BIOENERGY AND BIOFUELS

Structure and dynamics of the microbial communitiesunderlying the carboxylate platform for biofuel production

Emily B. Hollister & Andrea K. Forrest & Heather H. Wilkinson & Daniel J. Ebbole &

Stephanie A. Malfatti & Susannah G. Tringe & Mark T. Holtzapple & Terry J. Gentry

Received: 1 June 2010 /Revised: 9 July 2010 /Accepted: 11 July 2010 /Published online: 31 July 2010# Springer-Verlag 2010

Abstract The carboxylate platform utilizes a mixed micro-bial community to convert lignocellulosic biomass intochemicals and fuels. While much of the platform is wellunderstood, little is known about its microbiology. Mesophilic(40 °C) and thermophilic (55 °C) fermentations employing asorghum feedstock and marine sediment inoculum wereprofiled using 16S rRNA tag-pyrosequencing over the courseof a 30-day incubation. The contrasting fermentation temper-atures converted similar amounts of biomass, but themesophilic community was significantly more productive,and the two temperatures differed significantly with respect topropionic and butyric acid production. Pyrotag sequencingrevealed the presence of dynamic communities that respondedrapidly to temperature and changed substantially over time.Both temperatures were dominated by bacteria resembling

Clostridia, but they shared few taxa in common. The species-richmesophilic community harbored a variety of Bacteroidetes,Actinobacteria, and γ-Proteobacteria, whereas the thermophiliccommunity was composed mainly of Clostridia andBacilli. Despite differences in composition and produc-tivity, similar patterns of functional class dynamics wereobserved. Over time, organisms resembling knowncellulose degraders decreased in abundance, while organ-isms resembling known xylose degraders increased.Improved understanding of the carboxylate platform’smicrobiology will help refine platform performance andcontribute to our growing knowledge regarding biomassconversion and biofuel production processes.

Keywords Mixed alcohol bioreactor . Lignocellulosicbiofuels . Tag-pyrosequencing .Microbial communities .

Carboxylate platform

Introduction

As energy demands continue to increase and the world’sfossil fuel reserves steadily decline, the need to developstable, alternative energy sources continues to grow. Next-generation biofuels (i.e., those derived from cellulosicfeedstocks) have the potential to contribute to meeting aportion of those needs (Perlack et al. 2005; Rubin 2008). Itis estimated that the United States has the potential toproduce in excess of 1.3 billion tons of biomass feedstockeach year for the production of lignocellulosic biofuels(Perlack et al. 2005). Although biomass-based fuels cannotbe produced on such a scale as to replace petroleum-basedfuels entirely, it is thought that they have the potential tofulfill ∼30% of the U.S. annual transportation fuel demands(Perlack et al. 2005).

Electronic supplementary material The online version of this article(doi:10.1007/s00253-010-2789-7) contains supplementary material,which is available to authorized users.

E. B. Hollister (*) : T. J. GentryDepartment of Soil and Crop Sciences, Texas A&M University,370 Olsen Blvd, TAMU 2474,College Station, TX 77843-2474, USAe-mail: ehollister@tamu.edu

A. K. Forrest :M. T. HoltzappleDepartment of Chemical Engineering, Texas A&M University,College Station, TX 77843-3122, USA

H. H. Wilkinson :D. J. EbboleDepartment of Plant Pathology and Microbiology,Texas A&M University,College Station, TX 77843-2132, USA

S. A. Malfatti : S. G. TringeDOE Joint Genome Institute,Walnut Creek, CA 94598, USA

Appl Microbiol Biotechnol (2010) 88:389–399DOI 10.1007/s00253-010-2789-7

Multiple strategies have emerged to convert lignocellu-losic feedstocks into fuels. The two most common are thesugar platform (i.e., the enzymatic hydrolysis of celluloseinto simple sugars followed by aseptic fermentation) andthermochemical conversion (i.e., gasification followed bycatalysis). Although each of these has proven useful, theirlarge-scale implementation has been hampered by highcosts associated with their construction, processing, andmaintenance (Granda and Holtzapple 2008; Lin and Tanaka2006). A lesser known alternative to these strategies is thecarboxylate platform, also known as the MixAlco process(Holtzapple et al. 1999).

The carboxylate platform is a well-optimized, flexible, andcost-effective means to convert lignocellulosic feedstocks—including crop residues, dedicated energy crops, municipalsolid waste, and biosolids—into chemicals and liquid fuels(Aiello-Mazzarri et al. 2006; Granda et al. 2009; Holtzappleet al. 1999; Thanakoses et al. 2003). Contributing to its cost-effectiveness, the carboxylate platform operates under non-sterile conditions and uses a mixed community of naturallyoccurring anaerobic microorganisms to provide the enzymesnecessary for biomass conversion. The mixed-cultureapproach of this system underlies its flexible nature as well,allowing the platform to convert sugars, as well as proteins,pectin, oils, and fats. The products of the mixed-culturefermentation are carboxylate salts, which can be transformedthrough chemical means to a wide variety of products,including alcohols, jet fuel, and gasoline. In addition, theproduct spectrum from this process is temperature dependent(Chan and Holtzapple 2003; Fu and Holtzapple 2010) andcan be varied in response to market demands.

Over the course of its development, a number of differentinoculum sources have been utilized in the carboxylateplatform (Thanakoses et al. 2003). Despite such attempts toidentify better-performing microbial communities, the plat-form’s microbial component has remained a “black box”relative to other well-optimized and understood aspects ofthis system. One of the best performing inoculum sourcesthat has been found, to date, is a marine sediment collectedalong the Gulf of Mexico coastline at Galveston Island, TX,USA (Thanakoses et al. 2003). While this communityappears to be well-adapted to the conditions of thecarboxylate platform, it is unclear which microorganismsare present in this consortium and, of those, which areinvolved in the process of biomass conversion in thecarboxylate platform.

The objective of this study was to characterize thecarboxylate platform’s bacterial communities, seeded fromthe Galveston inoculum, over the course of a 30-day,laboratory-scale incubation, and under both mesophilic(40 °C) and thermophilic (55 °C) conditions. Thiscommunity-focused approach, involving the production oflarge-scale tag-pyrosequence libraries, was used in combi-

nation with reactor performance metrics to characterize thecomposition of the reactor microbial communities andthe relationships that they share with the system’sproductivity and performance. It is anticipated thatincreased understanding of the microbial communitiesinvolved in the carboxylate platform will help refine thedesign and operation of the bioreactor system, facilitatethe development of inocula for large-scale implementa-tion of this process, and contribute to the developingbody of knowledge regarding biomass conversion andbiofuel production.

