Bioresource Technology 127 (2013) 181–187

Contents lists available at SciVerse ScienceDirect

Bioresource Technology

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

Characterization of aerobic granular sludge treating high strength agro-basedwastewater at different volumetric loadings

Norhayati Abdullah a,⇑, Ali Yuzir b,⇑, Thomas P. Curtis c, Adibah Yahya a, Zaini Ujang d

a Department of Industrial Biotechnology, Faculty of Biosciences and Bioengineering, Universiti Teknologi Malaysia, 81310 Johor Bahru, Malaysiab Faculty of Civil Engineering, Universiti Teknologi Malaysia, 81310 Johor Bahru, Malaysiac Department of Civil Engineering, Newcastle University, Newcastle upon Tyne, NE1 7RU, United Kingdomd Institute of Environment and Water Resource Management (IPASA), Universiti Teknologi Malaysia, 81310 Johor Bahru, Malaysia

h i g h l i g h t s

" Fractal dimension averaged at 1.90 indicating good compactness of granules." Significant microbial evolutionary shift was observed during aerobic granulation." Raup–Crick indices decreased upon formation of mature aerobic granular sludge.

a r t i c l e i n f o

Article history:Received 5 May 2012Received in revised form 11 September 2012Accepted 15 September 2012Available online 25 September 2012

Keywords:Aerobic granular sludgePalm oil mill effluent (POME)Denaturing gradient gel electrophoresis(DGGE)Sequencing batch reactorMicrobial community

0960-8524/$ - see front matter � 2012 Elsevier Ltd. Ahttp://dx.doi.org/10.1016/j.biortech.2012.09.047

⇑ Corresponding authors. Current address: Departmogy, Faculty of Biosciences and Bioengineering, Un81310 Johor Bahru, Malaysia. Tel.: +60 137036730, +6

E-mail addresses: norhayatiabdullah@fbb.utm.myutm.my (A. Yuzir).

a b s t r a c t

Understanding the relationship between microbial community and mechanism of aerobic granulationcould enable wider applications of granules for high-strength wastewater treatment. The majority ofgranulation studies principally determine the engineering aspects of granules formation with littleemphasis on the microbial diversity. In this study, three identical reactors namely R1, R2 and R3 wereoperated using POME at volumetric loadings of 1.5, 2.5 and 3.5 kg COD m�3 d�1, respectively. Aerationwas provided at a volumetric flow rate of 2.5 cm s�1. Aerobic granules were successfully developed inR2 and R3 while bioflocs dominated R1 until the end of experiments. Fractal dimension (Df) averagedat 1.90 suggesting good compactness of granules. The PCR–DGGE results indicated microbial evolutionaryshift throughout granulation despite different operating OLRs based on decreased Raup and Crick simi-larity indices upon mature granule formation. The characteristics of aerobic granules treating highstrength agro-based wastewater are determined at different volumetric loadings.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction

Palm oil production has gained significant attention in recentyears due to its many competitive advantages over other compet-ing oils i.e. olive, vegetable and sunflower oils, for having low costproduction, high yield and being free from trans-fatty acids. Theutilization of palm oil has also increased rapidly owing to its multi-ple uses in both food and non-food industries contributing togreater demand and higher prices for palm oil production. The glo-bal aspiration to substitute fossil fuel with renewable fuel has gi-ven rise to increased demand for palm oil which is one of thesources for biofuel. The risks of pollution generated from the indus-

ll rights reserved.

ent of Industrial Biotechnol-iversiti Teknologi Malaysia,0 75532711.

(N. Abdullah), muhdaliyuzir@

try have been escalating following the rapid expansion of palm oilindustry worldwide. The production of palm oil generates a largeamount of solid and liquid wastes in the form of empty fruit bunch(EFB) and palm oil mill effluent (POME), respectively. Malaysia’spalm oil industry produced almost 80 million dry tonnes of solidbiomass per annum (Agensi Inovasi Malaysia, 2011). This volumeis projected to increase to 85–110 million dry tonnes by 2020. Sim-ilarly, the current POME volumes are expected to increase from 60million tonnes to 70–110 million tonnes by 2020. The untreatedPOME is to comply with legislation limits of BOD5 of 20 mg L�1

for Standard A as outlined in the Fifth Schedule Paragraph 11(1)(a) Environmental Quality (Industrial Effluents) Regulations 2009(Federal Subsidiary Legislation, 1974). The new regulations alsooutlined the effluent discharge standard to comply with colordischarge of 100 ADMI. Therefore, color removal is fast becomingan important research parameter in relation to agro-basedindustrial wastewater treatments. Additionally, in recent years,the significance of technological improvements in handling of res-

Peristaltic pumps controlled by timers

Flow meter

Gas out Air

compressor

182 N. Abdullah et al. / Bioresource Technology 127 (2013) 181–187

idues from palm oil production compared to conventional prac-tices are also being increasingly addressed through application oflife cycle assessment (LCA) tools (Hansen et al., 2012).

Ponding systems which are commonly used for POME treatmenthave certain drawbacks including: methane emissions, long reten-tion times, large area requirements and excessive sludge accumula-tion (Chan et al., 2010). Novel POME treatment methods have beenproposed including anaerobic treatment methods and high-rateanaerobic digesters (Poh and Chong, 2009). Conventional anaerobicdigesters require large reactors and long retention times to ensuresatisfactory digestion of the treated effluent. High-rate anaerobicbioreactors have been used to overcome the problem of long reten-tion time when treating POME (Zinatizadeh et al., 2007).

