2008). Interestingly, the number of coincident, or highlyrelated, ribotypes found in this work in relation to the previouslycited studies is remarkable (Fig. 2). These results suggest that theenvironmental conditions present in the thermophilic anaerobicdigestion of solid wastes promote de formation of relatively stablemicrobial consortia.

In relation to the archaeal domain, a single predominant bandwas observed in the DGGE profiles (band l). The associated se-quence was 97% homologous to that of the Methanosarcina thermo-phila type strain. This thermophilic archeon is a methanogen thathas been found in a wide variety of thermophilic anaerobic digest-ers treating organic wastes. Sequence homology of band l washigher in relation to another strain of the same species that wasenriched in a thermophilic anaerobic digester operated at highconcentration of volatile fatty acids (Hori et al., 2006). Mladenovs-ka et al. (2003) compared the digestion of cattle manure at meso-philic conditions to the digestion of a mixture of manure withglycerol trioleate (2%, w/w). Despite different reactor performanceno differences were found in the diversity of archaea, being thevast majority of the detected ribotypes phylogenetically close toMethanosarcina siliciae. Karakashev et al. (2005) studied the influ-ence of environmental conditions and feeding on methanogenicpopulations in a real scale biogas plants, reporting a dominanceof Methanosarcinaceae members on manure digesters. Kaparajuet al. (2009) also reported the predominance of Methanosarcinaceaeon the pilot plant (Kogens-Lyngby, Denmark), which was used assource of inoculum for semi-continuous reactors sampled in thepresent study (Palatsi et al., 2009). Hence, the origin of the inocu-lum, the daily manure feeding and the thermophilic regime mighthave exerted a strong influence on the enrichment of specificmethanogenic populations.

3.3. Mathematical modelling and parameter estimation

Data from batch activity assays were used to test the threemodel approaches (A1–A3) as summarized in Table 2. The mainaim was to determine whether the observed biomass adaptationprocess to LCFA can be explained by an increase of specific degrad-ing populations (Xi), and/or a modification of the adsorption–inhi-bition process, once a species composition shift has been excludedas the reason for the observed adaptation.

In order to estimate the initial biomass content of specific tro-phic groups, Xi(0), the experimental data from batch activity tests,with H2/CO2, Ac and Bu as substrates and not inhibited biomass(batch with samples I, III and V), was used in a direct implementa-tion of IWA ADM1 Model and a sequential parameter estimationprocedure (as described in Section 2). Estimates on the initial bio-mass content of specific trophic groups, Xi(0), are summarized inTable 5a. Goodness-of-fit coefficient R2 of modelled results rangedfrom 0.78 to 0.99 (data not shown). As an example, the simulationsand experimental data for Bu batch activity tests (which includedthe previously initial estimated concentrations for hydrogeno-trophic and acetoclastic methanogens), and the corresponding R2

coefficients, are depicted in Fig. 3. When the initial population con-centration, Xi(0) and the maximum uptake rate, km,i were simulta-neously estimated at each step, the obtained km,i values wererelatively close to those suggested by Batstone et al. (2002) andno significant differences in the coefficients of determination werefound. Moreover, at the tested initial substrate concentrations (inactivity assays, Sið0ÞoKsi), the sensitivity of the system to varia-

tions on the half saturation constants (Ksi) was extremely low, asexpected (Dochain and Vanrolleghem, 2001), and this constantwas not possible to be identified. For this reason, the sequentialparameter estimation of initial Xi values by adopting the suggestedbiochemical parameters by IWA ADM1 (Batstone et al., 2002), wasconsidered adequate. Due to technical difficulties on the measure-ment of the methane production in batch V, caused by anoperational problem on the GC-FID, model fitting in this particularbatch was based mainly on the VFA production–degradation pro-file (Fig. 3).

Based on the estimated initial biomass content of specific micro-organisms, Xi(0), an initial acetoclastic methanogenic populationstability can be outlined (Table 5a). However, the initial hydrogeno-trophic methanogenic population, Xh2(0), increased along sampling

Table 5bEstimated parameters values for inhibited batch tests data-sets II and IV.

