Biochemical and Regulatory Analyses of Xylanolytic Regulons in Caldicellulosiruptor bescii Reveal Genus-Wide Features of Hemicellulose Utilization

ABSTRACT

Caldicellulosiruptor species scavenge carbohydrates from runoff containing plant biomass that enters hot springs and from grasses that grow in more moderate parts of thermal features. While only a few Caldicellulosiruptor species can degrade cellulose, all known species are hemicellulolytic. The most well-characterized species, Caldicellulosiruptor bescii, decentralizes its hemicellulase inventory across five different genomic loci and two isolated genes. Transcriptomic analyses, comparative genomics, and enzymatic characterization were utilized to assign functional roles and determine the relative importance of its six putative endoxylanases (five glycoside hydrolase family 10 [GH10] enzymes and one GH11 enzyme) and two putative exoxylanases (one GH39 and one GH3) in C. bescii. Two genus-wide conserved xylanases, C. bescii XynA (GH10) and C. bescii Xyl3A (GH3), had the highest levels of sugar release on oat spelt xylan, were in the top 10% of all genes transcribed by C. bescii, and were highly induced on xylan compared to cellulose. This indicates that a minimal set of enzymes are used to drive xylan degradation in the genus Caldicellulosiruptor, complemented by hemicellulolytic inventories that are tuned to specific forms of hemicellulose in available plant biomasses. To this point, synergism studies revealed that the pairing of specific GH family proteins (GH3, -11, and -39) with C. bescii GH10 proteins released more sugar in vitro than mixtures containing five different GH10 proteins. Overall, this work demonstrates the essential requirements for Caldicellulosiruptor to degrade various forms of xylan and the differences in species genomic inventories that are tuned for survival in unique biotopes with variable lignocellulosic substrates.

IMPORTANCE Microbial deconstruction of lignocellulose for the production of biofuels and chemicals requires the hydrolysis of heterogeneous hemicelluloses to access the microcrystalline cellulose portion. This work extends previous in vivo and in vitro efforts to characterize hemicellulose utilization by integrating genomic reconstruction, transcriptomic data, operon structures, and biochemical characteristics of key enzymes to understand the deployment and functionality of hemicellulases by the extreme thermophile Caldicellulosiruptor bescii. Furthermore, comparative genomics of the genus revealed both conserved and divergent mechanisms for hemicellulose utilization across the 15 sequenced species, thereby paving the way to connecting functional enzyme characterization with metabolic engineering efforts to enhance lignocellulose conversion.

INTRODUCTION

Direct microbial conversion of lignocellulose to industrial chemicals depends on the effective deconstruction and utilization of both hemicellulose and cellulose (1). The most attractive sources of lignocellulose for industrial chemical production are those that minimize the impact on the food supply and land use; examples include energy crops like poplar (Populus sp.) and switchgrass (Panicum virgatum) as well as waste products such as corn fiber and wheat straw (2–4). For all lignocellulose conversion processes, access to the cellulosic portion requires the deconstruction of the highly heterogeneous hemicellulose fraction, which is variable depending on the type of plant (5). Xylan, the major polymer backbone of hemicellulose, consists of β-1,4-linked xylose, with substitutions of glucuronic acid, acetate, and l-arabinose. The degree and nature of the substitutions vary between dicots and monocots, with the former being richer in glucuronic acids and galactomannans and the latter being richer in arabinose (5, 6). Furthermore, xylan can be covalently linked to lignin through a 4-O-methyl-d-glucuronic acid ester linkage, adding to the recalcitrance of lignocellulose (7). While the access to, and conversion of, cellulose is critical for developing lignocellulose-based processes, attention to the hemicellulose conversion process for bioprocessing applications is also paramount, as this can account for over 25% of the carbon content present in lignocellulose (3, 4). Ultimately, industrial hosts need to be able to adequately process variable polysaccharide structures to maximize carbon conversion.

Bacteria from the genus Caldicellulosiruptor are the most thermophilic organisms capable of growth on lignocellulose without pretreatment, with Caldicellulosiruptor bescii receiving the most attention due to its genetic tractability and relatively simple metabolic product profile (acetate, lactate, hydrogen, and carbon dioxide) (8). Caldicellulosiruptor species degrade plant biomass primarily through multidomain, secreted, and surface (S) layer-associated glycoside hydrolases (GHs), with several enzymes having multiple catalytic domains (9). In C. bescii and several other Caldicellulosiruptor species, a genomic region known as the glucan degradation locus (GDL) encodes several cellulases and carbohydrate-processing enzymes; in vivo and in vitro studies have determined the collective roles of these enzymes in biomass deconstruction (10, 11). Interestingly, the deletion of a bifunctional cellulase/xylanase, CelC, in a genomic parent strain of C. bescii led to a large decrease in the biotic solubilization of lignocellulosic substrates (11). Thus, it is important to understand how hemicellulases, individually and synergistically, within the genus execute xylan conversion and thereby provide access to cellulose.

Previous efforts to investigate hemicellulose utilization by Caldicellulosiruptor species focused on the functional characterization of several xylanases to develop designer thermostable enzyme cocktails for industrial use or to improve in vivo xylan conversion by C. bescii (12–19). For xylan conversion, Caldicellulosiruptor species primarily use extracellular GH family 10 (GH10) xylanases that either exhibit both glucanase and xylanase activities (14, 19–21) or are xylan specific (22). Several species also utilize a GH11 endoxylanase, which is generally more specific for linear than for branched xylan degradation (14). The remaining complement of extracellular hemicellulases that Caldicellulosiruptor species utilize for deconstruction include secreted GH43 enzymes for arabinoxylan (16) and GH5 enzymes for hexose-rich hemicellulose, such as glucomannan (9). Intracellularly, enzymes from GH families 3, 39, and 43 appear to hydrolyze primarily linear β-1,4-linked xylooligosaccharides, with auxiliary enzymes from GH families 2, 51, and 67 removing heterogeneous sugar groups such as mannose, galactose, arabinose, and glucuronic acids (8, 16, 17). Because the activities of these GH families can be numerous, further characterization of the enzymes in Caldicellulosiruptor species can aid in understanding biomass deconstruction processes.

In vitro studies with recombinant GH10 endoxylanase, GH51 arabinofuranosidase, and GH3 β-xylosidase enzyme mixtures from C. bescii produced high concentrations of monosaccharides, demonstrating that both extracellular and intracellular enzymes are important for liberating fermentable sugars (17). Interestingly, C. bescii cultures do not completely solubilize highly heterogeneous xylan (e.g., oat spelt xylan [OSX]) in vivo without additional metabolic engineering (12). Xylan conversion has been improved by the addition of a surface layer-bound GH10 enzyme from Caldicellulosiruptor kronotskyensis, a secreted GH10 enzyme from Acidothermus cellulolyticus that contains carbohydrate binding module 6 (CBM6), or a β-xylosidase from Thermotoga maritima (12, 13, 23). While these efforts improve xylan utilization, a functional assessment of C. bescii’s hemicellulase inventory could identify its limiting factors for hemicellulose utilization, especially considering combinations of different GH family enzymes that could synergize to facilitate xylan degradation (24).

Recently, a metabolic and regulatory reconstruction of C. bescii identified the genomic clusters for polysaccharide deconstruction and catabolism (25). Pentose utilization is controlled by 64 different genes over 9 genomic loci with several distal genes. These genes are predicted to be controlled by a combination of two-component local regulators and three global repressors, XylR, XynR, and AraR. Within these loci, most of the GHs are encoded within six of the nine genomic clusters or by distal genes. Furthermore, C. bescii encodes no surface layer-associated hemicellulases and shows limited binding to hemicellulose (12), thereby making it a model within the genus for understanding xylanase deployment and available biochemical catalytic activities.