POME in its untreated form is classified as a high strength agro-based wastewater with COD and BOD concentrations ranging from50,000 to 90,000 mg L�1. Recently, a respirometric study on POMEfound the Activated Sludge Model (ASM) heterotrophic yield coef-ficients and some COD fractionations of POME which could be usedas basis for design and optimization of a POME treatment process(Damayanti et al., 2010). The typical characteristics of POME andASM model coefficients are given in Table 1.

Aerobic granular sludge is a novel, compact consortium of self-immobilized bacteria with high-rate biological wastewater treat-ability. Aerobic granular sludge characteristically contains higherbiomass concentrations within the same reactor than floccularsludge systems (Beun et al., 1999). Other main advantages of theaerobic granular sludge are high biomass retention in the reactor,good settling properties and the capacity to withstand high OLR,which all contribute to the very small footprint of this technologyin comparison to conventional activated sludge based systems.Aerobic granulation has been observed in an SBR fed with variousorganic substrates including industrial wastewaters and landfillleachates (Cassidy et al., 2005; Kishida et al., 2009). To date, aero-bic granular sludge has been successfully developed using POME atOLR 2.5 kg COD m�3 d�1 (Abdullah et al., 2011). Within the labora-tory scale granulation SBR, numerous operational parameters maybe manipulated to actively select for stable aerobic granular sludgeformation including settling time, aeration intensity, feeding re-gime, substrate composition and organic loading rate. Relativelylittle is known about the microbial ecology of granular sludge sys-tem and the little we do know is derived from the systems fed withsynthetic wastewater. Thus, this study sought to investigate the

Table 1Characteristics of POME and ASM estimated model parameters.

Parametersa POME values(this study)

POMErangec

ASM heterotrophicyield coefficientsd

ASM modelcoefficient

Values

Chemical oxygendemand (COD)

69,500 15,000–100,000 Total COD 45,500

Biochemical oxygendemand (BOD)b

25,000 10,300–44,000 Ss 50

pH 3.8 3.4–5.2 Si 16,600Total solid 55,000 11,500–79,000 Xs 25,500Suspended solid 33,600 5,000–54,000 Xi 2800Volatile suspended

solid24,000 9,000–72,000 YH 0.44

Total nitrogen 800 80–1,400 Cell COD 14,100Ammoniacal

nitrogen45 4–80 lA 0.76

Total phosphorus 38 – lH 0.78

PO3�4

6 – Ks 100

Oil and grease Not measured 150–18,000 bH 0.33

a All parameters unit in mgL-1 except pHb Sample is incubated for 3 days at 30�Cc Source: Ujang et al., 2010d Adapted from Damayanti et al., 2011

relationship between the microbial community structure and aer-obic granular sludge formation using real wastewater such asPOME.

2. Methods

2.1. Reactor set-up

The schematic design of the reactor setup is given in Fig. 1.Three identical reactors with internal diameter of 50 mm andeffective height to diameter (H/D) ratio of 17 were setup with sev-eral modifications to accommodate a working volume of 3 L. Thereactor was constructed using Borosilicate glass due to its chemicalresistance, inert characteristics and transparency, allowing visualmonitoring of the mixed liquor and reactor contents. Each reactorwas equipped with an internal down-comer tube located at thebottom of the reactor for influent feeding, an outlet port placedat mid-height of the reactor yielding a volumetric exchange rate(VER) of 50% and two sampling ports for mixed liquor and aerobicgranular sludge samplings. Aeration was provided inside eachreactor by means of air bubble diffusers at a volumetric flow rateof 3 L h�1, which is equivalent to a superficial air velocity of2.5 cm s�1. The reactors were operated at room temperature(27 �C) in successive cycles of 3 h, which comprised of influentfeeding (5 min), aeration (110 min), anaerobic reactions (45 min),settling (15 min) and finally effluent withdrawal (2 min).

2.2. Wastewater and seed sludge preparation

Both POME seed sludge and raw POME were kept at 4 �C to min-imize degradation. Prior to feeding, the raw POME was centrifuged

DO

Effluent

Influent

Bubble diffuser

pH

Sampling ports

Effluent tank

Fig. 1. Schematic diagram of operational reactor setup.

Table 2Lists of all the primers used for the PCR reaction, classified according to theoligonucleotide probe database.

Primer Sequence (from 50 end to 30

end)Annealingsiteb

Reference

Vr ATTACCGCGGCTGCTGG 518–534 Muyzeret al.(1993)

Vf CCTACGGGAGGCAGCAG 341–357 Muyzer et al.(1993)

CTO654r CTAGCYTTGTAGTTTCAAACGC 654–674 Kowalchuk et al.(1997)

CTO189fa GAGRAAAGYAGGGGATCG 189–207 Kowalchuk et al.(1997)

a Added GC-clamp (CGCCCGCCGCGCGCGGCGGGGGCGGGGGCACGGGGGG) to theforward primers.

b The position on the 16S rRNA gene is represented by comparison to the Esch-erichia coli 16S rRNA gene sequence.

N. Abdullah et al. / Bioresource Technology 127 (2013) 181–187 183

at 15,000 rpm for 40 min to eliminate suspended solids and debriswhich can cause clogging to the influent tubes. A suitable amountof tap water was added to the POME according to the desired or-ganic loading rate (OLR). Nutrients and trace elements were addedin order to provide a balanced feed to the reactors (i.e. COD:N:Pratio of 100:5:1). Prior to feeding the pH was adjusted to a level be-tween 6.5 and 7.0 using 2 M NaOH resulting in alkalinity of mixedliquor in the reactor of between 1000 and 2000 mg L�1 as CaCO3.The seed sludge was sieved with a mesh of 1.0 mm twice to re-move large debris before inoculation. The reactors were inoculatedwith 500 mL of seed sludge resulting in MLSS concentration of3000 mg L�1 in each reactor.