Materials and methods

Bioreactor substrate, inoculum source, and pre-incubationpreparation

A photoperiod-sensitive, high-tonnage sorghum cultivar(Sorghum bicolor (L.) Moench) was obtained from WilliamRooney (Sorghum Breeding and Genetics Program, TexasA&M University) and used as the biomass feedstock forthis series of fermentations. The sorghum was chipped,dried, and treated with hot water and lime (0.1 g Ca(OH) 2)and 10 mL distilled H2O per g dry biomass; 2 h at 100 °C)to enhance its digestibility (Chang et al. 1998).

The inoculum used in the reactor system was obtainedfrom marine sediment collected from Galveston, TX inMay 2008. Sediment was collected from multiple shorelinepits dug to 0.5 m depth, adjacent to the high-water mark atthe time of sampling. At 0.5-m depth, there was a distinctcolor change from yellow/brown sand to dark gray/blacksediment. Upon collection, sediment samples were placedinto bottles filled with deoxygenated water, 0.275 g L−1

sodium sulfate, and 0.275 g L−1 cysteine hydrochloride, asdescribed by Thanakoses et al. (2003). The bottles werecapped and held on ice. Upon return to the laboratory,samples were frozen at −20 °C until later use. Just prior toinoculation, a sediment sample was thawed, shakenvigorously, and allowed to settle by gravity. Aliquots ofthe resulting supernatant were then used to inoculate thebioreactor vessels.

Bioreactor construction and performance monitoring

Bioreactor vessels were constructed using 1-L polypropylenecentrifuge bottles fitted with a stirring and venting apparatus,as described by Fu and Holtzapple (2010). Each reactorcontained 50 mL mixed microbial inoculum, 36 g lime-treated sorghum, 4 g dried chicken manure (included as anutrient source and obtained from the Poultry Science Centerat Texas A&M University, College Station, TX), 350 mLdeoxygenated water, a calcium carbonate buffer (CaCO3,

390 Appl Microbiol Biotechnol (2010) 88:389–399

15 g L−1), and iodoform (CH3I, 20 g L−1) and was rolledcontinuously at 2 rpm over the course of the incubation (Fuand Holtzapple 2010). Reactors were flushed with N2 duringthe preparation process to create an anaerobic environment,and the iodoform was used to inhibit methane production.Two fermentation temperatures (40 and 55 °C) wereemployed, and the experiment was designed so that a set(n=3) of reactor vessels from each temperature treatmentcould be sacrificed for microbial community characterizationat the early (day 6), middle (day 16), and late (day 30) stagesof the incubation.

Carbon dioxide (CO2) and methane concentrations, aswell as pH and total carboxylic acid concentrations weremonitored every 2 days over the course of the incubation.The volume of gas produced was quantified by displacing aCaCl2 solution, and its CO2 and CH4 concentrations weremeasured using an Agilent 6890 gas chromatograph(Agilent Technologies, Palo Alto, CA, USA) equippedwith a thermal conductivity detector. pH was measured andmonitored using an ORION pH meter (Thermo ElectronCorporation, Beverly, MA, USA), and total carboxylic acidconcentrations in the fermentation broth were measured withan Agilent 6890 gas chromatograph (Agilent Technologies)equipped with a flame ionization detector and anAgilent 7683series injector (Agilent Technologies).

The products in the fermentation broth consisted of amixture of carboxylic acids and carboxylate salts. Samplesof the fermentation broth were combined with 1.162 g L−1

of a 4-methyl-n-valeric acid internal standard and wereacidified with 3 M phosphoric acid prior to injection intothe gas chromatograph. By acidifying this mixture, eachcomponent was converted to its corresponding acid, allowingproduct concentrations to be reported as carboxylic acids ing L−1. Column head pressure was maintained at 2 atm abs,and samples were heated from 40 °C to 200 °C at a rate of20 °C min−1 prior to being held at 200 °C for an additional2 min. Helium was used as the carrier gas, and the total runtime per sample was 11 min.

As each subset of the fermentation reactions wasterminated (i.e., early, middle, and late stages of incubation),samples of both the solid and liquid phases werecollected from the reactor vessels for chemical analysis.The reactor bottles were centrifuged, in a Beckman J-6Bcentrifuge (Beckman Coulter, Inc., Brea, CA, USA) witha swinging bucket rotor at 3,297×g for 30 min to separatethe solids and liquids from one another. An aliquot of thesupernatant was collected and subjected to carboxylic acidanalysis, as described above, and the solids were analyzedto determine the mass of undigested volatile solids (VS)first by drying at 105 °C and then by ashing at 550 °C (Fuand Holtzapple 2010). VS content was calculated as thedifference between the oven dry weight and the ashedweight of each sample.

DNA extraction and pyrotag sequencing

Reactor materials were collected early in the incubation(day 6), mid-incubation (day 16), and at the end of theincubation (day 30). Solids and liquids from each replicatewere combined into one vessel in equal volumes to create asingle composite sample for each treatment, at each timepoint. The composite samples were frozen and storedat −80 °C until DNA extraction. Just prior to extraction,the reactor samples were thawed and centrifuged at4,000×g for 10 min. DNA was extracted from the pelletmaterials using a PowerMax soil DNA extraction kit (MoBio Laboratories, Inc., Carlsbad, CA, USA), using alysozyme-modified version of the manufacturer’s protocol(Hollister et al. 2010). Briefly, 15 g aliquots (wet weight) ofreactor material and 15 mL of bead solution were added toeach bead beating tube. After 5 min of bead beating,lysozyme was added (final concentration of 1 mg mL−1),and samples were incubated at 37 °C for 1 h. Followinglysozyme treatment, solution “C1” was added and sampleswere incubated at 65 °C for 30 min. The manufacturer’sprotocol was followed from this point onward. Followingelution, DNA samples were concentrated and purified usingillustra MicroSpin S-400 HR columns (GE Healthcare Bio-Sciences Corp, Piscataway, NJ, USA).

