Functional Diversity of Carbohydrate-Active Enzymes Enabling a Bacterium to Ferment Plant Biomass

Summary

Plant-fermenting bacteria are important for the global carbon cycle, human nutrition, and industrial production of renewable fuels and commodities from cellulosic biomass. Plants are primarily composed of heterogeneous polysaccharides, requiring plant-degrading microbes to encode many carbohydrate-active enzymes (CAZymes) to cleave different sugar linkages. Here we develop a broadly applicable method to study how microbes catabolize plant biomass by determining the combination of CAZymes that depolymerize each polysaccharide into sugars, how the cell alters global mRNA expression, and the efficiency with which each polysaccharide is metabolized. We apply this method to investigate how Clostridium phytofermentans, a bacterium encoding 171 CAZymes, ferments polysaccharides. We assimilate our results into a genetic model of how this bacterium metabolizes plant biomass and discuss how these results further our understanding of microbial plant fermentation.

Introduction

Plants annually produce 200 billion tons of lignocellulosic biomass [1], which is metabolized by specialized microbes in diverse environments. For instance, recycling of plant biomass by soil [2] and marine [3] microbes is a key part of the global carbon cycle and intestinal bacteria ferment indigestible plant fiber to short chain fatty acids that constitute 60–85% of calories in ruminants and 5–10% in humans [4]. Further, as only 2% of cellulosic biomass is currently used by humans [5], it is a vast potential feedstock that industrial microbes could convert into energy and commodities. Elucidating how microbes depolymerize and metabolize plant biomass is thus important to understand carbon flow in the environment, to promote healthy human nutrition and prevent disease [6], and to develop industrial processes based on cellulosic bioconversion.

Most of plant biomass is in the cell wall, a macromolecular network of phenolic lignin and three types of polysaccharides (cellulose, hemicelluloses, and pectins) whose relative abundances vary widely among species and tissues (Table S1). The load bearing structure of the cell wall consists of cellulose fibrils tethered by various types of hemicellulose. Hemicellulose is enriched in xylan [7] and xyloglucan [8] in dicots, arabinoxylan in monocots [9], and galacto- and glucomannans in gymnosperms [10]. Outside the cell wall, mannans also act as storage polysaccharides in seeds [11], similar to starch. Pectins are cross-linked galacturonic acid-based polysaccharides that act in cellular adhesion and primary wall extension. More than 60% of pectin is often homogalacturonan (HG) [12], which is esterified with methanol to various degrees. Rhamnogalacturonan I (RGI) [13], the second most abundant pectin, can have galactan and arabinan side chains on the rhamnose residues [14]. Because plant tissues are composed of such heterogeneous polysaccharides, plant-degrading microbes express a myriad of carbohydrate-active enzymes (CAZymes) [15], each of which modifies or cleaves a specific type of sugar linkage.

Here we demonstrate a strategy for systematic analysis of the enzymatic machinery used by microbes to degrade and metabolize plant polysaccharides. Among these microbes, the plant-fermenting clostridia are of particular interest for being a dominant group in the human gut microbiome [6] and top candidates to transform cellulosic biomass into fuels and commodities [16], [17]. We studied Clostridium phytofermentans [18], a soil bacterium with 171 CAZyme-encoding genes (Table S2) including 116 glycoside hydrolases in 44 different families. We first quantified growth on comprehensive panel of plant polysaccharides and sugars (Table S3). Strand-specific RNA sequencing revealed all genes whose expression changed on the various substrates. In particular, we focused on up-regulated CAZyme genes and determined how they are organized into regulons that respond to specific polysaccharides. A set of 56 up-regulated CAZymes were cloned, purified, and an “each enzyme versus each substrate” screen quantified their abilities to bind and cleave plant polysaccharides. These enzymes were then tested in combination to identify synergies for polysaccharide degradation. We discuss how the results can be integrated to further our knowledge of how microbes metabolize plant biomass.