2.3. Experimental procedures

Three reactors namely R1, R2 and R3 were operated at OLRs of1.5, 2.5 and 3.5 kg COD m�3 d�1, respectively. Supernatant liquorand aerobic granular sludge samples were taken from each reactorfor analysis, accordingly. In addition, the formation of aerobic gran-ular sludge ranging from seed sludge, bioflocs and mature granuleswas closely monitored under scanning electron microscopy (SEM)and stereo microscopic examinations. Parameters such as MLSS,MLVSS, COD, NH3, SVI and color were carried out as previously de-scribed in Abdullah et al. (2011). COD was analyzed by using color-imetric method using HACH DR 4000 Spectrophotometer. MLSScontent was determined by oven drying of sample at 104 �C,whereas MLVSS was determined by ashing the dry sample in a550 �C muffled furnace for 15 min. Dissolved oxygen (DO) andpH were measured by using a pH/DO meter (Orion 4-Star BenchtoppH/DO Meter). The morphological and structural observations ofgranular sludge were conducted by using a stereo microscopeequipped with digital image analyzer (PAX-IT� v6, ARC PAX-CAM). The microstructure and microbial distribution within thegranule were observed with scanning electron microscope(FESEM-Zeiss Supra 35 VPFESEM). Prior to gold sputter coating,the granules were left to dry at room temperature (BioRad PolaronDivisions SM Coating System).

2.4. Microbial structure analysis by PCR–DGGE

2.4.1. Nucleic acid extractionDNA was isolated from biomass after being fixed with 50% (v/v)

ethanol using the FastDNA�SPIN kit for soil according to manufac-turer instruction (MP Biomedicals, Santa Ana, CA).

2.4.2. Polymerase chain reaction (PCR)The extracted DNA was used as the template for PCR amplifica-

tion based on methods described by Pholchan et al. (2010). The to-tal eubacterial population was analyzed using PCR amplificationwith primers Vr and Vf and the betaproteobacterial ammonia oxi-dizing bacteria (AOB) community was analyzed using AOB specificprimers; CTO654r and CTO189f as shown in Table 2. The 16S rRNAgene fragments were amplified in a PCR reaction mixture using aThermo Hybrid PX2 Thermal cycles. Once the PCR was completed,the templates were stored at �20 �C.

2.4.3. Denaturing gradient gel electrophoresis (DGGE)The generated PCR products were analyzed using the general

bacterial primers 2/3 (Muyzer et al., 1993) by denaturing gradientgel electrophoresis (DGGE). In addition, another gel was run toanalyze the PCR products amplified using beta-proteobacterialAOB specific primers CTO189/654 (Kowalchuk et al., 1997).

The PCR products were loaded onto 10% (w/v) polyacrylamidegels prepared in 1 � TAE buffer (40 mM Tris–acetate, 1 mM EDTA,pH 8.3, Eppendorf Scientific Inc., USA) with a denaturing gradientranging from 30% to 55% (100% is 7 M urea and 40% (v/v) formamide).

The denaturing gradient used in the present study is an optimal gra-dient used for analysis of PCR products amplified using the generalbacterial primers used in the experiments. Polymerization was en-hanced with 0.1% v/v N,N,N0,N0-Tetramethylethylenediamine(TEMED) (Sigma–Aldrich, UK) and Ammonium Persulfate, respec-tively. Wells were loaded accordingly with 11 lL of PCR productsand loading buffer (0.25% v/v bromophenol blue, 0.25% v/v xylenecyanol, 30% v/v glycerol in water).

The reference marker of 11 lL was obtained from a series ofclones selected from the activated sludge sample was loaded intothe respective well prior to loading of PCR products and loadingbuffer. These markers dyes can be used to determine when to stopa DGGE run. When the buffers disappear off the bottom of the gel,the gel run can be stopped. The DGGE was performed with the D-Gene System (Bio-Rad, UK) running at constant voltage of 200 V for900 V h and a temperature of 65 �C for approximately 4 h. Theresulting gels were stained for 30 min in 1 � TAE containing CYBRgreen I (Sigma, Poole, UK; diluted 1/10 000 in 1 � TAE). DGGEbands were visualized under UV transilluminator with the pro-gram Quantity One (Bio-Rad) and the banding patterns were ana-lyzed using the BioNumerics� Version 3.5 (Applied Maths BVBA,Belgium).

2.4.4. Multivariate analysis of the DGGE gel (BioNumerics� Version 3.5Applied Maths BVBA)

The DGGE gel was normalized according to the reference mar-ker patterns to enable multivariate analysis on the gel. The bandingpattern was converted as a binary matrix corresponding to thepresence/absence of each band, further defined as species and sub-sequently used for the multivariate analysis. The similarities of thebinary banding matrix were analyzed using Raup and Crick simi-larity index. Cluster analysis, non-metric multi-dimensional scal-ing and analysis of similarity were performed using PASTSoftware (Hammer et al., 2001).

2.4.5. Basic Local Alignment Search Tool – nucleotide (BLASTn) andsequencing

All sequences obtained in this work were aligned and analyzedwith Chromas� Lite Software Informer 2.0 (Technelysium Pty Ltd.,1998–2004). Sequences were compared to the GenBank databaseusing the Basic Local Alignment Search Tool–nucleotide (BLASTn)algorithm (Benson et al., 2008) via www.ncbi.nlm.nih.gov/Blast.