Several studies have investigated the inventory of GHs across the genus either broadly or with a focus on cellulose deconstruction (9, 26, 27). Since these studies, another member of the genus, Caldicellulosiruptor diazotrophicus, has been sequenced (28), and two species, C. kristjanssonii and C. lactoaceticus, have been reclassified as subspecies of C. acetigenus despite the lack of a closed genome for the latter (29). Furthermore, there are several Caldicellulosiruptor species that have larger inventories of GHs, but fewer cellulases, than C. bescii (27), which may provide insight into what is required by members of the genus to degrade more heterogeneous xylan effectively.

To extend previous regulatory reconstruction and metabolic modeling (25, 30), this study investigated the endo- and exoxylanases encoded within the XynR and XylR regulons from the perspective of genomic structure, transcriptional deployment, and biochemical function in order to identify the limiting factors of xylan conversion by C. bescii. This was complemented by a comparative assessment of the genus-wide inventory of hemicellulases to determine the distribution of hemicellulases related to xylan degradation by Caldicellulosiruptor species.

RESULTS

Genomic and operonic structures of xylanolytic clusters in Caldicellulosiruptor bescii.

To further understand the mechanisms of hemicellulose conversion by the type strain of C. bescii (strain DSM 6725, referred to as C. bescii here), five genomic loci, plus distal genes, that encode GHs were further investigated (25) (Fig. 1). The five loci are deemed the xylan degradation locus (XDL), the conserved xylan utilization locus (CXUL), the glucuronoxylan utilization locus (GXUL), the xynMNC operon, and the xynA2 operon; all are controlled by either the XynR or XylR regulators described previously (Fig. 1) (25). There are 12 putative GHs from families 2, 3, 10, 11, 39, 43, 67, and 129 encoded within these loci; 5 of them are predicted to be extracellular (Table 1). Furthermore, these loci account for four of the five GH10 domain-containing enzymes (and two of the three extracellular enzymes); the only extracellular GH43 enzyme; and the only GH3, -11, -67, and -129 enzymes present in the C. bescii genome. Outside these loci, only three additional intracellular GHs were identified, from families 31, 43, and 51, which are used to process α-xyloglucooligosaccharides (GH31) and α-arabinooligosaccharides (GH43 and GH51). While these enzymes are important for hemicellulose utilization, they are predicted to be controlled by different regulatory mechanisms than those for the genes in the above-mentioned clusters (25).

FIG 1.

Genomic and operonic structures of key xylanolytic loci from C. bescii. (A) Three of the five loci in C. bescii were named 174 to 187 and deemed the XDL, 614 to 618 and deemed the CXUL, and 847 through 857 and deemed the GXUL. Terminators were identified in silico using ARNold and are shown using stem-loops after the gene of interest. (B) To confirm operonic structures, RT-PCR was performed using gDNA as a positive control (reaction A), cDNA from xylose (reaction B), and cDNA from arabinose (reaction C). A no-RT control (reaction D) was also performed.

TABLE 1.

Glycoside hydrolases present in C. bescii xylanolytic clusters

ORF	Protein name	GH family(ies)	Secreteda	Locus
Athe_0089	XynA2	11	Y	XynA2
Athe_0178		129	Y	XDL
Athe_0182	XynF	43	Y	XDL
Athe_0183	XynE	10	Y	XDL
Athe_0184	XynD	39	N	XDL
Athe_0185	Xyn10A	10	N	XDL
Athe_0187	XynB	39	N	XDL
Athe_0618	XynA	10	Y	CXUL
Athe_0854	AguX	67	N	GXUL
Athe_0857	BgaL	2	N	GXUL
Athe_1857	CelC	5, 10	N	GDL
Athe_2354	Xyl3A	3	N	Distal
Athe_2724	XynC	10	N	XynMNC

^aY, yes; N, no.

The largest of these genomic clusters, the XDL (Fig. 1), encodes 14 genes (Athe_0174 to Athe_0187 [Athe_0174–0187]). The first eight genes account for two ABC transporters (three genes each), a putative GH129 protein, and one protein of unknown function, while the last six genes encode GHs from families 10, 39, and 43 and one carbohydrate esterase family 4 (CE4) protein. From the XDL, the GH129 (Athe_0178), the dually catalytic GH43 (Athe_0182 [XynF]), and the GH10 (Athe_0183 [XynE]) proteins are all predicted to be secreted. The five-gene CXUL locus contains a putative β-xylooligosaccharide transporter (Athe_0614–0616 [XynUVW]), the XylR transcriptional regulator (Athe_0617), and an extracellular enzyme with a GH10 domain (Athe_0618 [XynA]). Finally, the GXUL harbors 11 genes related to glucuronoxylan degradation, including 2 intracellular GHs from families 2 and 67 (BgaL and AguX). xynMNC, xynA2, and distal genes encode putative GHs from families 2, 3, 10, and 11, with the GH11 enzyme being freely secreted. Across these different loci, there are only four secretome-bound xylanases (XynF, XynE, XynA, and XynA2) that could contribute to lignocellulose deconstruction from GH families 10, 11, and 43. Note that the above-described glucan degradation locus (GDL) encodes a multicatalytic domain protein, CelC (Athe_1857), with an N-terminal GH10 domain, accounting for a fifth extracellular xylanase in C. bescii that is examined in parallel with the enzymes from the xylanolytic loci.

Within the xylanolytic clusters, the operonic structures are of interest for understanding the codeployment of genes related to xylan degradation, especially in the larger genomic clusters such as the 25-kb XDL. To this end, Rho-independent transcriptional terminators in these regions were identified using ARNold (31). Putative terminators are shown as stem-loops in Fig. 1A; only intergenic regions were considered possible terminator sites (see Table S1 in the supplemental material). Between the five genomic loci of interest, putative terminators were identified after Athe_2724, Athe_0618, Athe_0187, and Athe_0089, indicating that the CXUL, XDL, XynMNC, and XynA2 loci could exist as single transcriptional units. Within the GXUL, putative terminators were identified after Athe_0848, Athe_0851, and Athe_0857, indicating three transcriptional units. Overall, the in silico-predicted operonic structures are consistent with the transcriptional units inferred previously based on transcriptomic and regulon reconstruction data (25).

To complement the terminator predictions, the operonic structure of XDL was experimentally assessed since this represents the largest xylanolytic cluster in the genome. C. bescii was grown on either d-xylose or l-arabinose for intergenic reverse transcriptase PCR (RT-PCR) (Fig. 1B; Fig. S1). For C. bescii grown on either arabinose or xylose, the region from Athe_0182 to Athe_0187 showed no transcriptional breaks by RT-PCR, and no predicted terminators were found using ARNold. Taken with the short intergenic regions (Table 2), this region consists of a single transcriptional unit, which includes all of the XDL GHs and the CE4 protein. Only three transcriptional breaks were detected between the xylose- and arabinose-grown cells. On xylose, this break was between Athe_0176 and Athe_0177, which would separate a three-gene ABC transporter from a hypothetical protein; however, this break was not detected in arabinose-grown cells. On arabinose, a break occurred between Athe_0179 and Athe_0180, encoding two components of an ABC transporter. The faint RT-PCR product between Athe_0181 and Athe_0182 suggests that there may be a transcriptional break or readthrough of an unknown terminator, which would separate the transport cluster of the XDL from the GH/CE cluster in the XDL. However, the presence of an amplicon during growth on xylose by RT-PCR for both transcriptional breaks indicates that the genes are likely coexpressed. The genomic DNA (gDNA) showed connections for all genes tested, while the control showed no product. Combined, the complementary RT-PCR analysis of xylose and arabinose, the in silico terminator predictions, and the small intergenic regions suggest that the XDL is transcribed as a single operon.