DNA samples were quality-checked according to USDOE Joint Genome Institute protocols and were submit-ted to the Joint Genome Institute for pyrotag sequencing.Hypervariable regions V6–V8 of the 16S rRNA genewere amplified with universal primers 926F and 1392R,including the titanium adapter sequences as well as afive-base barcode on the reverse primer (Engelbrektsonet al. 2010). Reactions were performed in triplicate with25 cycles of amplification, checked by agarose gelelectrophoresis then quantified and pooled in equalamounts for sequencing by standard 454-titanium protocol(Engelbrektson et al. 2010).

Analysis of reactor performance

Reactor performance was characterized through a series ofmetrics that included conversion, selectivity, yield, andproductivity. Conversion was calculated as the fraction ofVS that was digested over the course of the incubationrelative to the quantity of VS contained in the biomass thatwas initially loaded into the reactor. Selectivity wascalculated as the proportion of digested material thatresulted in carboxylic acid production. Yield was deter-mined by calculating the ratio of total carboxylic acidsproduced relative to the quantity of VS contained in thebiomass that was initially loaded into the reactor, andproductivity was calculated as the overall rate of acidproduction (g acid L−1 day−1). Comparisons of these values,

Appl Microbiol Biotechnol (2010) 88:389–399 391

as well as the relative abundances of various acid products,between fermentation temperatures were conducted usingpaired, two-tailed Student’s t tests, and p values <0.05 wereconsidered to represent significant differences.

Microbial community characterization and comparisons

Bioreactor community DNA sequences were quality-checked, trimmed to a common length, and clustered(97% similarity) using the Joint Genome Institute’sPyroTagger (http://pyrotagger.jgi-psf.org/cgi-bin/index.pl)pipeline. A representative sequence from each cluster wascompared against the Greengenes NAST-aligned database(DeSantis et al. 2006) and used to assign putativeidentities to the entire sequence data set. The quality-checked output was reformatted and input into theRibosomal Database Project (RDP) pyrosequencing pipe-line (Cole et al. 2009) and the Mothur software package v.1.6 (Schloss et al. 2009) to perform comparisons of thebioreactor bacterial communities across temperature treat-ments and over time. RPD was used to align sequences,and Mothur was used to calculate distance matrices, assignsequences to operational taxonomic units (OTU, 97%similarity), calculate diversity indices and richness esti-mates, and determine the degree of overlap shared amongthe reactor communities. Overlap was calculated using theYue–Clayton similarity estimator (θYC), a metric that isscored on a scale of 0–1, where 0 represents completedissimilarity and 1 represents identity (Schloss et al. 2009;Yue and Clayton 2005). When comparing any given set ofcommunities, θYC considers the distribution of OTUsbetween the communities, as well as their relativeabundance. Additional comparisons of community struc-ture were carried out using nonmetric multidimensionalscaling of the OTU data using the Bray–Curtis similaritymetric, as implemented in the PAST software program(Hammer et al. 2001).

In an attempt to link community composition to reactorperformance, the relative abundance of each OTU wasregressed against the reactor performance metrics describedabove. Given the size of the data set, only those OTUs thataccounted for ≥1% of the entire community, at any given

time point, were considered. Those that had r2 correlationsof 0.5 or better were considered to be “performance related”and were subject to additional analysis. Nearest-neighborsequences for each of the performance-related OTUs, aswell as the five most abundant OTUs present in the middle-of-incubation communities were identified and used tobuild phylogenetic trees to illustrate the relationships sharedamong them and aid in their classification. Three nearest-neighbor sequences for each reactor OTU were obtainedfrom the Greengenes chimera-checked database (DeSantiset al. 2006). Sequences were aligned using the Alignerfeature at the Ribosomal Database Project (Cole et al.2009), they were bootstrapped with 100 replicates usingPhylip’s Seqboot function, and then the correspondingdistance matrices were calculated using DNAdist with theJukes–Cantor evolutionary model (Felsenstein 2005). Phylip’sNeighbor function was used to construct neighbor-joiningphylogenetic trees, and the Consense function was used toidentify a consensus tree (Felsenstein 2005).

Pyrotag sequence reads were submitted to the NCBIShort Read Archive under accession number SRA020048.

Results

Reactor performance

After 30 days’ incubation, the 40 and 55 °C fermentationtemperatures resulted in similar rates of biomass conversion(Table 1); however, the reactors incubated under thesetemperature regimes differed significantly with respect totheir selectivity, yield, and productivity. Each of theseperformance metrics was significantly greater at 40 °C thanat 55 °C, representing increases of 40%–60% over the 55 °Cvalues. The two temperature conditions also differedsignificantly with respect to the abundances of multipleacids within their product spectra (Table 2). Although aceticacid (C2) dominated and was equally abundant in bothreactor systems, the fermentations at 40 °C produced asignificantly larger proportion of propionic (C3) acidproducts and a significantly smaller proportion of butyric(C4) acid products than those at 55 °C.

Table 1 Reactor performance metrics following 30 days’ incubation

Treatment Conversiona Selectivityb Yieldc Productivity (g acid L−1 day−1)

40 °C incubation 0.231±0.023a 0.371±0.024a 0.085±0.004a 0.230±0.012a

55 °C incubation 0.259±0.032a 0.232±0.027b 0.059±0.001b 0.159±0.001b

Values represent the mean of three replicates ± SE. Different letters within a column indicate a statistically significant difference at p<0.05a Conversion represents the ratio of volatile solids digested to the biomass that was originally loaded into the reactorb Selectivity represents the proportion of digested material that resulted in carboxylic acid productionc Yield represents the ratio of total carboxylic acids produced to the biomass that was originally loaded into the reactor

392 Appl Microbiol Biotechnol (2010) 88:389–399

Microbial community characterization and comparisonsbetween temperatures and over time

Following the removal of short, ambiguous, and/or low-quality pyrotag reads, the final data set, representing bothfermentation temperatures and all time points, consisted of160,957 high-quality partial rRNA sequences. Despite theuse of universal primers in sequencing these communities,99% of all of the sequences recovered were of bacterialorigin (data not shown). Sequence library size ranged from15,344 sequences in the 55 °C mid-incubation sample to46,359 sequences in the 40 °C early incubation sample(Table 3). Among all samples 2,955 unique OTUs (97%similarity cutoff) were identified. The number of OTUscontained within each sequence library covered an order ofmagnitude difference, with as few as 140 OTUs occurringin the 55 °C mid-incubation sample and as many as 1,382OTUs occurring in the 40 °C-early incubation sample.Likewise, estimates of community diversity (H′) andrichness (Chao I) varied widely along temperature lines(Table 3).