2.5. Fractal dimension (Df) analysis

The Df of the aerobic granular sludge was determined usingimage analysis. A Zeiss Axio Imager Z2 microscope with Apotomeequipped with AxioCam MRc (color) and AxioCam MRm

184 N. Abdullah et al. / Bioresource Technology 127 (2013) 181–187

(monochrome) cameras connected to a computer via a grabbingboard was used for the image analysis. A drop of mixed liquorwas carefully deposited and covered with a cover slip. No stainingor fixation was required. The slide was systematically examinedusing �40 lens and a series of images acquired. The pixel size cal-ibration was done with a stage micrometer. The image was thenanalyzed by ImageJ Software (Collins, 2007). The gray-level imagesare automatically segmented resulting in a binary image. AnEuclidean distance map (EDM) of images is generated (Mu andYu, 2006). Mu and Yu (2006) explained that in the EDM image,the gray-level b associated to each pixel is related to its distanceto the nearest border. The gray-levels distribution b(S) then givesthe number of pixels at each distance S. The perimeter P(S) is cal-culated by dividing the number of pixels having a gray-level largeror equal to S by S:

PðSÞ ¼Ps

i¼1bðiÞs

ð1Þ

Therefore, if k is the slope of log P versus log S, the Df can be cal-culated with the following equation (Mu and Yu, 2006):

Df ¼ 2� k ð2Þ

The fractal dimension analysis is commonly associated with thecompactness and stability of aerobic granular sludge. The higherthe value of a fractal dimension, more stable and compact granulestructure will be achieved.

3. Results and discussion

3.1. Characteristics of aerobic granular sludge

Compact structured aerobic granular sludge was not detected inlow organic loading of 1.5 kg COD m�3 d�1 in R1. Instead, the bio-mass content in R1 was dominated by bioflocs towards the endof the experiment on day-60. At a relatively low operating OLR of1.5 kg COD m�3 d�1, it was found that most of the seed sludgewas washed out of the reactor within 24 h of startup. The MLSSvalues fluctuated between low of 0.5 g L�1 and high of 2.2 g L�1

throughout the experiments in R1 indicating an average qualityof biomass accumulation in the reactor. The excessive eliminationof poor settling activated sludge during effluent withdrawal in R1resulted in bad settling of activated sludge and an almost completewashout of biomass from the SBR. Poor reactor stabilization wasobserved during the initial start-up and throughout the experi-ments course in R1. These observations are consistent with previ-ous findings by Tay et al. (2004) in an SBR treating a simplesynthetic wastewater which found that the reactor experiencedbiomass washed out upon reactor startup at similarly low OLR of1.0 kg COD m�3 d�1 with MLVSS value averaged at only 5.1 g L�1.

An early study on physical conditioning of activated sludge byParker et al. (1971), suggested that a filamentous network is postu-lated to provide a backbone for the buildup of bioflocs or bioparti-cles in an activated sludge system. The PAX-IT� images revealedthat bioflocs cultivated in R1 had a fluffy and extremely loose mor-phology indicating similarities to those of the original seed sludge.Previously, it has been reported that low organic loading rate couldfavor filamentous growth. Evident filamentous filaments werevisually observed during PAX-IT� examinations protruding the sur-face of the bioflocs in R1. This feature is closely associated withprevious report by Martins et al. (2003) who hypothesized thatat low substrate concentration filamentous bacterial structuresgive easier access to the substrate on the outside of the flocs andthereby proliferate.

However, domain R2 and R3 were operated at OLRs of 2.5 and3.5 kg COD m�3 d�1, respectively. Good settling sludge was

observed despite the biomass being washed out during the start-up. The MLSS concentrations were maintained at above 2.0 g L�1

in both reactors as the biomass became acclimatized to the substrateand operating conditions. R2 and R3 exhibited good settling sludge;good settling was highly correlated with good reactor start-up per-formances. The suspended solids (SS) with good settling ability weresuccessfully retained in the reactor which subsequently improvedthe MLSS concentrations in the SBR. After 17 days of operation, light-er biomass fractions were eliminated and small aerobic granularsludge started to appear in both R2 and R3. The amount of granulesin both reactors subsequently increased over time and dominatedthe biomass content by day-60. The MLSS concentrations attainedwere at 7.6 g L�1 and 6.5 g L�1 in R2 and R3, respectively. The in-creased in reactor biomass concentrations indicated good biomassaccumulation which substantially promoted good maturation ofaerobic granular sludge in the SBR system.

Microscopic examinations showed that the seed sludge evolvedfrom seed suspended activated sludge to bioflocs before growinginto mature granules. The size of granules increased from 2.0 mmto 4.0 mm in R2 while R3 exhibited a slightly larger granular sludgesize of up to 5.0 mm. Microscopic examinations also showed thatgranules cultivated in R2 at an OLR of 2.5 kg COD m�3 d�1 werespherical-shaped, firm and compact with a clear outline boundary.Porous, irregular and diffused granular structures were observed inR3 which was operated at an OLR of 3.5 kg COD m�3 d�1.