TABLE 2.

XDL operon PCR primers and intergenic region sizes^a

Reaction	Primer sequence	Primer location	IGR amplicon (PCR product/IGR size [bp])
1	GGATAAATCCAAGAGAAAGTATGTGG	F Athe_0174	Athe_0174-Athe_0175 (268/53)
	AATCTTATATAATTTGACCATCCAACCC	R Athe_0175
2	CAGATTTGAAATGGGCTATGCTTCAG	F Athe_0175	Athe_0175-Athe_0176 (245/40)
	TTTCATCAAATGGCTTAAATGCGTTGAC	R Athe_0176
3	TTGCGCTTGTGCTAATGATTCC	F Athe_0176	Athe_0176-Athe_0177 (276/30)
	CAGCAGCGCTTGGTATTTCATG	R Athe_0177
4	GCACAATAGCTTTGAGTGTTTTGGG	F Athe_0177	Athe_0177-Athe_0178 (371/28)
	CTGTGTTTTCATACTTCCTGCTGC	R Athe_0178
5	CAAGGTATTAAAGGAGGGACTTG	F Athe_0178	Athe_0178-Athe_0179 (298/7)
	CATCCAATAAATTCAAACTCCAAACC	R Athe_0179
6	GCAACCAAGTAATGAGAATGGTTG	F Athe_0179	Athe_0179-Athe_0180 (289/30)
	TGGATACAATAGTATGTAGCTCATGC	R Athe_0179
7	CCTGATTGGTTCTAATGTACTTGAAGC	F Athe_0180	Athe_0180-Athe_0181 (245/75)
	TTTGTGGAGACTTGCCCACAATC	R Athe_0181
8	TTGAGAATTGGCGTTCAAGAAGG	F Athe_0181	Athe_0181-Athe_0182 (278/10)
	ATTATCGGATTTGTTGCAGCTGTTC	R Athe_0182
9	GTGGACAAAGGTAGAAACAAAGG	F Athe_0182	Athe_0182-Athe_0183 (270/17)
	ATTGTTATTTTTGCTTTGCCATATGC	R Athe_0183
10	GATGAGCATTATGATGGCAAACCTG	F Athe_0183	Athe_0183-Athe_0184 (304/42)
	GTAGATTCCAACATCGTCATGAAGC	R Athe_0184
11	AGCAAGAGAGGGTTATGTAACAC	F Athe_0184	Athe_0184-Athe_0185 (298/20)
	GTTGTAAAAATCTTCTTTCGGATGAATTC	R Athe_0185
12	GACATACATGGAAAGACAATTTTCCGG	F Athe_0185	Athe_0185-Athe_0186 (274/47)
	CATGTGATTGCCCGCATATCC	R Athe_0186
13	TTCCTTGTTCATGAGAACAGAGAATTG	F Athe_0186	Athe_0186-Athe_0187 (245/90)
	CTTTGTCCAGAATTTATTTATCTTGCC	R Athe_0187

^aF, forward; R, reverse; IGR, intergenic region.

Biochemical characterization of xylanases from C. bescii.

To determine the individual and collective roles of each GH for xylan degradation by C. bescii, full-length, active versions of 7 of the 12 GHs from the xylanolytic loci, along with CelC (Athe_1857; GH10-CBM3-CBM3-GH48) from the GDL, were examined (Fig. 2). While some of these enzymes or their homologs have been characterized previously (10, 14, 16, 17), this work aimed to investigate the key endo- and exo-acting xylanases from C. bescii in a side-by-side comparison to determine the relative importance of these enzymes in xylan and biomass deconstruction. This included all five GH10 proteins (XynA, Xyn10A, XynC, XynE, and CelC), the sole versions of GH11 (XynA2) and GH3 (Xyl3A) proteins, and one of the two GH39 proteins. The expression and purification of full-length Athe_0182 (XynF; GH43) were unsuccessful in either the Escherichia coli or C. bescii host, consistent with previously reported attempts to make full-length versions of this protein (16, 32). Additionally, one GH2 protein (Athe_0857 [BgaL]) and a protein of unknown function, XynM (Athe_2722), were successfully expressed but did not have activity on any of the substrates tested (e.g., see those listed in Fig. 2B) (data not shown).

FIG 2.

Verification of the activity of xylanases from C. bescii. (A) Domain structure and GH family for each open reading frame (ORF). (B) Activity screening for polysaccharides and pNP-linked substrates was performed at 75°C (pH 6.5) with 0.5 μM enzyme. Substrate abbreviations: OSX, oat spelt xylan; BBG, barley β-glucan; TXG, tamarind xyloglucan; CMC, carboxymethylcellulose; BeX, beechwood xylan; WAX, wheat arabinoxylan; BiX, birchwood xylan; BA, beet arabinan; pNP, para-nitrophenol; pNPX, pNP–β-d-xylopyranoside; pNPA, pNP–α-l-arabinofuranoside; pNPM, pNP–β-d-mannopyranoside; pNPGlu, pNP–β-d-glucopyranoside; pNPGal, pNP–β-d-galactopyranoside. The last lines of reactions in panel B represent the no-enzyme control.

The dominant activities of the C. bescii xylanases were evaluated by the hydrolysis of polysaccharides and para-nitrophenol (pNP)-linked monosaccharides at 75°C (pH 6.5) with 0.5 μM enzyme. Glucan substrates were included in the activity screen since several GH10 proteins from C. bescii have been reported to be bifunctional glucanases/xylanases (Fig. 2). None of the enzymes tested were active on xyloglucan or arabinan, as expected, since the primary xylose linkages in xyloglucan are α-1,2 xylose with a glucan backbone, while arabinan is a polymer of primarily α-1,5 arabinose linkages. CelC had weak activity on carboxymethylcellulose (CMC), while none of the other enzymes had detectable CMC activity. All enzymes, except for XynD (GH39) and Xyl3A (GH3), were active on oat spelt xylan (OSX), beechwood xylan (BeX), wheat arabinoxylan (WAX), and birchwood xylan (BiX). In general, for the enzymes tested, visual inspection of the activities on OSX showed much lower levels than those of the other xylans, likely because there are both glucose and arabinose side chains on OSX, in contrast to WAX and BeX, which have arabinose and no side chains, respectively.

Only XynD, XynE, Xyn10A, and Xyl3A exhibited activity on pNP–β-1,4-xylopyranoside (pNPX) or pNP–α-1,5-arabinofuranoside (pNPA), indicating the potential to liberate only free xylose or arabinose from oligosaccharides. XynD had activity only on pNPX, while Xyl3A had activity on pNPA and pNPX. Interestingly, both XynE and Xyn10A exhibited activity on both pNPX and pNPA, activities uncommon for GH10 enzymes. A previous characterization of an intracellular GH10 enzyme from the unsequenced Caldicellulosiruptor sp. strain Rt8B.4 indicated a combination of β-xylosidase, α-arabinosidase, and endoxylanase activities (33). These activities are rare for GH10 proteins, warranting further investigation.