The reactor communities were dominated by members ofthe phylum Firmicutes, and taxa representing the bacterialclasses Clostridia and Bacilli were common to bothfermentation temperatures and all time points. Althoughthe reactor communities were similar to one another atbroad taxonomic levels, their OTU composition differeddramatically with respect to incubation temperature (Fig. 1)and changed substantially over time (Table 4). Despite thepredominance of Clostridia and Bacilli in the reactorscommunities (Fig. 2), very little overlap was detectedbetween the two fermentation temperatures with respect toOTU-level community membership and structure (Table 4).Among the 2,955 OTUs identified, only 2% were found atboth 40 and 55 °C. The segregation of taxa by fermentationtemperature can also be seen in Fig. 3, a neighbor-joiningtree of the most abundant OTUs encountered in the mid-incubation communities. Among the OTUs represented,only clusters #1 (55 °C-specific) and #28 (40 °C-specific)share a close, common position on the tree.

Differences among the reactor communities that occurredwith respect to time are indicated by Yue–Clayton similarityvalues in the highlighted portions of Table 4. Samples thatwere taken closer in time (i.e., early- vs. mid-incubation ormid- vs. late-incubation) were more similar to one anotherthan were those that were taken following longer gaps intime (i.e., early- vs. late-incubation). In the early stages ofincubation, members of the class Clostridia dominated thereactor communities at both temperatures. A pair ofClostridiales-like OTUs accounted for 34% and 19% of theearly incubation sequences in the 55 °C and 40 °Ccommunities, respectively (Supplemental Table 1). Withtime, other taxa increased in abundance and displaced theseClostridiales-like OTUs as the predominant members of eachcommunity. In the 40 °C communities, Bacteroidetes,particularly OTUs resembling members of the Prevotellaceae,became the most commonly detected taxa, and members ofthe γ-Proteobacteria and Actinobacteria increased in abun-dance as well. In the 55 °C communities, OTUs resemblingThermoanaerobacterium spp. dominated the mid-incubationsequence library, and a Clostridium cellulolyticum-like OTUdominated the late-incubation sequence library. Concurrent

Table 2 Distribution of fermentation products under contrasting reaction incubation temperatures following 30 days’ incubation

Treatment Relative abundance of acid products (%)

Acetic (C2) Propionic (C3) Butyric (C4) Valeric (C5) Caproic (C6)

40 °C incubation 63.40±2.62a 14.43±1.14a 16.46±1.82a 2.72±0.44a 2.98±0.88a

55 °C incubation 64.47±1.86a NDb 34.23±1.86b 1.26±0.02a NDa

Values represent the mean of three replicates ± SE. Different letters, within a column, indicate a statistically significant difference at p<0.05 asdetermined by Student’s t test

ND refers to acid products that were not detected

Table 3 Summary of community diversity attributes based uponOTUs and their relative abundances

Reactorcommunity

Community characteristics

Sequencelibrary size

Number ofOTUs

Shannon(H’)

Chao I richnessestimate

Overall 160,957 2955 – –

40 °C early 46,359 1382 3.61 2878

40 °C mid 41,732 978 3.59 1882

40 °C late 17,033 526 3.46 1162

55 °C early 22,406 387 2.52 847

55 °C mid 15,344 140 2.34 271

55 °C late 18,083 479 3.21 888

OTUs were defined at a 97% similarity level

Early refers to samples taken after 6 days’ incubation, mid refers tosamples taken after 16 days’ incubation, Late refers to samples takenafter 30 days’ incubation

Appl Microbiol Biotechnol (2010) 88:389–399 393

with shifts in dominance among various types of Clostridiain the 55 °C community, increases were also detected withrespect to the relative abundance of Bacillus spp. Additionalinformation regarding the relative abundances of these OTUsand their identities at both temperatures and at each timepoint along the study can be found in Supplemental Table 1.

Subpopulations correlating with reactor performance

Between the two fermentation temperatures, ten OTUs wereidentified as sharing strong correlations with reactor produc-tivity (Table 5). While few of these OTUs displayedtemperature-exclusive distributions (i.e., Clusters #2, 27,and 28 only occurred at 40 °C), most displayed strongtemperature affiliations (i.e., the majority of sequencesaffiliated with each OTU were found at one fermentationtemperature only, rather than being equally distributed acrossboth). Of the ten OTUs identified, half were found to be

correlated positively with reactor productivity rates. Theseincluded Clusters #1, 2, 5, 14, and 28, each of which wasfound to be closely related to a Clostridium sp (Fig. 4). Incontrast, the other five OTUs were negatively correlated withreactor productivity rates and drew from a much broadertaxonomic range, including Actinobacteria, Proteobacteria,members of the class Bacilli, Symbiobacteria, and otherClostridium-like organisms. The phylogenetic relationshipsof these organisms and their closest characterized isolates areillustrated in Fig. 4, and additional information regarding therelative abundances of each OTU can be found inSupplemental Table 2.

Discussion

Although many aspects of the carboxylate platform arewell-optimized and understood, this study represents thefirst attempt to characterize the microbial portion of thissystem. While PCR-based characterizations of communitycomposition, including tag-pyroseqencing, are not withoutpotential biases [e.g., DNA extraction efficiency, thepotential for bacteria to carry multiple copies of the 16Sgene, primer bias, and sequencing error (Kunin et al. 2010;Lee et al. 2009; von Wintzingerode et al. 1997)], theyprovide valuable insight into communities that contain taxathat are difficult to cultivate and can be useful in studyingtemporal changes within the same ecosystem (Fernandezet al. 1999). Furthermore, recent evidence suggests thatthorough controls on data quality and clustering can help toeliminate some of the potential biases associated withpyrotag-induced errors (Kunin et al. 2010). With thesecaveats in mind, these results suggest that the bacterialcommunities that underlie biomass conversion in thecarboxylate platform are dynamic, respond rapidly toaltered temperature conditions, and change substantiallyover time. The reactors were dominated, in large part, bybacteria associated with the class Clostridia. Although the

Fig 1 Nonmetric multidimensional scaling of reactor communities,based upon their OTU composition, shows distinct separation of thebioreactor communities with respect to fermentation temperature (40vs. 55 °C) and time. Early, mid, and late refer to incubation lengths of6, 16, and 30 days, respectively