Results/Discussion

Growth on polysaccharides and sugars

We developed a high resolution, microtiter anaerobic growth assay that shows C. phytofermentans ferments diverse plant polysaccharides (Fig. 1) and their constituent monosaccharides (Fig. S1), but with widely varying cell yields and growth rates (Table S4). It also forms colonies on solid medium containing each polysaccharide except arabinogalactan II (AGII) (Fig. S2). Growth was fastest on HG (Fig. 1A, generation time 0.70h), similar to rumen microbes that digest pectin more rapidly than cellulose and hemicellulose [19]. Although C. phytofermentans ferments both galacturonic acid (Fig. S1F) and rhamnose (Fig. S1H), cell yield was low on RGI (Fig. 1B). C. phytofermentans grows well on galactan (Fig. 1C), xylans (Fig. 1F–G), mannans (Fig. 1H–I), xyloglucan (Fig. 1J), and starch (Fig. 1L). Limited growth on AGII (Fig. 1E) relative to galactan supports that C. phytofermentans cleaves β-1,4 galactan, but not the β-1,3 and β-1,6-galactose bonds in AGII. Poor growth on arabinan (Fig. 1D) is similar to arabinose (Fig. S1G), suggesting this sugar is transported or metabolized inefficiently. C. phytofermentans grows well on cellulose plates (Fig. S2) and solubilizes cellulosic substrates such as filter paper and raw corn stover (Fig. S3), but weak growth on carboxymethylcellulose (CMC) might result from either lack of a suitable endoglucanase or carboxymethyl side groups inhibiting its metabolism.

C. phytofermentans shows diauxic growth on the mixed sugar polysaccharides galactomannan (Fig. 1I) and xyloglucan (Fig. 1J). For each of these substrates, one of the component sugars (galactose or glucose) supports faster growth than the other (mannose or xylose) (Fig. S1, Table S4). Growth on various mixtures of galactose/mannose (Fig. S4) and of glucose/xylose (Fig. S5) shows rapid metabolism of the preferred sugar followed by slower growth on the other one. However, in both cases when the favored sugar reached 75% of the total, the other sugar does not appear to be metabolized.

Similar to some ruminal [19] and human gut microbes [20], C. phytofermentans often grows faster on polysaccharides than the constituent sugars (Table S4). When presented with mixtures of xylan and xylose, this bacterium shows diauxic growth with preferential metabolism of xylan (Fig. S6), which is surprising because xylan must be cleaved to xylose to be metabolized. Growth on polysaccharides could be energetically favorable if significant ATP is saved by simultaneous transport of multiple sugar units in a single oligosaccharide [21] or by intracellular phosphorolysis of oligosaccharides [22]. C. phytofermentans encodes at least a dozen phosphorylases [23] [24], which cleave oligosaccharides without using ATP. Although the mechanisms regulating sugar metabolism in C. phytofermentans are unknown, diauxic growth supports carbon catabolite repression prioritizes growth on preferred sugars and polysaccharides.

Gene expression

We quantified mRNA expression by strand-specific RNA sequencing during log-phase growth on 8 polysaccharides, 3 monosaccharides, and raw corn stover as a complex biomass substrate. An average of 17.3 million mRNA reads were mapped per sample (Table S5), yielding expression (RPKM) values (Table S6) that were highly correlated (r² = 0.96–0.99) between duplicate cultures for all conditions (Fig. S7). The reads were also highly strand-specific (Fig. S8), which will facilitate their future use for de novo transcriptome assembly, gene annotation and detection of antisense transcription. The fraction of reads mapping to CAZymes during growth on glucose was 2.0%, but this increased greatly on polysaccharides, especially cellulose (11.9%) and stover (31.0%). We assessed which genes were significantly differentially expressed on each polysaccharide relative to glucose using DESeq [25] (Table S7). Expression of CAZyme genes on polysaccharides relative to glucose (Fig. 2) shows that between 15 (cellobiose) and 40 (stover) CAZymes were significantly up-regulated per treatment (Table S8) with a total of 92 CAZymes up-regulated on at least one polysaccharide.