Of the five GH10 domain enzymes in C. bescii, either the C. bescii protein or a homolog from another Caldicellulosiruptor species has been partially characterized, except for XynC. XynC is an intracellular GH10 protein encoded in the xynMNC locus, with visual substrate screening indicating that this enzyme could be less active than the other GH10 proteins. To determine if reaction conditions influence the lower activity, the pH and temperature optima as well as thermostability were determined for XynC using BeX as the substrate (Fig. 3). XynC showed optimal activity at 80°C and pH 6.5. At 75°C, the enzyme retains 60% of its activity after 4 h and 40% of its activity after 24 h. However, at 85°C, less than 20% of the original activity was observed after 2 h. These optima and thermostability parameters are consistent with those of other GH10 enzymes characterized from C. bescii (17).

FIG 3.

Biochemical characterization of XynC. The pH optimum (A), temperature optimum (B), and thermostability (C) were determined for recombinant XynC.

To determine the relative differences between the individual xylanases, further quantitation of the specific activity for xylan-based substrates was conducted (Fig. 4). CelC has been previously characterized as a bifunctional glucanase/xylanase (10). In this study, the highest activity for CelC was observed on WAX (9,172 ± 21 U/mg protein), with significantly less activity on BeX (1,093 ± 138 U/mg protein) and minimal activity on OSX (257 ± 39 U/mg protein). The most active enzyme on OSX was Xyn10A, an intracellular GH10 enzyme, at 3,612 ± 107 U/mg protein. Both CBM22-GH10 proteins (XynA and XynE) had similar levels of activity on OSX and WAX, with the difference being that XynA had 36% higher activity on BeX (7,863 ± 184 U/mg protein) than XynE (5,766 ± 50 U/mg protein); the activity for XynA on BeX is approximately the same as that for Xyn10A, which was previously reported to be the more active enzyme (17) (Fig. 4). As XynA is colocated with a β-xylooligosaccharide transport cluster (XynUVW), the higher activity on less branched xylan is expected. Interestingly, XynC had the lowest activity of any of the xylanases in C. bescii, indicating the need for accessory enzymes to perform its function, such as XynM and XynN, which are putative carbohydrate esterases from families 20 and 4, respectively. These genes likely remove ester branches from xylose, allowing easier enzymatic access to cleave the backbone.

FIG 4.

Specific activities of xylanases on polysaccharides and pNP-linked substrates. (A) All polysaccharide substrates were loaded at 1% (wt/vol) and incubated at 75°C (pH 6.5) with 0.2 μM enzyme for 15 min, with reducing sugars being measured by a DNSA assay using xylose as the standard. Abbreviations: OSX, oat spelt xylan; WAX, wheat arabinoxylan; BeX, beechwood xylan. (B) For pNP-linked substrates, the enzyme was incubated for 2 min at final concentrations of 10 μg for Athe_0185 and Athe_0183 and 1 μg for Athe_2354 and Athe_0184. All reactions were performed in triplicate.

The occurrence of β-xylosidase and α-arabinosidase activities from both GH10 enzymes (XynE and Xyn10A) in the XDL during the activity screens was unexpected as GH10 proteins are typically glucanases or xylanases. Additionally, α-arabinosidase activity was detected for Xyl3A, which had not been identified in the previously reported characterization of this enzyme (17). To further examine the activity on pNP-linked substrates, enzymatic reactions were conducted using 4 mM pNPX or pNPA as the substrate and 10 μg protein due to the uncertainties in the putative activities (Fig. 4B). XynE, Xyn10A, XynD, and Xyl3A were all active on pNPX. Of these four enzymes, only XynD exhibited no activity on pNPA, consistent with the substrate specificity investigation described above. Both GH10 proteins that exhibited activity on pNPX and pNPA had significantly lower activities than those of the GH3 and GH39 exohydrolases. For Xyn10A, the activities on pNPX and pNPA were similar, at 0.508 ± 0.04 and 0.527 ± 0.03 μmol pNP released/min/mg protein, respectively. XynE had slightly higher activity, at 1.70 ± 0.03 and 1.42 ± 0.1 U/mg protein on pNPA and pNPX, respectively. Surprisingly, the GH39 enzyme XynD had 3.3-fold the activity of the GH3 enzyme Xyl3A on pNPX (18.8 ± 0.01 compared to 5.77 ± 0.7 U/mg protein) despite not being active on pNPA. On a molar basis, this ratio of activity differences (not shown) would be slightly higher as XynD has a lower M_r than that of Xyl3A.

All four families of enzymes tested (GH3, -10, -11, and -39) are retaining enzymes and therefore can have transglycosylation reactions that compete with hydrolysis (34). The low activity of the GH10 proteins on pNP-linked substrates is likely not biologically relevant as endohydrolysis is their primary activity, with over 10,000-fold-higher activity in some cases. However, for the GH3 and GH39 intracellular exohydrolases, transglycosylation may be relevant. To this end, enzymatic reactions were conducted using 5 mM either xylobiose, xylotriose, or xylopentose at 75°C with 0.2 μM enzyme (Fig. S3). Additional reactions were performed, one with 5 mM (each) xylotriose and xylopentose and another with 5 mM xylotriose plus 10 mM xylose. For Xyl3A, the complete conversion of xylooligosaccharides was observed, except in the case of supplementing xylotriose with xylose, where a small xylobiose peak was observed. Incubation of XynD with xylotriose produced a combination of xylotetrose, xylobiose, and xylose along with some unknown products. The addition of additional xylose slightly increased the amounts of xylobiose and xylotetrose formed. The use of xylopentose or an equimolar mixture of xylopentose and xylotriose as the substrate produced xylotetrose, xylobiose, and xylose as the products; however, the same reaction with Xyl3A produced only xylose, which is consistent with previous work (17). No transglycosylation products were detected using xylobiose as the substrate. Despite the higher activity on pNPX for the GH39 enzyme, there was significant transglycosylation activity, suggesting that the GH3 enzyme is the primary enzyme for producing free xylose intracellularly. However, the transglycosylation reactions may in turn allow C. bescii to retain carbon intracellularly, consistent with its isolation from nutritionally poor environments. Other characterized GH39 enzymes also have transglycosylation activity, with pNP-xylose as the donor, xylotriose as the acceptor molecule, and xylotetrose as a product, indicating the consistency of the C. bescii GH39 transglycosylation products (34).

Transcriptional regulation of hemicellulases in Caldicellulosiruptor bescii.

To determine how Caldicellulosiruptor species regulate hemicellulases and supporting proteins for hemicellulose deconstruction, transcriptomic data reported previously by Rodionov et al. (25) were examined in the context of the xylanolytic clusters described above. Two separate analyses were performed. The first analysis compared differential expression on various growth substrates to determine the relative importance of a gene during biomass conversion. The second analysis was a rank order analysis to determine the relative transcription level of a gene compared to the transcription of every other gene in the genome, with emphasis on the transcriptional rank for the xylanases studied in this work (Fig. 5).

FIG 5.

Transcriptional regulation of xylanases in Caldicellulosiruptor bescii. Previously obtained RNA sequencing data were reanalyzed in the context of the glycoside hydrolases from the xylanolytic clusters investigated in this work using rank order analysis (A) or fold change values (B).

From RNA sequencing data for C. bescii obtained using five different carbon sources, four differential expression comparisons were made between C. bescii growth substrates: Avicel versus cellobiose, Avicel versus xylose, xylan versus Avicel, and xylan versus xylose. XynA2 showed the highest overall fold change when comparing polysaccharides to monosaccharides: an 18.4-fold increase on Avicel compared to cellobiose, a 15.1-fold increase on Avicel compared to xylose, and a 33.8-fold increase on xylan compared to xylose. Interestingly, there was only a 2.2-fold increase in the expression of Athe_0089 (GH11) on xylan compared to Avicel. Low, statistically insignificant levels of induction were observed between xylan and Avicel for the XDL GHs XynF (GH43), XynE (GH10), XynD (GH39), Xyn10A (GH10), and XynC (GH39); however, greater induction of these genes was found on polysaccharides than on simpler sugars. The largest fold changes for GHs on xylan compared to Avicel were for Athe_2354 (7.6-fold increase) and Athe_0618 (5.0-fold increase). The non-XDL GHs Athe_0857 (GH2), Athe_2722 (GH2/CE20), Athe_0854 (GH67), and Athe_2724 (GH10) had between 3.1- and 3.5-fold increases, suggesting that the role of these enzymes may be related more directly to xylan conversion than to overall biomass deconstruction.