Table 4 Overlap of OTUs among the reactor communities, as represented by θYC, the Yue–Clayton estimator of similarity

Reactor community 40 ºC early 40 ºC mid 40 ºC late 55 ºC early 55 ºC mid 55 ºC late

40 ºC early --- 40 ºC mid 0.4236 ---

---

40 ºC late 0.0615 0.5885 55 ºC early 0.0001 < 0.0001 0.0412 ---

--- ---

55 ºC mid 0.0001 0.0003 0.0488 0.7555 55 ºC late 0.0020 0.0044 0.0346 0.2983 0.4071

θYC is scored on a scale of 0–1, with a score of 0 representing complete dissimilarity and a score of 1 representing identity

Clusters of shading have been placed within the table to highlight the changes within each temperature treatment over time

394 Appl Microbiol Biotechnol (2010) 88:389–399

best match to many of these were unclassified Clostridium-like organisms originating from a broad range of habitats,this group also included organisms similar to a number ofwell characterized biomass-degrading and potentially cel-lulolytic Clostridium spp. and Thermoanaerobacterium spp(Supplemental Table 1). In addition, members of the classBacilli were particularly abundant in the 55 °C community,and members of the Bacteroidetes were commonly encoun-tered in the 40 °C community.

Studies of other biomass conversion systems haveisolated and characterized similar taxa in attempts toidentify organisms involved in the conversion of biomassand solid waste into acids, alcohols, biogas, and hydrogen(Burrell et al. 2004; Goberna et al. 2009; Klocke et al.2007; Ren et al. 2007; Shin et al. 2004). Clostridia oftendominate these systems, though they are frequently foundin co-culture with a number of other species. The bacterialclass Clostridia includes a number of anaerobic, thermo-

Fig 2 Distribution of bacterialclasses (foreground) versus bio-reactor acid products (back-ground) over 30 days’ incubationat a 40 °C and b 55 °C

Prevotella sp. str. DJF CP65 (EU728727.1)

Cluster 3 (14% of the 40 °C community) Prevotella sp. str. RZ (AF040719.1)

Prevotella albensis str. M384 (AJ011683.1)

Bacteroides caccae str. ATCC 43185T (X83951.1)

Bacteroides sp. str. XDT-1 (AB363973.1)

Cluster 6 (10% of the 40 °C community) Bacteroides sp. str. 253c (AY082449.1)

Bacillus sp. str. SL170 (EU266058.1)

Bacillus sp. str. SL168 (EU266059.1)

Bacillus sp. str. SL167 (EU266057.1)

Cluster 20 (9% of the 40 °C community) Bacillus sp. str. ICPS10 (U62623.1)

Bacillus thermozeamaize (AY288912.1)

Bacillus sp. NS2-1 (AY466702.1)

Cluster 39 (12% of the 55 °C community) Paenibacillus brasilensis str. 172 (AF273740.1)

Clostridium cellulosi (L09177.1)

Cluster 14 (14% of the 55 °C community) Ethanologenbacterium harbinense str. CGMCC1152 (AY833426.1)

Clostridium sp. str. rennanqilfy3 (AY363375.2)

Symbiobacterium toebii str. SC-1 (AF190460.1)

Thermoanaerobacterium saccharolyticum str. B6A (FJ268955.1)

Thermoanaerobacterium aotearoense str. JW/SL-NZ613T (X93359.1)

Cluster 8 (25% of the 55 °C community) Thermoanaerobacterium saccharolyticum str. B6A-RI (L09169.1)

Cluster 15 (8% of the 55 °C community) Caldicellulosiruptor hydrothermalis str. 108 (EF100908.1)

Clostridium mesophilum str. SW408 (EU037903.1)

Clostridium sp. str. BL-20 (DQ196623.2)

Clostridium sp. CYP8 (DQ479418.1)

Cluster 2 (11% of the 40 °C community) Clostridium lavalense str. CCRI-9842 (EF564277.1)

Clostridium jejuense str. HY-35-12T (AY494606.1)

Clostridium xylanovorans str. HESP1 (AF116920.1)

Clostridiaceae str. 80Wc (AB078860.1)

Cluster 28 (6% of the 40 °C community)

Cluster 1 (20% of the 55 °C community) Escherichia coli (U00096)

72

100

100

100

83

100

100

10097

100

100100

72

69

97

99

86

52

100

100

100

56

100

100

100

62

100100

Fig 3 Neighbor-joining phylo-genetic tree of the five mostabundant OTUs found withineach bioreactor communityfollowing 16 days’ incubation at40 °C (blue) and 55 °C (red) andtheir three nearest-neighbor iso-lates from the Greengeneschimera-checked database.Bootstrap values are from 100replicates and values <50 are notpresented. The relative abun-dance of each cluster (i.e., OTU)within its respective communityis provided in parentheses

Appl Microbiol Biotechnol (2010) 88:389–399 395

philic species that are commonly associated with thedecomposition of lignocellulose and municipal solid waste(Burrell et al. 2004; Chynoweth and Pullammanappallil1996), and many are thought to work most effectively whengrown in mixed culture (Demain et al. 2005; Kato et al.2004).

At the laboratory scale, co-cultures of Clostridiumthermocellum with Thermoanaerobacterium thermosac-charolyticum, Thermoanaerobacter ethanolicus, or otherspecies have been used to maximize the hydrolysis andfermentation of lignocellulosic and hemicellulosic feed-stocks (Bàez-Vàsquez and Demain 2008; Demain et al.2005, and references therein). C. thermocellum producescellobiose and cellodextrins from cellulose for its own use,and it participates in the enzymatic saccharification ofhemicelluloses to xylose and xylobiose, substrates that canbe used by the co-cultured xylose-degrader (Demain et al.2005). Such interactions have been demonstrated in anumber of systems, including natural environments, indus-trial settings, and synthetically constructed communities(Demain et al. 2005; Kato et al. 2004; Lynd et al. 2002;Shiratori et al. 2006), and they are considered to representthe model for co-culture cellulose degradation (Bàez-Vàsquez and Demain 2008).