The differentially expressed CAZymes are putatively classified by the CAZy database as 67 glycoside hydrolases, 6 carbohydrate esterases, 4 polysaccharide lyases, 14 glycosyl transferases, and 2 CBM proteins. We analyzed the specificity of the CAZyme transcriptional response by K-means clustering the expression profiles of these genes (Fig. 3, Table S9). Cluster A consists six genes that were highly up-regulated on multiple substrates: the GH26 cphy1071, the GH11 cphy2105, two GH18 chitinases cphy1799 and cphy1800 [26], the GH9 cellulase cphy3367 [27] [28] and the GH48 cellulase cphy3368 [29]. Clusters B–F respond to specific polysaccharides such as homogalacturonan (clusters B,C), starch (cluster D), xylan (cluster E), and cellulose/arabinan (cluster F). C. phytofermentans thus perceives signals from individual polysaccharides and responds by up-regulating specific transcriptional regulons that enable it to tailor its complement of CAZymes to the polysaccharide substrate.

CAZyme activities

A set of 56 CAZymes up-regulated on polysaccharides were His-tagged, overexpressed, purified, and their abilities to bind and cleave polysaccharides were quantified. The CAZy database classifies these enzymes putatively as 47 glycoside hydrolases, 4 polysaccharide lyases, and 4 carbohydrate esterases (Table S2); putative glycosyltransferases were not examined as they are not involved in polysaccharide catabolism [6]. Thirty-two enzymes have significant cleavage or binding activities (Fig. 4, Table S11). Some substrates such as β-1,4-galactan appear to be cut by a single, highly active enzyme, while multiple CAZymes from the same family degrade other substrates such as xylan (GH10), mannan (GH26), starch (GH13), and HG (PL9). CAZymes from multiple families together depolymerize substrates such as xyloglucan (GH2,5,12,31), glucomannan (GH5,GH9,GH26) and galactomannan (GH5,GH26).

We found 15 CAZymes that bind insoluble polysaccharides, most commonly cellulose and mannan (Fig. 4B). Unexpectedly, no CAZyme bound corn stover, suggesting that partial digestion of raw biomass is needed to facilitate enzyme binding. Nine enzymes that bound substrates have carbohydrate binding modules (CBM), but some enzymes such as the cellulase Cphy1163 can bind their substrate without one. While CBM are known to discriminate between polysaccharides such as cellulose and mannan [30], we observed overlap with cellulase CBMs binding mannan and vice versa. Further, CBM from xylanases can bind cellulose and mannose, but with lower affinity, showing that CBM often bind a range of polysaccharides. Consistent with their cleavage activities, GH13 were the only enzymes to bind starch. Enzymes with CBM usually also have catalytic modules, but Cphy1713, a CAZyme with a CBM32 and no catalytic module, binds galactomannan. CBM32 are known to bind galactose and this protein may function similar to one in Yersinia that is proposed to bind oligosaccharides to prevent them from leaking out of the cell [31].

Thirty-two CAZy families have multiple members, which often have divergent cleavage activities and cellular localizations. Cphy1510 has the highest activity among the four GH10 active on xylan (Fig. 5A). Cphy3010, the GH10 with lowest activity, is the only one lacking a secretion signal, supporting it acts intracellularly on xylo-oligosaccharides while the other GH10 are extracellular. Members of the GH5 family act on a wide range of polysaccharides [32]. C. phytofermentans encodes 3 GH5 enzymes, among which one is active on galactomannan and two on xyloglucan (Fig. 5B). The GH5 Cphy1163 has no activity on either of these substrates, but is the most active on cellulose and glucomannan. The 3 GH26 also vary in substrate specificities (Fig. 5C); all the GH26 are similarly active on β-mannan, but only Cphy1071 has cellulase activity and it has lower activity on gluco- and galactomannan. Sequenced-based families are thus useful to make general substrate predictions for CAZymes, but experiments are needed to determine substrate range and catalytic efficiency.