To determine the relative importance of the GHs on different substrates, a rank order analysis looked at the expression of individual genes with respect to all genes in the genome. For all growth substrates, CelC was the most highly expressed hemicellulase. This is not surprising as it is part of the highly expressed GDL and is transcriptionally on par with other GDL enzymes (Table S2). Along with CelC (GH10/48), XynA and Xyl3A were in the top 90th percentile of genes expressed on xylan, suggesting their critical role in xylan conversion. The transcriptional rank of the GHs from the XDL decreases from the 85th percentile for XynF (GH43) to below the 40th percentile for XynB (Athe_0187; GH39); however, these genes are coexpressed, as demonstrated above, so the decreased expression compared to those of the other xylanases may be due to the putative cross-regulation of the XDL by arabinose/arabinoxylan, as predicted previously (25). Despite having a lower specific activity than those of other GH10 proteins, XynC is in the 70th percentile of all genes expressed on xylan, which is higher than that of the other intracellular GH10 protein within the XDL, Xyn10A. Interestingly, the high-level induction of the expression of XynA2 (GH11) on xylan does not correlate with high expression levels, as it is in only the 60th percentile of genes expressed on xylan. Its specificity for xylan, its lower relative size than that of the GH10 enzymes, and its relatively high activity may offset the lower transcriptional levels. Furthermore, the expression levels of the two intracellular β-xylosidases, XynD (GH39) and Xyl3A (GH3), on xylan show a 6-fold-higher expression level of the GH3 enzyme despite it being 3.3-fold less active. This further suggests that the GH3 enzyme is the primary β-xylosidase that C. bescii utilizes for xylose generation.

From the available RNA sequencing data, the transcriptional levels of the additional genes located in the XDL were further investigated (Fig. 6). While the RT-PCR data presented above suggest a single transcriptional unit, the RNA sequencing analysis shows broad variation in the transcriptional ranks of the XDL genes on all substrates. The most interesting feature is the expression of Athe_0181, the extracellular component of the Athe_0179–Athe_0181 (AxoFGE) ABC transporter, which is in the top 95% of genes expressed for both xylan and Avicel. A recent characterization of this protein showed binding to cellulose and an enhancement of GH5 activity for Avicel hydrolysis (35) despite it being encoded in a locus designed primarily for xylan conversion. However, the reason for the increased expression is unknown as searches for internal promoters within the XDL using PePPER (36) (Table S3) found putative promoter sequences within the coding sequences, except before Athe_0174, which is the promoter that likely controls the XDL.

FIG 6.

Transcriptional regulation of the xylan degradation locus in Caldicellulosiruptor bescii. Previously obtained RNA sequencing data were reanalyzed for the additional genes located in the XDL but not characterized using rank order analysis (A) or fold change values (B).

Reassessment of the core and pangenomes within Caldicellulosiruptor and defining the core and panhemicellulase inventories.

Updates to the pan- and core genomes were made to the genus Caldicellulosiruptor with the addition of sequenced genomes of Caldicellulosiruptor diazotrophicus YA01 and Caldicellulosiruptor changbaiensis CBS-Z (28, 37, 38) (Fig. S2). The two new species reduce the core genome size to 1,280 orthologous groups from 1,401 groups and increase the pangenome size from 3,493 orthologous groups to 3,853 groups since the last comprehensive analysis (27). Based on a power law fit of the pangenome size versus the number of species having a positive coefficient, the pangenome of Caldicellulosiruptor remains open.

To determine if there are conserved strategies for xylan utilization, the genus-wide inventory of GHs was reevaluated using dbCAN (Table 3) (39). GH inventories were further curated to include all proteins from GH families 2, 3, 10, 11, 39, 43, 51, and 67 as the core set of xylanases and accessory enzymes (Fig. 7). A total of 223 GH-encoding genes were identified from these 8 GH families across the genus. Interestingly, the most recently sequenced species, C. diazotrophicus, had the most GHs related to xylan degradation, followed by C. kronotskyensis, C. owensensis, and C. changbaiensis (Table 3). To further delineate the conservation of the hemicellulases within the genus, an analysis was performed to identify the orthologous xylanase genes across all 15 species (“core hemicellulase”) and unique xylanase genes present in at least one species (“panhemicellulase”). From the 223 identified GH-encoding genes, there are only 28 unique genes from the 8 GH families considered and only 3 hemicellulases from GH families 2, 3, and 10 that are conserved across all 15 members of the genus (Fig. 7), 2 of which are homologs of XynA (secreted GH10 enzyme) and Xyl3A (intracellular GH3 enzyme). These two enzymes represent a possible minimum combination of GHs for xylan utilization by Caldicellulosiruptor species as they have endohydrolase activity (GH10) to degrade long-chain β-1,4-d-xylose polymers and β-xylosidase activity (GH3) to convert oligosaccharides into monomers. Coupled with the biochemical activities for these genes, this indicates that the diversity of hemicellulases across the genus is likely for processing heterogeneous xylans.

TABLE 3.

GH inventory from Caldicellulosiruptor species^a

Species	No. of GHs
	Total GHs	HC GHs	GH2	GH3	GH10	GH11	GH39	GH43	GH51	GH67
C. morganii	49	11	2	1	5	1	0	0	1	1
C. naganoensis	44	6	1	1	4	0	0	0	0	0
C. lactoaceticus	44	8	2	1	3	0	0	0	1	1
C. bescii	52	17	4	1	5	1	2	2	1	1
C. kronotskyensis	77	22	4	1	6	1	3	4	2	1
C. danielii	69	16	4	1	5	1	1	2	1	1
C. saccharolyticus	59	19	4	2	6	0	2	3	1	1
C. obsidiansis	47	13	5	1	3	0	0	2	1	1
C. kristjanssonii	37	8	2	1	3	0	0	0	1	1
Caldicellulosiruptor sp. F32	45	9	1	3	2	1	0	1	1	0
C. owensensis	51	21	3	1	6	1	2	5	2	1
C. hydrothermalis	62	13	3	2	2	0	0	3	2	1
C. acetigenus	66	16	3	1	5	0	2	2	2	1
C. changbaiensis	70	20	4	1	7	0	2	3	2	1
C. diazotrophicus	75	25	8	1	6	0	2	5	2	1

^aShown is a genomic inventory of GHs and the hemicellulase (HC) GHs from families 2, 3, 10, 11, 39, 43, 51, and 67.

FIG 7.

Distribution and conservation of xylanases in the Caldicellulosiruptor genus. Manual curation of dbCAN results for GH families 2, 3, 10, 11, 39, 43, 51, and 67 from 15 Caldicellulosiruptor type strain genomes was performed. Unique genus-wide ORFs associated with hemicellulose conversion (A) and percent identities relative to the reference ORF (B) are shown. The light-blue star next to CF32 xyl3C represents a unique gene from Caldicellulosiruptor sp. F32 that does not have a unique ORF identification due to an unclosed genome.