Sorghum is a complex material that contains 29%cellulose, 26% hemicellulose, and 8% lignin on a dryweight basis (Rooney et al. 2007), and as a substrate it

should be able to support of a wide range of metabolisms,including the co-culture model. The results of this studysuggest this to be the case, as the communities associatedwith each of our fermentation temperatures harboredsubsets of OTUs that correlated well with reactor produc-tivity and adhere to the cellulose degrader/non-cellulosedegrader co-culture model (Demain et al. 2005; Kato et al.2004; Lynd et al. 2002). Each community contained two orthree dominant Clostridium-like OTUs that were positivelycorrelated with reactor productivity and are likely to play anintegral role in cellulose degradation within the carboxylateplatform. In complement, each community also harboredtwo or three emergent OTUs that were negatively correlatedwith reactor productivity and possibly represent the xylose-degrading component of the platform. Among this group werea Bacillus-like bacterium and a Clostridium stercorarium-like bacterium, both of which were detected in the middle-and late stages of the thermophilic fermentation. In themesophilic fermentation, this group consisted of taxa thatwere similar to members of the Actinobacteria and Proteo-bacteria (Table 5).

Previous work has demonstrated that the spectrum ofacids produced by the carboxylate platform can be variedby altering temperature conditions (Chan and Holtzapple2003; Fu and Holtzapple 2010), and the divergent patternsof community succession that we observed between the twotemperature treatments may explain this phenomenon.

Table 5 Characteristics of OTUs correlated with productivity in the 40 °C and 55 °C reactor communities

OTU ID Correlation (r2) Closest GenBankmatch

Accessionnumber

Similarity (%) Potential function Reference(s)

40 °C

2 0.992 (+) UnculturedClostridiales

GU214156 98 Cellulose degradation

5 0.999 (+) Clostridiaceaestr 80Wc

AB078860 98 Cellobiose degradation Akasaka et al. (2003)

28 0.987 (+) Clostridiumaminovalericum

NR_029245 96 Cellobiose degradation Hardman and Stadtman (1960)

13 0.825 (–) Actinomycetaceaebacterium

FJ542912 95 Uncertain

27 0.720 (–) Azotobacterchroococcum

EF634040 100 Xylose degradation; possibledegradation of phenoliccompounds

Abd-Alla et al. (1992);Juarez et al. (2005)

55 °C

1 0.996 (+) Unculturedbacterium

AB221356 99 Cellulose degradation Shiratori et al. (2006)

14 0.559 (+) Clostridiumcellulosi

FJ465164 100 Cellulose degradation Yanling et al. (1991)

16 0.871 (–) Clostridiumstercorarium

L09176 96 Cellulose and xylosedegradation

Madden (1983)

70 0.739 (–) Symbiobacteriumsp KA13

AB455239 95 Uncertain

174 0.964 (–) Bacillus sp S13 AF403022 99 Likely xylose degradation

The symbol following the r2 value indicates positive (+) or negative (–) correlation between OTU relative abundance and reactor productivity. Accessionnumbers refer to those belonging to the closest GenBank matches

396 Appl Microbiol Biotechnol (2010) 88:389–399

Although both fermentations were seeded with a commoninoculum and dominated by Clostridia early on, they sharedvery few taxa in common. With time, these differencesincreased, as Bacteroidetes overtook the 40 °C communityand the 55 °C community remained Clostridia-rich. At thesame time, the acid product spectra of the two communitiesalso began to differentiate themselves from one another,resulting in the 40 °C community producing significantlymore propionic acid than the 55 °C community, and the55 °C community producing significantly more butyric acidthan the 40 °C community (Table 2). Members of theBacteroidetes have been identified as xylose degraders thatgenerate propionate as a fermentation product (Chassardet al. 2008; Ren et al. 2007), whereas many Clostridium sppare known cellulose degraders that produce butyrate and/orhydrogen gas as major fermentation products (Ren et al.2007). Thus, it appears that community composition andsuccession patterns, particularly with respect to temperatureand the type of substrate remaining within the reactorsystem, may play an important role in determining productoutcomes between the two fermentation temperatures.

Community divergence may also contribute to thedifferences we observed between the two fermentationtemperatures with respect to productivity. Although both

temperatures resulted in the conversion of similar amountsof biomass, the 40 °C community was significantly moreproductive than the 55 °C community (Table 1). The 40 °Ccommunity also harbored a wider variety of taxa than the55 °C community (Fig. 2). A richer and more diversecommunity is likely to harbor individuals capable offulfilling a wider array of niches, particularly with respectto substrate utilization (Levén et al. 2007), and a morecomplete utilization of substrate within the reactor couldlead to increased productivity. Alternatively, the differencesobserved with respect to productivity may be rooted in thefact that our definition of productivity was limited tocarboxylic acids and did not include other potentialconversion products, such as hydrogen gas or methane.Given the similar conversion rates observed between thetwo fermentation temperatures, the differences that resultedbetween them with respect to productivity, the use ofiodoform to inhibit methanogenesis, and the identities ofdominant taxa that emerged within each community, wespeculate that a substantial portion of the potential productin the 55 °C fermentation may have been lost as hydrogengas. Further work will be needed to confirm whether or notthis is the case and will be a focus of future investigationsof the carboxylate platform.

Azotobacter tropicalis str. BKK.2 (AB236162.1)

Pseudomonas indica (AF302795.1)

Cluster 27 Pseudomonas indica (AF302796.3)

Actinobaculum schallii str. LL (AF487680.1)

Actinobaculum sp. P1 str. P2P 19 (AY207066.1)

Actinobaculum sp. str. BM#101342 (AY282578.1)

Cluster 13

Cluster 70 Symbiobacterium sp. str. KA13 (AB455239.1)

Symbiobacterium thermophilum str. IAM 14863 (AP006840.1)

Clostridium sp. str. rennanqilfy3 (AY363375.1)

Ethanologenbacterium barbinense str. CGMCC1152 (AY833426.1)

Clostridium cellulosi (L09177.1)

Cluster 14 Bacillus sp. str. S13 (AF403022.1)

Cluster 174 Ureibacillus sp. 6T19 (AY850379.1)

Ureibacillus sp. 6T29 (AY850380.1)

Cluster 16 Clostridium stercorarium subsp. leptospartum str. ATCC 35414 (AF266461.1)

Clostridium stercorarium subsp. thermolacticum str. TC21 (L09176.1)

Clostridium stercorarium str. DSM 8532T (AJ310082.1)

Clostridium sp. str. BL-20 (DQ196623.2)

Clostridium sp. CYP8 (DQ479418.1)

Clostridium mesophilum str. SW408 (EU037903.1)

Cluster 2 Clostridium aminovalericum (M23929.1)