Model approach Estimatedparameter

Results

II IV

(A1) IWA ADM1 Xfa 3.00 � 10�4 3.70 � 10�3

km;fa 22.37 22.37

(A2) Inhibition Model Xfa 2.40 � 10�3 4.45 � 10�2

km;fa 21.69 21.69K I 3.35 3.35

(A3) Inhibition–Adsorption Model Xfa 9.89 � 10�4 1.30 � 10�3

km;fa 124.33 124.33K I 2.37 � 103 2.37 � 103

Units: Xi (kg COD m�3); Km,fa (kg COD_S kg COD_X�1 d�1); KI and K 0I (kg COD m�3).

2250 J. Palatsi et al. / Bioresource Technology 101 (2010) 2243–2251

time, which could explain the observed improvement of the hydro-genotrophic activity (Table 3). From the microbial communityanalysis, it was not possible to differentiate between methanogenicpopulations, because the most abundant isolated archaeae wasaffiliated to the genus Methanosarcina (Fig. 2).

In the analysis of batch reactors with inhibited biomass(data-sets II and IV), the initial amount of hydrogenotrophicmethanogens Xh2(0), aceticlastic methanogens Xac(0), and butyrateacetogens Xbu(0), was assumed to be the same as in tests withnon-inhibited biomass (samples I and III), as explained in Section2 (Table 5a). The initial content of LCFA in batch tests II and IV(Sfa(0), 2.23 and 2.64 kg COD m�3) was identical to that from theprevious LCFA pulse in the reactor, adsorbed on the biomass. Asgeneral procedure, in each tested approach with inhibited batchtests data, a multiple parameter estimation (Xfa(0), km,fa and KI)was performed for batch test II and the obtained kinetic parametervalues were then used in the estimation of the initial LCFA degrad-ing population, Xfa(0), as the sole parameter optimized in batch testIV (Table 5b).

The first approach (A1) to estimate Xfa(0) and km,fa parameterswas the IWA ADM1 Model (Table 2). Fig. 4 shows, as example, theexperimental and predicted values for the inhibited sample IV.Although the predicted methane production curve and Sac or Sbu

evolution values are acceptable in some cases, it was not possibleto find an unique set of parameters (Xfa(0) and km,fa) able to fitall experimental data together, with sufficiently high coefficientsof determination (Fig. 4). Hence, the need to introduce modifica-tions in IWA ADM1 model, in order to express adequately the LCFAinhibition process is justified.

0.E+00

1.E-02

2.E-02

3.E-02

4.E-02

5.E-02

Sg C

H4

(km

ol/m

3)

H2/CO2+LCFA_IVA1 R2=0.30A2 R2=0.66A3 R2=0.90

0.0

0.5

1.0

1.5

2.0

2.5

3.0

0 2 4 6 8 10 12 14

Sac

(kg

CO

D/m

3)

days

A1 R2<0A2 R2<0A3 R2= 0.54

0.E+00

1.E-02

2.E-02

3.E-02

4.E-02

5.E-02

Sg C

H4

(km

ol/m

3)

Ac+LCFA_IV

A1 R2=0.91A2 R2=0.95A3 R2= 0.95

0.0

0.5

1.0

1.5

2.0

2.5

3.0

0 2 4 6 8 10 12 14

Sac

(kg

CO

D/m

3)

days

A1 R2=0.19A2 R2=0.96A3 R2= 0.85

Fig. 4. Experimental data (point markers) and different model assumptions (A1–A3) r(kmol CH4 m�3) cumulative production and substrate degradation (kg COD m�3) for inhibevery plot.

The second tested approach (A2), Inhibition Model, introducedthe Haldane inhibition kinetics for the b-oxidation and the revers-ible non-competitive inhibition kinetics for acetate or hydrogenmethanogenesis (Table 2), as previously reported (Angelidakiet al., 1999; Salminen et al., 2000; Lokshina et al., 2003). The esti-mated parameter values for batch test II are shown in Table 5b. Anincrease in the initial LCFA degrading population, Xfa(0), from2.40 � 10�3 to 4.45 � 10�2 kg COD_X m�3, in batch test IV wasdetected, maintaining a maximum degradation rate and inhibitionconstant of km,fa = 21.69 kg COD_S kg COD_X�1 d�1 and KI = 3.35 kgCOD m�3, respectively. Coefficients of determination and modelfitting for sample IV are shown in Fig. 4.