In general, the conservation of GH10 enzymes varied across the genus. C. naganoensis, C. danielii, and C. morganii all have homologs of C. bescii CelC and XynA, which allows xylan processing with a smaller hemicellulolytic inventory. Furthermore, orthologs of XynC are present in all Caldicellulosiruptor species; however, in the case of C. naganoensis, it is fused to carbohydrate binding modules and extracellularly expressed, as opposed to the other Caldicellulosiruptor species, where it is intracellular. Despite unsuccessful attempts to express full-length, XynF, previous analyses investigated the role of secreted GH43 containing enzymes by subcloning the individual domains from the XynF homolog present in C. saccharolyticus (16). The protein contains two GH43 domains from subfamilies GH43_10 (exo-α-arabinosidase) and GH43_16 (α-l-1,3-arabinosidase) and is the only secreted GH43 containing enzymes protein in C. bescii. Homologs of this multidomain, secreted protein were identified in 6 additional species of Caldicellulosiruptor. Interestingly, a surface layer-associated GH43 domain-containing protein is present in C. hydrothermalis and C. diazotrophicus, indicating possible biomass de-cross-linking activity by genes associated with the cell surface. Both C. diazotrophicus and C. hydrothermalis contain additional intracellular and secreted GH43 domain-containing enzymes, which could allow them to process arabinoxylans that are recalcitrant to other Caldicellulosiruptor species.

While many of the xylanases are conserved throughout the genus, the XDL is somewhat more diverse (Fig. S4). Three different organizations of the XDL were identified, with the first being the structure present in C. bescii, with 14 genes, which is also present in C. kronotskyensis, C. owensensis, C. diazotrophicus, C. acetigenus, and C. changbaiensis. C. saccharolyticus has a structure similar to that of C. bescii except for the presence of an additional GH10 enzyme. Five additional species have the eight-gene transporter and hypothetical protein cluster without the GHs from the XDL, except for C. danielii, which encodes the homolog of XynB. In general, species that lack an XDL also tend to lack homologs of the five XDL xylanases (Fig. 7) (XynF, XynE, XynD, Xyn10A, and XynB). Furthermore, these species generally lack GH39 enzymes and have fewer intracellular GH10 enzymes for oligosaccharide processing as the conserved GH3 proteins can serve as the primary β-xylosidase.

Assessment of synergistic interactions between C. bescii xylanases.

With the conservation of the hemicellulases across the genus further understood, individual and equimolar mixtures of C. bescii xylanases were incubated with OSX to screen for synergistic interactions (Fig. 8). The combination of XynA and XynD (GH10 and -39) resulted in the largest amount sugar release under the conditions tested (2.01 ± 0.11 g/L xylose equivalents), with the combination of XynA and Xyl3A showing slightly smaller amounts, at 1.94 ± 0.25 g/L xylose released. Because there was low activity for the β-xylosidases (XynD and Xyl3A) on OSX, the synergy values under these two conditions were 3.50 ± 0.41 and 2.80 ± 0.72, respectively, representing the highest degree of synergy of the pairwise mixtures tested under these conditions.

FIG 8.

Synergism of GH10 and -11 xylanases. Endoxylanases were loaded at 0.2 μM total protein either individually or at an equimolar (Eq) mixture of the two proteins indicated. (A to D) Synergy results for XynA2 (GH11) (A), Xyn10A (GH10) (B), XynE (GH10) (C), and XynA (GH10) (D). (E) Comparisons between supplementation with XynD (GH39) and Xyl3A (GH3). For the reactions, the following enzyme letters were used: A, XynA2; B, XynE; C, Xyn10A; D, XynA; E, CelC; F, XynC; G, XynD; H, Xyl3A. (F) Top 10 degree of synergy (DOS) values observed under all conditions tested.

For the GH10 domain-containing proteins, of the 10 equimolar mixtures tested, the most biologically relevant combinations are Xyn10A and XynC since they are both intracellular GH10 proteins; XynE and Xyn10A because they are encoded within the same genomic region, despite one protein being secreted; XynE and XynA since they have similar domain organizations; and the pairings of XynE and XynA with CelC because they are secreted GH10 enzymes. Of these five pairs, the highest synergy was observed between XynE and Xyn10A (1.43 ± 0.29), followed by XynE and CelC (1.30 ± 0.21). The other pairs of secreted enzymes (XynA/CelC and XynA/XynE) and the pair of intracellular enzymes (Xyn10A/XynC) showed synergism values of <1.2 or were not statistically significant. Overall, the low synergism values within the GH10 domain-containing enzymes suggest that the different secreted enzymes are likely binding to distinct portions of the biomass to access hemicellulose rather than working in coordination to perform depolymerization.

Interestingly, the most synergistic interactions, not involving a β-xylosidase, are for mixtures containing XynA2, the sole secreted GH11 enzyme in C. bescii. The highest synergy values of XynA2 were found with the secreted GH10 enzyme XynA (1.69 ± 0.31) and with the intracellular GH10 enzyme Xyn10A (1.54 ± 0.10), which also had the highest sugar release under any condition without a β-xylosidase (1.74 ± 0.22 g/L and 1.68 ± 0.062 g/L xylose equivalents). Furthermore, the addition of β-xylosidases seemed to have no effect on the sugar released despite synergy values of >1 for each β-xylosidase. This may indicate an antagonistic effect of the enzymes, which in vivo could be due to their differential localizations in the cell (secreted versus intracellular) or smaller hydrolysis products generated by the GH11 enzyme. While the synergism values for the GH11 enzyme with different GH10 enzymes, especially those that are secreted, are not high, XynA2 is the smallest of the secreted xylanases, so it may be helping to degrade minimally branched xylan, while the GH10 enzymes are cleaving in areas that are more branched. However, homologs of XynA2 are found in only 5 additional species of Caldicellulosiruptor (C. kronotskyensis, C. owensensis, C. morganii, C. danielii, and Caldicellulosiruptor sp. strain F32), 3 of which (C. kronotskyensis, C. morganii, and C. danielii) are more cellulolytic. The presence of a smaller xylanase in the secretome acting synergistically with GH10 domain-containing proteins may be a mechanism for processing linear xylan more efficiently.

DISCUSSION

In contrast to the GDL of C. bescii, the xylanolytic clusters are distributed throughout the genome in smaller genomic loci and to some extent as single operons. While the large locus devoted to glucuronoxylan degradation is multioperonic, the XDL is likely transcribed as a single unit despite the measurement of higher transcription levels in the middle of the locus. The secreted enzymes from GH families 10 and 11 degrade both branched- and straight-chain xylan, with all three secreted GH10 enzymes showing activity on glucan as well. Despite the unsuccessful characterization of the full-length multicatalytic GH43 domain from the XDL, it is likely that this enzyme is providing key de-cross-linking activity in a manner similar to that of Cellvibrio japonicus, which uses two freely secreted and one membrane-bound GH10 proteins, two secreted GH11 proteins, and a GH43 protein for arabinoxylan utilization (34).

For the further processing of xylooligosaccharides, C. bescii employs a GH3 and two GH39 enzymes, one of which (XynD) is characterized in this work. Although Xyl3A had slightly lower activity than XynD, it is transcribed at a much higher level, suggesting that it is the primary enzyme for xylooligosaccharide conversion to xylose. The GH39 protein characterized in this work also had transglycosylation activity, consistent with other GH39 proteins (40). Furthermore, the cooperonic nature of the GH39 enzyme with both secreted and intracellular GH10 enzymes could be to move free xylose to odd-chain xylooligosaccharides to allow processing by the two intracellular GH10 enzymes (Athe_0185 and Athe_2724). Interestingly, there are fewer intracellular GH10 proteins in Caldicellulosiruptor species that lack GH39 proteins, suggesting that the GH3 homolog primarily degrades xylooligosaccharides.