SEM examination was used to characterize the evolution inmicrostructure of aerobic granular sludge during this study. Theseed sludge had a comparatively loose, fluffy and irregular struc-ture. The boundaries of primary particles were not clearly definedand the arrangement of bacteria is unclear. The microstructure ofgranules altered over time indicating rich microbial populationaccommodating the granule surface. The aerobic granular sludgeproduced also dominated by aerobic bacteria which proof thatthe method used for granule formation has successfully enrichedaerobic bacteria for the treatment of POME. Dense mixture of bac-teria on aerobic granular sludge surface is due to the occurrence ofthe bacteria in large morphological variety when POME is used assubstrate. The consortium of bacteria is embedded in a thick EPSmatrix as single cells or as dense architecture of microbial arrange-ment contributing to the dense surface of granule. EPS matrixformed part of the microbial mechanism of a successful aerobicgranular sludge formation. The EPS molecules form buffering layerfor microbial cells against the harsh external environment and pro-vides carbon and energy source during bacterial starvation phases.

The granules were made up of various bacterial microcolonies toform large and compact granular aggregates. The reactor conditionsin R2 and R3 seemed to favor the growth and formation of stableaerobic granules with mature granules dominating the reactor con-tent at the end of the experiment. It appears that the combination ofa short settling time and sufficient OLR could presumably resultedin the transition of activated sludge from a loose, fluffy structure toa compact, and well-defined aerobic granular sludge.

Settling ability and strength are important characteristics thatenable granules to be successfully retained in the reactor. Thesecharacteristics generally improved with increased OLR. The aver-age SVI values for the granules improved from 41.9 mL g�1 SS at1.5 kg COD m�3 d�1 to 19.9 mL g�1 SS at 3.5 kg COD m�3 d�1 (seedsludge SVI of 203 mL g�1 SS). A higher OLR contributed to a muchlower SVI with better settling granular sludge. These results areconsistent with observations made by Thanh et al. (2009) insequencing batch airlift reactors (SBARs) treating synthetic waste-water, who found that the SVI decreased from 243 mL g�1 SS to24.4 mL g�1 SS at OLR 2.5 kg COD m�3 d�1. Tay et al. (2004) furthersuggested that stable and compact granules with high specificgravity and high strength were developed under OLR of4 kg COD m�3 d�1.

Fig. 2. Raup and Crick similarity index calculated from the DGGE profiles. Theacclimation and granulation correspond to the acclimation and granulationprocesses, respectively; and maturation refers to mature aerobic granular sludgesampled at steady state. The labeled data points refer to (a) Pure seed sludgeoriginated from POME facultative pond; (b) mature granules obtained at OLR1.5 kg COD m�3 d�1; (c) mature granules obtained at OLR 3.0 kg COD m�3 d�1; (d)mature granules obtained at OLR 2.5 kg COD m�3 d�1.

N. Abdullah et al. / Bioresource Technology 127 (2013) 181–187 185

3.2. Size and fractal dimension of aerobic granular sludgeinterpretation

The fractal theory developed by Mandelbrot (2006) gives a sys-tematic approach to the characterization of surface morphology ofaerobic granular sludge. The theoretical values of Df, varying from 1to 3 (Mu and Yu, 2006) provide a useful index for describing thedegree of floc compactness and how particles are packed (Leeand Hsu, 1994). Jin et al. (2003) explained that the high value ofthe Df is related to the compact and dense structure of the granularsludge. In this study, the Df of the granular sludge averaged at 1.90suggesting good compactness of the granules. At above an equiva-lent diameter of 700 lm, the fractal dimension is relatively invari-ant. Below 700 lm there is an almost linear relationship betweenthe size of aerobic granular sludge and fractal dimension howeverthere was no relationship between SVI and Df. Granule size is alsostrongly associated with sludge volume index. Granules larger insize demonstrated poor compressibility with much higher SVI val-ues which could presumably affect the settling ability of granules.

3.3. Microbial community and structure

The well-resolved DGGE bands obtained upon mature granula-tion at different OLRs revealed that the relative populations of thebacterial species in the aerobic granular sludge gradually shiftedwith time. Pholchan et al. (2010) has recently reported that theinfluence of operating conditions and reactor format on the com-munity diversity is evidenced by the changes in banding patterns.Visual analysis of the DGGE profiles showed that some bands werecommon to all reactors albeit the different operating OLRs.

Although the dominant species found in the inoculating seedsludge were mostly preserved throughout the granulation process,new bands were observed in the mature granules at different OLRs.The dominant DGGE bands associated with the activated seedsludge did not co-migrate with the dominant bands obtained fromthe aerobic granular sludge. These results indicated that the majorpopulations in the activated seed sludge were noticeably differentfrom those in the mature aerobic granular sludge. This is presum-ably due to microbial community adaptation towards aerobic gran-ular sludge formation selection pressures in the SBR which forcedthe less dense seed sludge to aggregate into dense and compactgranules (Jiang et al., 2004). Moreover, the mild changes in theintensities of the bands indicated that the microbial communitiesmay have presumably shifted gradually throughout the granula-tion process as recently explained by Zhang et al. (2011).

The similarity indices for samples collected during acclimationand granulation phases were higher than for samples from matura-tion phases as shown in Fig. 2. The average value of the similarityindices obtained during the acclimation/granulation and the matu-ration phases were 0.95 ± 0.02 and 0.72 ± 0.06, respectively. Lightflocs were washed out allowing denser bioparticles to be retainedin the reactor. The maturation phase could be defined as the periodafter the aerobic granular sludge became matured when the set-tling ability were kept at a relatively stable value. The mature gran-ules maintained a good settling ability and a dominant granularmorphology as evidenced by the low SVI of 19.9 mL g�1 SS at3.5 kg COD m�3 d�1. However, the relationship between the abun-dance and band intensity in DGGE profiles of aerobic granularsludge remains unanswered as to whether microbial selection be-tween seed sludge and aerobic granular sludge plays an active rolein aerobic granulation process.