CAZymes mixtures can degrade polysaccharides more efficiently than individual enzymes. We assessed pairwise interactions between each CAZyme and a second enzyme on cellulose (Cphy3367), xylan (Cphy2105), glucomannan (Cphy1071), and homogalacturonan (Cphy1687) (Fig. 5D–G). Similar to results showing synergy between the GH9 Cphy3367 and a B. subtilis GH5 [33], we found that a mix of Cphy3367 and the GH5 Cphy1163 has higher activity on cellulose than either enzyme alone (Fig. 5D), supporting they have complementary roles in cellulolysis. CAZymes can also potentiate other enzymes that have no activity by themselves. For example, the xylanase Cphy2105 activates the putative xylosidases Cphy3009 and Cphy3207 on xylan (Fig. 5E). Similarly, Cphy1071 activates the putative mannosidase Cphy1719 on glucomannan (Fig. 5F). Activities of the GH28 Cphy2567 and Cphy3310 are enhanced by the carbohydrate esterase Cphy1687 (Fig. 5G), supporting this enzyme demethylesterifies homogalacturonan to facilitate its degradation. This carbohydrate esterase did not, however, increase cleavage by the PL9 enzymes that were the most active on homogalacturonan.

Global correlations between CAZyme mRNA expression and cleavage activities were weak for all polysaccharides (Fig. S12), mostly because many CAZyme genes are up-regulated on substrates upon which they have no activity. CAZymes up-regulated on multiple substrates (Fig. 3, cluster A) may act as ‘carbon scouts’ [34] that degrade complex substrates into inducing molecules used to fine-tune the expression of hydrolytic enzymes. As described above, some CAZymes such as xylosidases (Fig. 5D) are inactive on intact xylan, but are potentiated by other xylanases. The GH18 Cphy1799 and Cphy1800 are the most highly-upregulated CAZymes on cellulose (Fig. S12), but are chitinases with no activity on cellulose or other plant substrates [26]. As such, the set of up-regulated CAZymes is useful to identify active enzymes, but strong up-regulation does not necessarily indicate activity on a given substrate.

Conclusions

We assimilated our results into a model of C. phytofermentans polysaccharide catabolism that shows degradation by active CAZymes and uses mRNA expression profiles to predict how these substrates are transported and metabolized (Fig. 6). Unlike other clostridia that transport sugars with numerous phosphotransferase systems (PTS) [35] [36], C. phytofermentans encodes a single, lowly expressed PTS and also lacks the symporters to transport xylose and arabinose [37]. Instead, C. phytofermentans responds to carbon sources by up-regulating between two (galacturonic acid) and twenty-two (arabinan) ABC transporters (Fig. 6). Expression changes support that oligosaccharides and monosaccharides are uptaken by distinct transporters. For example, different ABC transporters are up-regulated on xylose and xylan. Similarly, different transporters respond to glucose, cellobiose, and cellulose. Intracellular cellodextrins are cleaved by at least one cellodextin phosphorylase (GH94); hexoses are phosphorylated, likely by a ROK hexokinase (Cphy0329) and a putative galactokinase (Cphy2237), and fed into glycolysis. While hexokinases may have wide substrate activity [38], poor growth on mannose could be due to inefficient mannose phosphorylation. The pentoses xylose and arabinose are isomerized and metabolized by the pentose phosphate pathway (PPP). Weak growth on arabinose could be due to inefficient transport or the lack of the phosphoketolase in the PPP enabling rapid L-arabinose metabolism by C. acetobutylicum [39].

Plant degrading microbes differ widely in their abilities to depolymerize and metabolize polysaccharides, likely reflecting niche differentiation to alleviate resource competition. Among soil clostridia, C. thermocellum ferments cellulose, but not xylan [40]. C. cellulolyticum grows faster on xylose than xylan and faster on cellobiose than glucose [41], both of which differ from C. phytofermentans. Similar specialization exists in the human gut microbiome where microbes catabolize different glycans in dietary fiber [20]. The strategy presented here of high-resolution anaerobic growth measurements, RNA sequencing, and CAZyme assays complements other methods such as proteomics [42] and metagenomics. Elucidating how microbes metabolize polysaccharides is key to understanding the function of plant-degrading microbial communities and to develop improved enzyme mixtures and recombinant microbes for industrial processing of plant biomass.