Despite the variation in the GH inventory in the genus Caldicellulosiruptor, several conserved strategies exist for xylan degradation, with the most notable being the presence of the GH3 enzyme for intracellular oligosaccharide processing and the multifunctional, secreted GH10 enzyme for polysaccharide utilization. The combination of these two enzymes was among the most synergistic pairs and had the highest sugar release from oat spelt xylan (OSX). C. bescii has 17 of the 28 unique proteins from GH families 2, 3, 10, 11, 39, 43, 51, and 67 for hemicellulose deconstruction; the additional unique enzymes were primarily from GH2 and GH43, which have broad substrate specificities. The presence of GHs in other species that are homologous to those found in the C. bescii XDL represents a partially conserved strategy for deploying additional xylanases and transporters to aid in the deconstruction and uptake of heterogeneous products from lignocellulose. However, the absence of an XDL does not mean that certain activities are missing, with one example being a surface layer-associated GH43 protein from C. hydrothermalis, which represents the primary extracellular GH43 protein for that organism.

Further analysis of recent RNA sequence data (25) for C. bescii indicated that the presence of xylan or cellulose led to transcriptional ranks above the 90th percentile for several critical enzymes. While most of the enzymes from the XDL are primarily xylanases, the ability to hydrolyze glucans suggests a role for the coexpression of these enzymes during growth on Avicel. For xylanases in the lower percentiles on xylan, there is still a large fold induction on xylan compared to glucose; the most notable is the GH11 protein XynA2, which is expressed at levels 30-fold higher on xylan than on xylose. In general, the transcription of the core xylanases in C. bescii is driven more by the presence of polysaccharides than by the presence of monomeric pentose. However, in some cases, such as XynD (GH39) and Xyl3A (GH3), the protein with less expression has slightly more activity, indicating that the enzymes from Caldicellulosiruptor species work in conjunction with each other to process oligosaccharides.

Overall, the coordinated enzyme expression from several distinct loci in C. bescii drives xylan degradation through a relatively minimal set of GHs. While there is some activity that is shared across all enzymes, unique sets of activities were observed for genes located at distinct loci. Despite the variability in the conservation of these GHs in the genus, the major activities to degrade xylan are present in all species, suggesting a minimal set of genes necessary for Caldicellulosiruptor species to grow on xylan. Further assessment of the transcription factors that control xylanase expression will provide more insight into the strategies that C. bescii uses to deploy certain secreted enzymes over others. This, coupled with assessing C. bescii on different lignocellulosic substrates, will allow the development of this organism into a more robust metabolic engineering platform that can handle different lignocellulosic substrates.

MATERIALS AND METHODS

Culturing of Caldicellulosiruptor species and preparation of genomic DNA.

Caldicellulosiruptor bescii DSM 6725 was grown on modified DSM516 medium, as described previously (10), using 5 g/L of xylose, arabinose, cellobiose, beechwood xylan (BeX) (Biosynth), or microcrystalline cellulose as the carbon source. All cultivation occurred at 75°C with agitation at 150 rpm using 50 mL of medium in a 125-mL serum bottle with a headspace of 80% N₂–20% CO₂. Five-milliliter samples of the culture were spun down at 6,000 × g for 10 min; next, genomic DNA was extracted and purified using the Monarch genomic DNA purification kit (New England BioLabs, Ipswich, MA), according to the manufacturer’s instructions, using Monarch tissue lysis buffer plus 25 mg/mL lysozyme for lysis.

Plasmid construction and maintenance and protein expression.

All genes of interest and the pET28a plasmid backbone were PCR amplified using Q5 polymerase (New England BioLabs, Ipswich, MA) with suitable overhangs to perform Gibson assembly (see Tables 4 and 5 for plasmids and primers). Gibson assembly was performed using the NEBuilder HiFi DNA assembly master mix. For the GHs Athe_0089, Athe_0184, Athe_0185, Athe_2354, and Athe_2724, the endogenous stop codon was removed to add a C-terminal 6×His purification tag. For Athe_0183 and Athe_0618, the stop codon was left in to generate untagged proteins. All plasmids were transformed into E. coli DH5α, sequence verified by Sanger sequencing (Azenta), and maintained as glycerol stocks in Luria-Bertani (LB) medium containing 15% glycerol and 50 μg/mL kanamycin at −80°C.

TABLE 4.

Plasmids used in this study

Plasmid	Purpose	Protein tag, location	Reference
pET28a-0089	Expression of Athe_0089 without the signal peptide	6×His, C terminus	This work
pET28a-0183	Expression of Athe_0183 without the signal peptide	None	This work
pET28a-0184	Expression of full-length Athe_0184	6×His, C terminus	This work
pET28a-0185	Expression of full-length Athe_0185	6×His, C terminus	This work
pET28a-0618	Expression of Athe_0618 without the signal peptide	None	This work
pJMC053	Expression of Athe_1857 in C. bescii with a tag	6×His, C terminus	10
pET28a-2354	Expression of full-length Athe_2354	6×His, C terminus	This work
pET28a-2724	Expression of full-length Athe_2724	6×His, C terminus	This work

TABLE 5.

Primers used for plasmid construction

Primer	Purpose	Sequencea
JRC 807	pET28a backbone R	GCCCATGGTATATCTCCTTCTTAAAG
JRC 808	pET28a backbone F	ACTCGAGCACCACCACC
JRC 811	pET28a-183 F	ctttaagaaggagatataccatgggcGCACAGACAACCTCAACAA
JRC 812	pET28a-183 R	ggtggtggtggtgctcgagTTTATTTTTTAGCCTTTACTTTTGGAATAGC
JRC 816	pET28a-618 F	ctttaagaaggagatataccatgggcGAAACAAAAAAGAGTTTTGTGGAGTATAA
JRC 817	pET28a-618 R	ggtggtggtggtgctcgagTTTATTCTTCTGGCACAACTGAC
JRC 825	pET28a-2354 F	ctttaagaaggagatataccatgggcTCAATTGAAAAAAGGGTAAACCAG
JRC 826	pET28a-2354 R	ggtggtggtggtgctcgagTTTCACACCATGCATGGC
JRC 780	pET28a-0089 F	ctttaagaaggagatataccatgggcGCAATAACCCTCACATCCAAT
JRC 781	pET28a-0089 R	ggtggtggtggtgctcgagTATTAACAAATAATCTGCATAAACATCC
JRC 827	pET28a-0184 F	ctttaagaaggagatataccatgggcACATATGTAAAAATTGAACGAGG
JRC 828	pET28a-0184 R	ggtggtggtggtgctcgagTTAACCTGGAATTTTACTATCATC
JRC 829	pET28a-0185 F	ctttaagaaggagatataccatgggcAGCGAAGATTATTATGAAAAGTC
JRC 830	pET28a-0185 R	ggtggtggtggtgctcgagAAAGTCAATTATTCTGAAAAATGCC
JRC 831	pET28a-2724 F	ctttaagaaggagatataccatgggcGATAGATACAATCATCGAAAAAGC
JRC 832	pET28a-2724 R	ggtggtggtggtgctcgagTAAAATTATCTGAGCAACTTCATG

^aLower case indicates Gibson assembly overlap region to pET28a backbone.