3.4. Identification of microbial population

Visible bands from the DGGE profiles in Fig. 3 were excised,re-amplified, purified and sequenced to determine the identity of

bacterial strains present in the granules as summarized in Table 3.Identification of dominant bacteria as a result from DGGE analysisindicated that most are highly related to uncultured bacteria inwhich some of these bacteria have never been successfully cul-tured. The novelty of uncultured bacteria may not be resolved,though utilization of DGGE method in combination with DNAsequencing technology made the identification of uncultured bac-teria excessively possible. Therefore, the community structure ofbacterial population can be significantly monitored.

As determined by comparison of the 16S rRNA sequences, themajority of bacteria were either dominant species in the seedsludge or gradually migrated as dominant culture in mature gran-ules as recently reported by Zhang et al. (2011). Members such asBacteriodetes and uncultured Propionibacteriaceae were dominantin the activated seed sludge also co-migrated in mature aerobicgranular sludge. Bacteriodetes were observed to be one of the dom-inant bacterial population commonly found in aerobic granules asinvestigated in various studies as previously reported by Dahalan(2011). On the contrary, Zhang et al. (2011) reported that Bacterio-detes bacterium were washed out at short settling times and didnot contribute to sludge granulation. Furthermore, the unculturedMethylobacillus, Bacterium and Flavobacterium were also found par-ticularly during the acclimation and granulation phases of aerobicgranules but were not detected during maturation of aerobic gran-ular sludge. The significant shift in microbial population from accli-mation and granulation phases towards the maturation phasecould be due to the microbial attachment and detachment pro-cesses which occur during these phases to allow for the formationof stable and compact granules in the reactors. Strains such asPseudomonas sp. and Flavobacterium sp. were previously reportedto have no contribution towards aerobic granular sludge formation(Adav et al., 2009).

The uncultured Propionibacteriaceae family presumably indi-cated the presence of phenol-degrading coccus which was previ-ously isolated from phenol-degrading aerobic granules (Maszenanet al., 2007). Phenolic compounds contain a benzene ring towhich one or more hydroxyl groups are attached. According toLimkhuansuwan and Chaiprasert (2010), the presence of phenoliccompounds associated with POME was presumably the cause forthe appearance of dark brown color of raw POME. It is also impor-tant to note that phenol and its derivatives are often associatedwith palm oil refineries and their wastewater (Whiteley and Bailey,

12 18

4

6

5

3

7

8

11

10

9

19

17

2

16

1

13

14

15

Marker MarkerMarker

Seed

slu

dge

Seed

slu

dge

OL

R 2

.5 a

OL

R 2

.5b

OL

R 2

.5c

OL

R 2

.5d

OL

R 2

.5 M

G

OL

R 2

.5e

OL

R 1

.5 M

G

OL

R 2

.5 M

G

OL

R 3

.5 M

G

OL

R 3

.5 M

G

OL

R 1

.5 M

G

OL

R 1

.5 M

G

Ano

xic

cond

.

Ano

xic

cond

.

Size

(bp

)

2000 1800

1600

1400

1200 1000 800 700

600

500

400

Fig. 3. DGGE profiles of the bacterial communities in the three SBRs namely R1 (OLR 1.5 kg COD m�3 d�1), R2 (OLR 2.5 kg COD m�3 d�1), and R3 (OLR 3.5 kg COD m�3 d�1)during the aerobic granulation process. The band references are tabulated in Table 3. (MG: Mature Granule; a, b, c, d, e indicates operation at day-17, 21, 31, 35 and 41,respectively).

Table 3Species identification of selected DGGE bands from the seed sludge and aerobicgranular sludge cultivated at different OLR.

Bands Closest relatives (accession no.) Length(bp)

Identity(%)

1 Kineosphaera limosa (AB550802.1) 524 932 Trichococcus sp. 0 (FJ374769.1) 710 973 Runella sp. (GU223115.1) 1341 984 Pseudomonas sp. (HM051242.1) 799 955 Uncultured Flavobacteriaceae bacterium

(EF651688.1)510 97

6 Uncultured Leadbetterella sp. (GU560170.1) 1465 997 Uncultured bacterium (JF048329) 1341 968 Uncultured bacterium (JF048329) 546 969 Uncultured Comamonadaceae bacterium

(HQ674824.1)1494 98

10 Uncultured bacterium (EU809170) 1274 9111 Uncultured bacterium (JF019340) 1354 10012 Uncultured Bacteroidetes bacterium

(FN669651.1)1474 93

13 Uncultured Propionibacteriaceae bacterium(EU812987)

1499 100

14 Uncultured bacterium (DQ906136) 590 9715 Uncultured Propionibacteriaceae bacterium

(EU812987.1)1499 100

16 Uncultured bacterium (GU562534) 808 9517 Uncultured bacterium (GU731738) 932 9718 Uncultured Methylobacillus sp. (GQ390421.1) 1475 9219 Uncultured bacterium (GQ289449) 166 93

186 N. Abdullah et al. / Bioresource Technology 127 (2013) 181–187

2000). The biodegradation of phenol and phenolic compounds inwastewater is known to be carried out by bacteria of various phy-logeny including members from both the Betaproteobacteria andGammaproteobacteria. These findings suggested the presence of

phenol-degrading bacteria in the cultivated aerobic granular sludgein POME which could be useful for treating phenolic wastewater ata load that would lead to failure in conventional activated sludgesystems (Jiang et al., 2004).