The last approach (A3), Inhibition–Adsorption Model, replacedthe constant inhibitory factor, KI, by a term KIFA proportional tothe ratio Xfa/Sfa (Table 2) to model the adsorption effect of LCFAon the cell walls. Estimated parameter values for test II were pre-sented in Table 5b. An increase in the initial LCFA degrading popu-lation Xfa(0), from 9.89 � 10�4 to 1.30 � 10�3 kg COD m�3, was alsodetected in sample IV (Table 5b), while initial KIFA(0) value re-mained around 1.15 kg COD m�3. An example of the obtained coef-ficients of determination and model fittings for sample IV areshown in Fig. 4. The best model fittings were obtained with theInhibition–Adsorption Model, which was able to reproduce not onlythe lower production rates when the system was inhibited but alsothe longer lag-phase during system inhibition. Although the ob-tained parameter set is probably not unique, these results couldbe considered as a first approach to express the importance ofthe LCFA/biomass ratio in the adsorption–inhibition process.

Modelling results suggest that adsorption plays an importantrole in the overall LCFA inhibition–adaptation process, and thatthere is a need to introduce modifications in IWA ADM1 modelwhen dealing with the degradation of lipids. Although the pro-posed Inhibition–Adsorption Model produces a satisfactory fittingof experimental results and provides a better representation ofthe physical nature of the overall LCFA inhibition process, addi-tional experimental data specifically designed to study biosorptionphenomena is needed to mathematically express the adsorption–inhibition process. It is important to notice that for all tested mod-elling approaches, an increase in the initial hydrogenotrophicmethanogens and LCFA degrading population occurred along time.The obtained batch experimental data and modelling results, to-gether with the apparent stability of the microbial communitystructure, might explain the observed LCFA adaptation process asthe result of a physiological acclimatation of existing populations

0.E+00

1.E-02

2.E-02

3.E-02

4.E-02

5.E-02

Sg C

H4

(km

ol/m

3)

Bu+LCFA_IV

A1 R2=0.81A2 R2=0.98A3 R2=0.97

0.0

0.5

1.0

1.5

2.0

2.5

3.0

0 2 4 6 8 10 12 14

Sbu

or S

ac (k

g C

OD

/m3)

days

Bu Data/*Ac DataA1 R2=0.55/<1A2 R2=0.89/0.90A3 R2=0.93/0.95A3 R2= 0.95

0.E+00

1.E-02

2.E-02

3.E-02

4.E-02

5.E-02

Sg C

H4

(km

ol/m

3)

Control (LCFA)_IV

A1 R2=0.95A2 R2=0.91A3 R2=0.97

0.0

0.5

1.0

1.5

2.0

2.5

3.0

0 2 4 6 8 10 12 14

Sac

(kg

CO

D/m

3)

days

A1 R2=0.12

A2 R2=0.42

A3 R2= 0.94

esults (lines) for activities to H2/CO2, Ac, Bu and LCFA (Controls) in terms of CH4

ited biomass IV. Coefficients of determination (R2) for model fitting are indicated in

J. Palatsi et al. / Bioresource Technology 101 (2010) 2243–2251 2251

or, at most, to the proliferation of specific, yet already existing,LCFA degrading bacteria and syntrophic methanogenic archaea.

4. Conclusions

Activity assays of anaerobic biomass exposed to successiveLCFA inhibitory pulses evidenced the recovery capacity of b-oxidiz-ing bacteria and syntrophic methanogens, while no significantmicrobial community shift occurred. A new LCFA inhibition kinet-ics was proposed within the IWA ADM1 model framework, whichresulted in better fits to the experimental results and provided anumerical expression of the process, in accordance to the adsorp-tive nature of the inhibition. The predicted increase in hydrogeno-trophic methanogens and LCFA degrading populations along time,together with the observed stability of the microbial community,indicate that the observed adaptation process is of physiologicalnature.

Acknowledgements

The authors would like to thank Miriam Guivernau (GIROTechnological Centre, Barcelona, Spain) for assistance in PCR-DGGEprofiling and ribotype sequencing. This work was supported by theSpanish Ministry of Science and Innovation (Projects ENE 2004-00724 and ENE 2007-65850) and from the Danish Energy Council(EFP-05 Journal no.: 33031-0029).

References

APHA, AWWA, WEF, 1995. Standard Methods for the Examination of Water andWastewater, 19th ed. American Public Health Association/American WorksAssociation/Water Environment Federation, Washington DC, USA.