Materials and Methods

Growth measurements

C. phytofermentans was cultured anaerobically at 30°C in GS2 medium [43]. Growth was quantified in medium containing 3 g l⁻¹ mono-, di-, or polysaccharides (Table S3 for product numbers) in 100-well microtiter plates (Bioscreen 9502550) that were sealed by press-fitting adhesive sheets (Qiagen 1018104) under the lids. Growth was measured every 15 minutes as OD₆₀₀ at 30°C using a Thermo Scientific Bioscreen C with 30 seconds shaking before each reading. Growth was not measured for ß-mannan and amorphous cellulose cultures because these low solubility substrates occluded optical density measurements. Growth on insoluble substrates (15 g l⁻¹ of 0.5×5 cm strips of filter paper cellulose or raw corn stover) was measured in 10 ml cultures as substrate consumption by collection on 11 µm filters by vacuum filtration, drying overnight at 65°C, and weighing. Growth on solid GS2 medium with 3 g l⁻¹ polysaccharide and 15 g l⁻¹ agar was tested by incubating plates anaerobically for 10 days at 30°C (Fig. S2). Regenerated amorphous cellulose (RAC) for GS2 agar plates and enzyme assays was prepared from Avicel PH-101 by phosphoric acid treatment [44].

RNA-seq

Cultures containing soluble substrates were sampled for RNA-seq in mid-log phase. Filter paper cellulose cultures were sampled under the same conditions as in [45]; corn stover cultures were sampled after 3 days. Cells were arrested with RNA stabilization buffer, collected by centrifugation (5 kg, 5 min, 4°C), and RNA was isolated with an Ambion Ribopure Bacteria Kit (AM1925). Twenty µg total RNA was treated with Turbo DNase (Invitrogen AM1907), phenol-chloroform extracted, ethanol precipitated, and resuspended in 15 µl 10 mM Tris-Cl, pH 8.5. Ribosomal RNA was depleted from 10 µg RNA using a MICROBExpress kit (Invitrogen AM1905), giving a typical yield of 1–2 µg RNA. 500 ng of rRNA-depleted RNA was fragmented with magnesium buffer (NEB E6101) for 2.5 minutes at 94°C, ethanol precipitated, and resuspended in 10 µl water. RNA was resolved on an Agilent Bioanalyzer 2100 to confirm it was 200–800 bp.

Single stranded cDNA was made from 500 ng fragmented RNA by Superscript III first strand synthesis (Invitrogen 18080-051) including 200 ng/µl actinomycin D (Invitrogen 11805017) and 120 ng/µl random hexamers (Invitrogen 48190-011). Single stranded DNA was twice phenol-chloroform extracted, ethanol precipitated, and resuspended in 52 µl water. The second cDNA strand was synthesized by the Invitrogen double stranded cDNA synthesis kit (Invitrogen 11917-010) using 250 µM dNTP (dATP, dCTP, dGTP, dUTP). Double stranded cDNA was purified using 1.8 volumes of Solid Phase Reversible Immobilisation (SPRI) beads [46] prepared from carboxylate-modified microparticles [47] (ThermoScientific 6515-2105-050250) and resuspended in 50 µl 10 mM Tris-Cl, pH 8.5. DNA was resolved on an Agilent Bioanalyzer to confirm cDNA was 100–500 bp.

Sequencing libraries were prepared from ∼250 ng cDNA using the Illumina Truseq DNA Kit (Illumina FC-121-2001). DNA was purified with 1.8 volumes of SPRI beads after end repair. Following adapter ligation, 200–500 bp DNA was selected by Pippin Prep gel extraction (Sage Science). The second cDNA strand (contains dUTP) was then degraded by incubating with 1 unit USER enzyme (NEB NEB M5505S) at 37°C for 15 minutes, followed by 5 minutes at 95°C. Single stranded DNA was PCR amplified and size selected with 1.8 volumes of SPRI beads. Multiplexed libraries were normalized to 2 ng µl⁻¹, pooled, and resolved on an Agilent Bioanalyzer to confirm the peak was at 200–500 bp. Libraries were sequenced on a Illumina HiSeq2000 sequencer. RNA-seq reads were mapped to the C. phytofermentans genome (NCBI NC_010001.1) using Bowtie [48] to report the single, best alignment (see Table S5 for parameters). Gene expression was measured as RPKM (sum of nucleotides in gene per million mapped nucleotides per gene kb) using RSEQtools [49] (Table S6), excluding rRNA reads from RPKM calculations. Differentially expressed genes (Tables S7, S8) were identified by DESeq [25] using the statistical thresholds defined in [50] of >4-fold expression and a p-value<10⁻⁵, which corresponds to a p-value<0.05 adjusted for multiple testing of the 3902 genes in C. phytofermentans genome by Bonferroni correction.