In addition, the presence of the uncultured Propionibacteriaceaefamily indicated that the aerobic granules developed in this studyhave successfully caused an integration of a highly concentratedpopulation of phenolic degraders to be retained in the reactor sys-tem. Whereas, in the conventional activated sludge system thebacteria presence as a suspended cell and is more prone towashout.

Moreover, the phylum Comamonadaceae dominated the reactorafter granulation but was not present in the seed sludge althoughComamonas testosteroni, which is a Gram-negative soil bacteriaoriginating from human pathogen was found in the seed sludge.Previous findings by Adav et al. (2009) and Ginige et al. (2005) sug-gested that the members of the Comamonadaceae family play a ma-jor role in denitrification processes in the presence of acetate in anaerobic granular sludge system. Leadbetterella sp. which is a Gram-negative rod-shaped carbohydrate degrader and Runella sp. whichis known for its phosphate removal capacity (Ryu et al., 2006) werepresent significantly in all reactors throughout the granulationprocess. Given the microbial dominance of mature granules underdifferent OLR, these bacterial species likely played an importantrole in the formation of aerobic granular sludge despite the typeof wastewater being used as substrate.

Additionally, the presence of another genus of bacteria knownas Trichococcus sp. which is one of the Gram-positive filamentouscocci was also detected from the 16S rRNA sequencing analysisof the selected DGGE band. The SEM study has shown the forma-tion of cocci-shaped bacteria which is similar to Trichococcus sp.

N. Abdullah et al. / Bioresource Technology 127 (2013) 181–187 187

Cocci often occurred in spherical clusters of numerous cells whichis in agreement with several studies indicating the presence of coc-ci colonies in aerobic granular sludge microstructures (Jiang et al.,2004).

The outgrowth of filamentous bacteria is usually detrimental toconventional activated sludge systems and may lead to operationaldisorders such as sludge bulking and foaming (Jenkins et al., 1993).The presence of filamentous cocci is also commonly associatedwith the formation of loose microbial structure in glucose-fedaerobic granular sludge with adequate settling and strengthcharacteristics.

4. Conclusions

Aerobic granular sludge was successfully developed at OLRs 2.5and 3.5 kg COD m�3 d�1 while bioflocs remained dominant at OLR1.5 kg COD m�3 d�1, respectively. Different volumetric loadingrates led to differences in morphology and structural features ofgranule in which higher loading rates promoted the formation oflarger and looser aerobic granular sludge. The PCR–DGGE resultsindicate the microbial evolution throughout aerobic granulation.Low similarity index was observed in the reactor with high organicloading rates. These findings represent important informationabout microbial community structure and the factors that are cru-cial for successful aerobic granulation in treating high strengthwastewaters.

Acknowledgements

We wish to thank Felda Bukit Besar palm oil for providing ac-cess to their sites for sampling. This work is funded by UniversitiTeknologi Malaysia (UTM) and Ministry of Higher Education (GrantNo. 72468). Norhayati Abdullah wishes to thank Dr. Russell Daven-port, Mrs. Fiona Read, Dr. Micol Belluci and Dr. Trevor Booth ofNewcastle University, England for their expert advice and guidanceon molecular experiments.

References

Abdullah, N., Ujang, Z., Yahya, A., 2011. Aerobic granular sludge formation for highstrength agro-based wastewater treatment. Bioresource Technology 102, 6778–6781.

Adav, S.S., Lee, D.J., Lai, J.Y., 2009. Aerobic granulation in sequencing batch reactorsat different settling times. Bioresource Technology 100, 5359–5361.

Agensi Inovasi Malaysia, 2011. National Biomass Strategy 2020: New WealthCreation for Malaysia’s Palm Oil Industry.

Benson, D.A., Karsch-Mizrachi, I., Lipman, D.J., Ostell, J., Wheeler, D.L., 2008. NucleicAcids Research 36, 25–30.

Beun, J.J., Hendriks, A., van Loosdrecht, M.C.M., Morgenroth, E., Wilderer, P.A.,Heijnen, J.J., 1999. Aerobic granulation in a sequencing batch reactor. WaterResearch 33 (10), 2283–2290.

Cassidy, D.P., Belia, E., 2005. Nitrogen and phosphorus removal from an abattoirwastewater in a SBR with aerobic granular sludge. Water Research 39, 4817–4823.

Chan, Y.J., Chong, M.F., Law, C.L., 2010. Biological treatment of anaerobicallydigested palm oil mill effluent (POME) using a lab-scale sequencing batchreactor (SBR). Journal of Environmental Management 91, 1738–1746.

Collins, T.J., 2007. ImageJ for microscopy. BioTechniques 43, S25–S30.Dahalan, F.A., 2011. Development of Photosynthetic Aerobic Granular Sludge. Ph.D.

Dissertation. Universiti Teknologi Malaysia, Johor Bahru, Malaysia.Damayanti, A., Ujang, Z., Salim, M.R., Olsson, G., Sulaiman, A.Z., 2010. Respirometric

analysis of activated sludge models from palm oil mill effluent. BioresourceTechnology 101, 144–149.

Federal Subsidiary Legislation-Environmental Quality Act, 1974. [ACT 127]Environmental Quality (Sewage and Industrial Effluents) Regulation 1979[Online]. Available: <http://www.doe.gov.my>.

Ginige, M.P., Jurg, K., Blackall, L.L., 2005. Investigation of an acetate-fed denitrifyingmicrobial community by stable isotope probing, fullcyle rRNA analysis, andfluorescent in situ hybridization microautobiography. Applied EnvironmentalMicrobiology 71, 8683–8691.