Angelidaki, I., Ellegard, L., Ahring, B.K., 1999. A comprehensive model of anaerobicbioconversion of complex substrates to biogas. Biotechnol. Bioeng. 63 (3), 363–372.

Angelidaki, I., Alves, M., Bolzonella, D., Borzacconi, L., Campos, L., Guwy, A., Jenicek,P., Kalyuzhnui, S., van Lier, J., 2009. Defining the biomethane potential (BMP) ofsolid organic wastes and energy crops: a proposed protocol for batch assays.Water Sci. Technol. 59 (5), 927–934.

Batstone, D.J., Keller, J., Angelidaki, I., Kalyuzhnyi, S.V., Pavlostathis, S.G., Rozzi, A.,Sanders, W.T.M., Siegrist, H., Vavilin, V.A., 2002. Anaerobic Digestion Model No.1 (ADM1), IWA Task Group for Mathematical Modelling of Anaerobic DigestionProcesses. IWA Publishing, London.

Boone, D.R., Liu, Y., Zhao, Z.-J., Balkwill, D.L., Drake, G.R., Stevens, T.O., Aldrichl, H.C.,1995. Bacillis infernus sp. nov., an Fe(III)- and Mn(IV)-reducing anaerobe fromthe deep terrestrial surface. Int. J. Syst. Bacteriol. 45, 441–448.

Cavaleiro, A.J., Salvador, A.F., Alves, J.I., Alves, M., 2009. Continuous high rateanaerobic treatment of oleic acid based wastewater is possible after a stepfeeding start-up. Environ. Sci. Technol. 43 (8), 2931–2936.

Dochain, D., Vanrolleghem, P.A., 2001. Dynamical Modelling and Estimation inWastewater Treatment Processes. IWA Publishing, London.

Dolfing, J., Bloemen, W.G.B.M., 1985. Activity measurements as a tool tocharacterize the microbial composition of methanogenic environments. J.Microbiol. Method 4, 1–12.

Goberna, M., Insam, H., Franke-Whittle, I.H., 2009. Effect of biowaste sludgematuration on the diversity of thermophilic bacteria and archaeae in ananaerobic reactor. Appl. Environ. Microbiol. 75, 2566–2572.

Hatamoto, M., Imachi, H., Yashiro, Y., Ohashi, A., Harada, H., 2007. Diversity ofanaerobic microorganisms involved in LCFA degradation in methanogenicsludges revealed by RNA-based stable isotope probing. Appl. Environ. Microbiol.73 (13), 4119–4127.

Hori, T., Haruta, S., Ueno, Y., Ishii, M., Igarashi, Y., 2006. Dynamic transition ofmethanogenic population in response of the concentration of volatile fatty acidsin thermophilic anaerobic digester. Appl. Environ. Microbiol. 72, 1623–1630.

Hwu, S.H., Tseng, S.K., Yuan, C.Y., Kulik, Z., Lettinga, G., 1998. Biosorption of long-chain fatty acids in UASB treatment process. Water Res. 32 (5), 1571–1579.

Karakashev, D., Batstone, D.J., Angelidaki, I., 2005. Influence of environmentalconditions on methanogenic compositions in anaerobic biogas reactors. Appl.Environ. Microbiol. 7 (1), 331–338.

Kaparaju, P., Ellegard, L., Angelidaki, I., 2009. Optimization of biogas productionfrom manure through serial digestion: lab-scale and pilot-scale studies.Bioresour. Technol. 100, 701–709.

Kröber, M., Bekel, T., Diaz, N.N., Goesmann, A., Jaenicke, S., Krause, L., Miller, D.,Runte, K.J., Prisca, V., Pühler, A., Schlüter, A., 2009. Phylogenetic characterizationof biogas plant microbial community integrating clone library 16S-rDNAsequences and methagenome sequence data obtained by 454-pyrosequencing. J. Biotechnol. 142, 38–49.

Lalman, J.A., Bagley, D.M., 2002. Effect of C18 long chain fatty acids on glucose,butyrate and hydrogen degradation. Water Res. 36, 3307–3313.

Leven, L., Eriksson, A.R., Schnurer, A., 2007. Effect of process temperature onbacterial and archaeal communities in two methanogenic bioreactors treatingorganic household waste. FEMS Microbiol. Ecol. 59 (3), 683–693.