Clostridium aminovalericum str. DSM 1238 (X73436.1)

Cluster 5 Clostridium phytofermentans str. ISDg (NZ_AAQT01000017.1)

Clostridium xylanovorans str. HESP1 (AF116920.1)

Clostridium jejuense str. HY-35-12T (AY494606.1)

Cluster 28 Clostridiaceae str. 80Wc (AB078860.1)

Cluster 1 Clostridium lavalense str. CCRI-9842 (Ef564277.1)

Escherichia coli (U00096)

65100

100

67

76

100

100

86

100

65100

100

100

100

62

61

100

100

61

100

79

100

9499

66

92

91

9499

Fig 4 Neighbor-joining phylo-genetic tree of the ten OTUs thatwere found to share strongcorrelations with reactorproductivity rates and theirnearest-neighbor isolates fromthe Greengenes chimera-checked database. Bootstrapvalues are from 100 replicatesand values <50 are notpresented. The symbol adjacentto each OTU entry indicates itsrelationship (triangle positivecorrelation, square negativecorrelation) with reactorproductivity, and the color of thetext indicated the communitywith which each OTU wasaffiliated (blue 40 °C, red 55 °C)

Appl Microbiol Biotechnol (2010) 88:389–399 397

This study provides a first look into the “black box” thatunderlies the MixAlco carboxylate platform for biomassconversion. Although the 16S rRNA surveys described herehave provided useful insight regarding the composition ofour reactor communities, the effects that temperature haveupon them, and the ways in which they change over time,we have only just begun to shed light on the microbi-ology of this system. Additional characterizations of thefunctional genes that are active within this system, andits metagenome at large, are needed to better understandthe microbiology of the carboxylate platform. It isanticipated that such information will provide tremen-dous strides forward in the optimization of this platformand enhance efforts to understand and improve otherbiomass conversion systems.

Acknowledgments This work was supported by grants from theDepartment of Energy’s Joint Genome Institute, the Texas AgriLifeResearch Bioenergy Program, and the Texas A&M University Officeof the Vice President for Research Energy Resources Program. Thiswork was performed under the auspices of the US Department ofEnergy’s Office of Science, Biological and Environmental ResearchProgram, and by the University of California, Lawrence BerkeleyNational Laboratory under contract No. DE-AC02-05CH11231,Lawrence Livermore National Laboratory under Contract No. DE-AC52-07NA27344, and Los Alamos National Laboratory undercontract No. DE-AC02-06NA25396. The authors would like toacknowledge Mukul Sherekar and Heidi Mjelde for their assistancein the lab; Kanwar Singh for library construction and sequencing; theSorghum Breeding and Genetics Program at Texas A&M Universityfor providing the sorghum that was used as the bioreactor feedstockfor this study; and two anonymous reviewers whose comments helpedto improve this manuscript.

References

Abd-Alla MH, Omar SA, Abdel-Wahab AM (1992) The role ofcellulose-decomposing fungi in nitrogenase activity of Azotobac-ter chroococcum. Folia Microbiol 37:215–218

Aiello-Mazzarri C, Agbogbo FK, Holtzapple MT (2006) Conversionof municipal solid waste to carboxylic acids using a mixedculture of mesophilic microorganisms. Biores Technol 97:47–56

Akasaka H, Izawa T, Ueki K, Ueki A (2003) Phylogeny ofnumerically abundant culturable anaerobic bacteria associatedwith degradation of rice plant residue in Japanese paddy fieldsoil. FEMS Microbiol Ecol 43:149–161

Bàez-Vàsquez MA, Demain AL (2008) Ethanol, biomass, andClostridia. In: Wall JD, Harwood CS, Demain A (eds) Bioenergy.ASM, Washington, DC, pp 49–54

Burrell PC, O’Sullivan C, Song H, Clarke WP, Blackall LL (2004)Identification, detection, and spatial resolution of Clostridiumpopulations responsible for cellulose degradation in a methano-genic landfill leachate bioreactor. Appl Environ Microbiol70:2414–2419

Chan WN, Holtzapple MT (2003) Conversion of municipal solidwastes to carboxylic acids by thermophilic fermentation. ApplBiochem Biotechnol 111:93–112

Chang VS, Nagwani M, Holtzapple MT (1998) Lime pretreatment ofcrop residues bagasse and wheat straw. Appl Biochem Biotech-nol 74:135–159

Chassard C, Delmas E, Lawson PA, Bernalier-Donadille A (2008)Bacteroides xylanisolvens sp. nov., a xylan-degrading bacteriumisolated from human faeces. Int J Syst Evol Microbiol 58:1008–1013

Chynoweth DP, Pullammanappallil P (1996) Anaerobic digestion ofmunicipal solid wastes. In: Palmisano AC, Barlaz MA (eds)Microbiology of solid wastes. CRC, Boca Raton, pp 74–114

Cole JR, Wang Q, Cardenas E, Fish J, Chai B, Farris RJ, Kulam-Syed-Mohideen AS, McGarrell DM, Marsh T, Garrity GM, Tiedje JM(2009) The ribosomal database project: improved alignments andnew tools for rRNA analysis. Nuc Acids Res 37:D141–D145

Demain AL, Newcomb M, Wu JHD (2005) Cellulase, clostridia, andethanol. Microbiol Mol Biol Rev 69:124–154

DeSantis TZ, Hugenholtz P, Larsen N, Rojas M, Brodie EL, Keller K,Huber T, Dalevi D, Hu P, Andersen GL (2006) Greengenes, achimera-checked 16S rRNA gene database and workbenchcompatible with ARB. Appl Environ Microbiol 72:5069–5072

Engelbrektson A, Kunin V, Wrighton KC, Zvenigorodsky N, Chen F,Ochman H, Hugenholtz P (2010) Experimental factors affectingPCR-based estimates of microbial species richness and evenness.ISME J 4:642–647

Felsenstein J (2005) PHYLIP (Phylogeny Inference Package) version3.6. Distributed by author. Department of Genome Sciences,University of Washington, Seattle

Fernandez A, Huang SY, Seston S, Xing J, Hickey R, Criddle C,Tiedje J (1999) How stable is stable? Function versus communitycomposition. Appl Environ Microbiol 65:3697–3704

Fu Z, Holtzapple MT (2010) Fermentation of sugarcane bagasse andchicken manure to calcium carboxylates under thermophillicconditions. Appl Biochem Biotechnol 162:561–578