Hammer, Ø., Harper, D.A.T., Ryan, P.D., 2001. PAST: paleontological statisticssoftware package for education and data analysis. Palaeontologia Electronica4(1), 9pp. Retrieved at: <http://palaeoelectronica.org/2001_1/past/issue1_01.htm>.

Hansen, S.B., Olsen, S.I., Ujang, Z., 2012. Greenhouse gas reductions throughenhanced use of residues in the life cycle of Malaysian palm oil derivedbiodiesel. Bioresource Technology 104, 358–366.

Jenkins, D., Richard, M.G., Daigger, C.T., 1993. Manual on the Causes and Control ofActivated Sludge Bulking and Foaming, second ed. Lewis Publishers, Chelsea,Michigan.

Jiang, H.L., Tay, J.H., Maszenan, A.M., Tay, S.T.L., 2004. Bacterial diversity andfunction of aerobic granules engineered in a sequencing batch reactor forphenol degradation. Applied and Environmental Microbiology 70 (11), 6767–6775.

Jin, B., Wilen, B.-M., Lant, P., 2003. A comprehensive insight into floc characteristicsand their impact on compressibility and settleability of activated sludge.Chemical Engineering Journal 95, 221–234.

Kishida, N., Tsuneda, J., Kim, J.H., Sudo, R., 2009. Simultaneous nitrogen andphosphorus removal from high-strength industrial wastewater using aerobicgranular sludge. Journal of Environmental Engineering ASCE, 153–158.

Kowalchuk, G.A., Stephen, J.R., Deboer, W., Prosser, J.I., Embley, T.M., Woldendorp,J.W., 1997. Analysis of ammonia-oxidizing bacteria of the subdivision of theclass proteobacteria in coastal sand dunes by denaturing gradient gelelectrophoresis and sequencing of PCR-amplified 16S ribosomal DNAfragments. Applied and Environmental Microbiology 63, 1489–1497.

Lee, D.J., Hsu, Y.H., 1994. Fast freeze/thaw treatment on excess activated sludges:floc structure and sludge dewaterability. Environmental Science Technology 28,1444–1449.

Limkhuansuwan, V., Chaiprasert, P., 2010. Decolorization of molasses meladoninsand palm oil mill effluent phenolic compounds by fermentative lactic acidbacteria. Journal of Environmental Sciences 22, 1209–1217.

Mandelbrot, B.B., 2006. Fractal analysis and synthesis of fracture surface roughnessand related forms of complexity and disorder. International Journal of Fracture138, 13–17.

Martins, A.M.P., Heijnen, J.J., van Loosdrecht, M.C.M., 2003. Effect of feeding patternand storage on the sludge settleability under aerobic conditions. WaterResearch 37, 2555–2570.

Maszenan, A.M., Jiang, H.L., Tay, J.H., Schumann, P., Kroppenstedt, R.M., Tay, S.T.L.,2007. Granulicoccus phenolivorans gen. nov., sp. nov., a Gram-positive, phenol-degrading coccus isolated from phenol-degrading aerobic granules.International Journal of Systematic and Evolutionary Microbiology 57, 730–737.

Mu, Y., Yu, H.-Q., 2006. Rheological and fractal characteristics of granular sludge inan upflow anaerobic reactor. Water Research 40, 3596–3602.

Muyzer, G., de Waal, E.C., Uitterlinden, A.G., 1993. Profiling of complex microbialpopulations by denaturing gradient gel electrophoresis analysis of polymerasechain reaction-amplified genes coding for 16S rRNA. Applied andEnvironmental Microbiology 59, 695–700.

Parker, D.S., Kaufman, W.J., Jenkins, T., 1971. Physical conditioning of activatedsludge floc. Journal Water Pollution Control Federation 43 (9), 1817–1833.

Poh, P.E., Chong, M.F., 2009. Development of anaerobic digestion methods for palmoil mill effluent (POME) treatment. Bioresource Technology 100, 1–9.

Pholchan, M.K., de C Baptista, J., Davenport, R.J., Curtis, T.P., 2010. Systematic studyof the effect of operating variables on reactor performance and microbialdiversity in laboratory-scale activated sludge reactors. Water Research 44,1341–1352.

Ryu, S.H., Nguyen, T.T., Park, W., Kim, C.J., Jeon, C.O., 2006. Runella limosa sp. Nov.,isolated from activated sludge. International Journal of Systematic EvolutionMicrobiology 56, 2757–2760.

Tay, J.H., Pan, S., He, Y., Tay, S.T.L., 2004. Effect of organic loading rate on aerobicgranulation II: characteristics of aerobic granules. Journal of EnvironmentalEngineering 130 (10), 1102–1109.

Thanh, B.X., Visvanathan, C., Aim, R.B., 2009. Characterization of aerobic granularsludge at various organic loading rates. Process Biochemistry 44, 242–245.

Whiteley, A.S., Bailey, M.J., 2000. Bacterial community structure and physiologicalstate within an industrial phenol bioremediation system. Applied EnvironmentMicrobiology 66, 2400–2407.

Zhang, B., Min, J., Zhigang, Q., Huina, L., Jingfeng, W., Junwen, L., 2011. Microbialpopulation dynamics during sludge granulation in an anerobic–aerobicbiological phosphorus removal system. Bioresource Technology 102, 2474–2480.

Zinatizadeh, A.A.L., Mohamed, A.R., Mashitah, M.D., Abdullah, A.Z., Isa, M.H., 2007.Optimization of pre-treated palm oil mill effluent digestion in an up-flowanaerobic sludge fixed film bioreactor: a comparative study. BiochemicalEngineering Journal 35, 226–237.