Enzyme purification and activities

A set of 56 CAZymes (Table S2) that were up-regulated on plant polysaccharides were His-tagged, overexpressed, and purified. Primers (Table S10) were designed to clone CAZyme genes by Ligation-Independent Cloning [51] into pET-22B(+). Genes were cloned with C-terminal His-tags except for cphy1687 and cphy2105, which were cloned with N-terminal tags to improve yields. The 26 enzymes predicted to be secreted by SignalP [52] were cloned as their mature forms, omitting the N-terminal secretion signals. Forward and reverse sequences of the 56 genes cloned in pET-22B(+) were confirmed by sequencing (Fig. S9). Plasmids were transformed into E. coli BL21(DE3) (Novagen 70235), grown in 50 ml TB medium to OD₆₀₀ = 1.0, and induced by adding 500 µM IPTG and incubating overnight at 20°C. Cells were pelleted, resuspended in lysis buffer: 50 mM phosphate buffer pH = 8, 0.5M NaCl, 10 mM imidazole, 15% glycerol, 1 mM pefabloc (Sigma 76307). Cells were lysed by sonication (Cole-Parmer Vibracell CV33) in the presence of lysozyme (Novagen 71230). His-tagged proteins were purified from 50 ml culture on Ni-NTA spin columns (Qiagen 31014) and quantified by Bradford assay, giving an average yield of 3 mg protein. Purified proteins were visualized on 12% SDS-PAGE gels (Novex 12% Bis-Tris Gel NP0342BOX) (Fig. S10).

Enzymatic polysaccharide cleavage was quantified by 3,5-dinitrosalicylic acid (DNS) assay [53] in 96 well clear, flat bottom plates (Molecular Devices X6011). Fifty µg enzyme was incubated with 0.25% polysaccharide in 25 mM potassium phosphate buffer pH = 7 for 30 minutes at 37°C, 1 volume DNS reagent was added, incubated at 95°C for 15 minutes, and read at OD₆₀₀. Enzyme activities (nmol reducing sugar per mg enzyme per min) were calculated from DNS readings by subtracting blanks (+polysaccharide,-enzyme) and calculating sugar produced using standard curves (Fig. S11). Polysaccharide cleavage by enzyme pairs was tested on amorphous cellulose, xylan, glucomannan, and homogalacturonan as the reducing sugar (mM) produced by 25 µg of each enzyme alone and combined with a second enzyme (25 µg Cphy3367 for cellulose, 5 µg Cphy2105 for xylan, 5 µg Cphy1071 for glucomannan, 25 µg Cphy1687 for homogalacturonan). Reducing sugar was measured by DNS assay after 30 minutes incubation as described above. Enzyme binding to insoluble polysaccharides was quantified using a method adapted from [54]. Fifty µg enzyme was suspended in 200 µl 0.1 M phosphate buffer pH = 7 with 6 mg polysaccharide and incubated with rotation for 5 h at 4°C. The polysaccharide was collected by centrifuging at 12 kg for 5 min and the enzyme concentration in the supernatant was calculated before and after incubation by Bradford assay, yielding the percentage of enzyme bound to the polysaccharide.

Supporting Information

Attachment 1

Attachment 2

Attachment 3

Attachment 4

Attachment 5

Attachment 6

Attachment 7

Attachment 8

Attachment 9

Attachment 10

Attachment 11

Attachment 12

Attachment 13

Attachment 14

Attachment 15

Attachment 16

Attachment 17

Attachment 18

Attachment 19

Attachment 20

Attachment 21

Attachment 22

Attachment 23