Lokshina, L.Y., Vavilin, V.A., Salminen, E., Rintala, J., 2003. Modeling of anaerobicdegradation of solid slaughterhouse waste. Inhibition effect of long-chain fattyacids or ammonia. Appl. Biochem. Biotechnol. 109 (1–3), 15–32.

Menes, R.J., Muxi, L., 2002. Anaerobaculum mibile sp. nov., a novel anaerobic,moderately thermophilic, peptide-fermenting bacterium that uses crotonate asan electron acceptor, and emended description of the genus Anaerobaculum.Int. J. Syst. Evol. Microbiol. 52, 157–164.

Mladenovska, Z., Dabrowski, S., Ahring, B.K., 2003. Anaerobic digestion of manureand mixture of manure with lipids: biogas reactor performance and microbialcommunity analysis. Water Sci. Technol. 48 (6), 271–278.

Mykhaylovin, O., Roy, J.M., Jing, J.A., 2005. Influence of C18 long chain fatty acids onbutyrate degradation by a mixed culture. J. Chem. Technol. Biotechnol. 80, 169–175.

Nielsen, H.B., Ahring, B.K., 2006. Responses of the biogas process to pulses of oleatein reactors treating mixtures of cattle and pig manure. Biotechnol. Bioeng. 95(1), 96–105.

Palatsi, J., Laureni, M., Andrés, M.V., Flotats, X., Nielsen, H.B., Angelidaki, I., 2009.Recovery strategies from long-chain fatty acids inhibition in anaerobicthermophilic digestion of manure. Bioresour. Technol. 100 (20), 4588–4596.

Pereira, M.A., Sousa, D.Z., Mota, M., Alves, M.M., 2004. Mineralization of LCFAassociated with anaerobic sludge: kinetics, enhancement of methanogenicactivity and effect of VFA. Biotechnol. Bioeng. 88 (4), 502–511.

Pereira, M.A., Pires, O.C., Mota, M., Alves, M.M., 2005. Anaerobic biodegradation ofoleic and palmitic acids: evidence of mass transfer limitations caused by longchain fatty acid accumulation onto the anaerobic sludge. Biotechnol. Bioeng. 92(1), 15–23.

Raskin, L., Stromley, J.M., Rittmann, B.E., Stahl, D.A., 1994. Group specific 16S rRNAhybridization probes to describe natural communities of methanogens. Appl.Environ. Microbiol. 60, 1232–1240.

Salminen, E., Rintala, J., Lokshina, L.Ya., Vavilin, V.A., 2000. Anaerobic batchdegradation of solid poultry slaughterhouse waste. Water Sci. Technol. 41 (3),33–41.

Sheskin, D., 2000. Handbook of Parametric and Nonparametric StatisticalProcedures, second ed. Chapman & Hall/CRC, Boca Raton, FL.

Sousa, D.Z., Pereira, A.A., Smidt, H., Stams, A.J.M., Alves, M.M., 2007. Molecularassessment of complex microbial communities degrading long chain fatty acidsin methanogenic bioreactors. FEMS Microbiol. Ecol. 60, 252–265.

Sousa, D.Z., Pereira, M.A., Alves, J.I., Smidt, H., Stams, A.J.M., 2008. Anaerobicmicrobial LCFA degradation in bioreactors. Water Sci. Technol. 57 (3), 439–444.

Tang, Y., Shigematsu, T., Morimura, S., Kida, K., 2004. The effects of micro-aerationon phylogenetic diversity of microorganisms in a thermophilic anaerobicmunicipal solid-waste digester. Water Res. 38, 2537–2550.

Templer, J., Lalman, J.A., Jing, N.J., Ndegwa, P.M., 2006. Influence of C18 long chainfatty acids on hydrogen metabolism. Biotechnol. Prog. 22, 199–207.

Wrighton, K.C., Agbo, P., Warnecke, F., Weber, K.A., Brodie, E.L., DeSantis, T.Z.,Hugenholtz, P., Andersen, G., Coates, J.D., 2008. A novel ecological role ofFirmicutes identified in thermophilic microbial. ISME J. 2, 1146–1156.

Yu, Z., Morrison, M., 2004. Comparisons of different hypervariable regions of rssgenes for use in fingerprinting of microbial communities by PCR-denaturinggradient gel electrophoresis. Appl. Environ. Microbiol. 70, 4800–4806.