Goberna M, Insam H, Franke-Whittle IH (2009) Effect of biowaste sludgematuration on the diversity of thermophilic Bacteria and Archaea inan anaerobic reactor. Appl Environ Microbiol 75:2566–2572

Granda CB, Holtzapple MT (2008) Biorefineries for solvents: TheMixAlco process. In: Wall JD, Harwood CS, Demain A (eds)Bioenergy. ASM, Washington, DC, pp 347–360

Granda CB, Holtzapple MT, Luce G, Searcy K, Mamrosh DL (2009)Carboxylate platform: the MixAlco process part 2: processeconomics. Appl Biochem Biotechnol 156:537–554

Hammer O, Harper DAT, Ryan PD (2001) PAST: paleontologicalstatistics software package for education and data analysis.Palaeontol Electronica 4:9

Hardman JK, Stadtman TC (1960) Metabolism of omega-amino acids:2. Fermentation of delta-aminovaleric acid by Clostridiumaminovalericum n sp. J Bacteriol 79:549–552

Hollister EB, Engledow AS, Hammett AM, Provin TL, WilkinsonHH, Gentry TJ (2010) Shifts in microbial community structurealong an ecological gradient of hypersaline soils and sediments.ISME J 4:829–838

Holtzapple MT, Davison RR, Ross MK, Aldrett-Lee S, Nagwani M,Lee CM, Lee C, Adelson S, Kaar W, Gaskin D, Shirage H,Chang NS, Chang VS, Loescher ME (1999) Biomass conversionto mixed alcohol fuels using the MixAlco process. Appl BiochemBiotechnol 77–9:609–631

Juarez B, Martinez-Toledo MV, Gonzalez-Lopez J (2005) Growth ofAzotobacter chroococcum in chemically defined media contain-ing p-hydroxybenzoic acid and protocatechuic acid. Chemo-sphere 59:1361–1365

Kato S, Haruta S, Cui ZJ, Ishii M, Igarashi Y (2004) Effectivecellulose degradation by a mixed-culture system composed of acellulolytic Clostridium and aerobic non-cellulolytic bacteria.FEMS Microbiol Ecol 51:133–142

Klocke M, Mähnert P, Mundt K, Souidi K, Linke B (2007) Microbialcommunity analysis of a biogas-producing completely stirredtank reactor fed continuously with fodder beet silage as mono-substrate. Syst Appl Microbiol 30:139–151

398 Appl Microbiol Biotechnol (2010) 88:389–399

Kunin V, Engelbrektson A, Ochman H, Hugenholtz P (2010)Wrinkles in the rare biosphere: pyrosequencing errors can leadto artificial inflation of diversity estimates. Environ Microbiol12:118–123

Lee ZM-P, Bussema C III, Schmidt TM (2009) rrnDB: documentingthe number of rRNA and tRNA genes in Bacteria and Archaea.Nuc Acids Res 37:D489–D493

Levén L, Eriksson ARB, Schnürer A (2007) Effect of processtemperature on bacterial and archaeal communities in twomethanogenic bioreactors treating organic household waste.FEMS Microbiol Ecol 59:683–693

Lin Y, Tanaka S (2006) Ethanol fermentation from biomass resources:current state and prospects. Appl Microbiol Biotechnol 69:627–642

Lynd LR, Weimer PJ, van Zyl WH, Pretorius IS (2002) Microbialcellulose utilization: fundamentals and biotechnology. MicrobiolMol Biol Rev 66:506–577

Madden RH (1983) Isolation and characterization of Clostridiumstercorarium sp. nov., cellulolytic thermophile. Int J SystBacteriol 33:837–840

Perlack RD, Wright LL, Turhollow AF, Graham RL, Stokes BJ,Erbach DC (2005) Biomass as feedstock for a bioenergy andbioproducts industry: the technical feasibility of a billion-tonannual supply DOE/GO-102005– 2135. Oak Ridge NationalLaboratory, Oak Ridge, TN. (http://feedstockreview.ornl.gov/pdf/billion_ton_vision.pdf)

Ren N, Xing D, Rittmann BE, Zhao L, Xie T, Zhao X (2007)Microbial community structure of ethanol type fermentation inbio-hydrogen production. Environ Microbiol 9:1112–1125

Rooney WL, Blumenthal J, Bean B, Mullet JE (2007) Designingsorghum as a dedicated bioenergy feedstock. Biofuels, BioprodBiorefin 1:147–157

Rubin EM (2008) Genomics of cellulosic biofuels. Nature 454:841–845Schloss PD, Westcott SL, Ryabin T, Hall JR, Hartmann M, Hollister EB,

Lesniewski RA, Oakley BB, Parks DH, Robinson CJ, Sahl JW, StresB, Thallinger GG, Van Horn DJ, Weber CF (2009) IntroducingMothur: open-source, platform-independent, community-supportedsoftware for describing and comparing microbial communities. ApplEnviron Microbiol 75:7537–7541

Shin H-S, Youn J-H, Kim S-H (2004) Hydrogen production from foodwaste in anaerobic mesophilic and thermophilic acidogenesis. IntJ Hydrogen Energy 29:1355–1363

Shiratori H, Ikeno H, Ayame S, Kataoka N, Miya A, Hosono K, Beppu T,Ueda K (2006) Isolation and characterization of a new Clostridiumsp. that performs effective cellulosic waste digestion in athermophilic methanogenic bioreactor. Appl Environ Microbiol72:3702–3709

Thanakoses P, Mostafa NAA, Holtzapple MT (2003) Conversion ofsugarcane bagasse to carboxylic acids using a mixed culture ofmesophilic microorganisms. Appl Biochem Biotechnol 105:523–546

von Wintzingerode F, Göbel UB, Stackebrandt E (1997) Determina-tion of microbial diversity in environmental samples: pitfalls ofPCR-based rRNA analysis. FEMS Microbiol Rev 21:213–229

Yanling H, Youfang D, Yanquan L (1991) Two cellulolytic Clostrid-ium species: Clostridium cellulosi sp. nov. and Clostridiumcellulofermentans sp. nov. Int J Syst Bacteriol 41:306–309

Yue JC, Clayton MK (2005) A similarity measure based on speciesproportions. Commun Stat Part ATheory Methods 34:2123–2131

Appl Microbiol Biotechnol (2010) 88:389–399 399