Plasmids were then transformed into E. coli Rosetta/pLysS with strains maintained in LB medium containing 15% glycerol, 50 μg/mL kanamycin, and 34 μg/mL chloramphenicol. Protein expression was performed in 2× YT medium (10 g/L yeast extract, 16 g/L tryptone, 5g/L sodium chloride) plus the appropriate antibiotics at 37°C until the optical density at 600 nm (OD₆₀₀) was between 0.6 and 0.8. Cultures were cooled to room temperature and then induced by adding isopropyl-β-d-thiogalactopyranoside (IPTG) to a final concentration of 0.5 mM, and cultivation was continued at 20°C for 16 to 18 h. Cells were harvested by centrifugation at 6,000 × g for 10 min prior to lysis and purification.

Protein purification.

E. coli cell pellets were resuspended (3 mL/g [wet weight] of cells) in either 50 mM sodium phosphate and 300 mM sodium chloride (pH 8.0) (immobilized-metal affinity chromatography [IMAC] buffer A) for His-tagged proteins or 50 mM morpholineethanesulfonic acid (MES) (pH 6.0) cation exchange (CEX A) for untagged proteins. Resuspended cells were lysed using 3 passes in a French press (Sim-Aminco) operating at 1,000 lb/in². Because the proteins were expected to be thermostable, lysates were heat treated at 65°C for 20 min, clarified by centrifugation at 20,000 × g for 60 min, and sterile filtered through a 0.22-μm filter prior to purification.

His-tagged proteins were purified by IMAC using a 500 mM imidazole gradient for elution on a 5-mL HisTrap HP column (Cytiva). Untagged proteins were purified by cation exchange chromatography using 50 mM MES (pH 6.0) as a binding buffer and a gradient elution with the binding buffer plus 1 M sodium chloride for elution on a HiTrap SP column (Cytiva). Proteins were further purified as necessary by an additional heat treatment step at 65°C for 20 min. All chromatography steps were performed on a Biologic DuoFlow fast protein liquid chromatography (FPLC) system (Bio-Rad).

Protein fractions containing the correct protein were pooled and concentrated through Vivaspin 20 filters with the appropriate molecular mass cutoff (10, 30, or 50 kDa). Cutoff sizes were chosen to allow for at least a 10-kDa difference between the filter size and the expected M_r of the protein. All GHs were buffer exchanged into 50 mM sodium phosphate (pH 6.5) after purification.

Expression of CelC.

The expression of recombinant CelC (Athe_1857) was performed as previously described (10). Briefly, the strain of C. bescii harboring the CelC expression plasmid was grown in a 30-L Biostat reactor (Sartorius) with 17 L of DSM516 medium plus 100 μg/mL kanamycin for 30 h. The fermentation broth was filtered through a 0.22-μm filter to remove cells, concentrated, and buffer exchanged into IMAC buffer A through a Pellicon 2 filter with a 100-kDa cutoff (Millipore) prior to purification by IMAC, as described above.

Protein quantitation and SDS-PAGE.

Protein purity was assessed by SDS-PAGE using 4 to 15% acrylamide MiniPROTEAN gels (Bio-Rad) with Laemmli’s reagent for denaturation and loading and Tris-glycine-SDS as the running buffer. Proteins were quantified using the Pierce bicinchoninic acid (BCA) protein assay kit, according to the manufacturer’s instructions. Bovine serum albumin was used as a protein quantitation standard.

Determination of substrate specificity and specific activity.

Reducing-assay-grade wheat arabinoxylan (WAX), BeX, beet arabinan (BA), carboxymethylcellulose (CMC), barley β-glucan (BBG), and tamarind xyloglucan (TXG) were purchased from Megazyme (Wicklow, Ireland). Oat spelt xylan (OSX), birchwood xylan (BiX), and the para-nitrophenol (pNP)-linked substrates pNP–β-xylopyranoside, pNP–α-arabinofuranoside, pNP–β-glucoside, pNP–β-galactoside, pNP–β-glucuronide, and pNP–β-mannoside were purchased from Sigma (St. Louis, MO). All activity assays were performed in 50 mM sodium phosphate (pH 6.5) at 75°C to screen for substrate specificity with polysaccharide substrates loaded at 1% (wt/vol) and pNP-linked substrates loaded at 4 mM. Polysaccharide substrates were incubated for 30 min, while pNP-linked substrates were incubated for 5 min, with 0.5 μM enzyme used in both cases. Activity on polysaccharides was assessed using the modified dinitrosalicylic acid (DNSA) method adapted for microtiter plates (39, 41), as previously described (12). Xylose standards were used for the quantitation of reducing sugars. Analysis of the specific activity of pNP-linked substrates was performed discontinuously. Briefly, after incubation under the target reaction conditions, reaction mixtures were quenched using an equal volume of 1 M sodium carbonate (pH 10). The release of pNP was quantified at 405 nm using pNP standards subjected to the same buffer conditions; the estimated extinction coefficient was determined to be 22,765 M⁻¹ cm⁻¹ for this buffering system.

Determination of the optimal pH and temperature for XynC.

The optimal pH was determined from a pH range of 4 to 10 in 0.5-U increments at 75°C. For buffers between pH 4 and 5.5, 50 mM sodium acetate was used; for buffers between pH 6.0 and 8.0, 50 mM sodium phosphate was used; and for buffers between pH 8.5 and 10, 50 mM sodium bicarbonate was used. The optimal temperature was determined using 5°C increments from 40°C to 95°C at the optimal pH of the enzyme. The enzyme was loaded at 0.5 μM, with 1% (wt/vol) beechwood xylan (Megazyme) as the test substrate. All reactions were performed in triplicate.

Determination of XynC thermostability.

Proteins were diluted to 1 μM in 50 mM sodium phosphate (pH 6.5) and then incubated at 75°C or 85°C. Samples were periodically pulled and assayed at 75°C using a 1% (wt/vol) final concentration of beechwood xylan as the substrate for 30 min. The relative activity was compared to the activity of the protein that had been at room temperature for 15 min and then incubated with 1% beechwood xylan at 75°C for 30 min. The protein concentration used in the assays was targeted to be 0.5 μM.

Determination of synergistic interactions between glycoside hydrolases.

Enzymatic reactions were conducted as described above except using 0.2 μM total enzyme. Individual and equimolar enzyme pairs (0.1 μM each) were tested. All reactions were conducted at 75°C for 30 min, with released sugars being measured by the DNSA method as described above. The degree of synergy (DOS) was determined by the formula DOS = AB/(0.5A + 0.5B), where AB is the sugar released from the equimolar mixtures and A and B are the sugars released from the individual enzymatic components with the individual components multiplied by 0.5 to represent the molar fraction of each pair in the assay.

Determination of hydrolysis products.

Enzymatic reaction mixtures were filtered through a 0.22-μm filter. Ten microliters of the filtered reaction mixture was injected onto a Shodex SP 0810 sugar column using a Waters Arc3 high-performance liquid chromatography (HPLC) system. The mobile phase used was 0.6 mL/min water at 85°C. A Waters 2414 refractive index (RI) detector was used for detection. This method was able to resolve the following compounds: d-xylose, β-1,4-d-xylobiose, β-1,4-d-xylotriose, and β-1,4-d-xylotetrose.

Transglycosylation reactions.

Samples (0.2 μM) of either XynD (Athe_0184) or Xyl3A (Athe_2354) were incubated at 75°C with 5 mM substrate. The following substrates were tested for transglycosylation: pNP–β-1,4-d-xylopyranoside, d-xylose, β-1,4-d-xylobiose, β-1,4-d-xylotriose, and β-1,4-d-xylotetrose. Reaction mixtures were incubated for 4 h and then quenched using wet ice. The products were then resolved as described above.

Operon profiling.

C. bescii was grown in DSM516 medium as described above, with 5 g/L d-xylose or l-arabinose as the carbon source. Cells were quenched at room temperature during early stationary phase, and RNAs were extracted as described above. Purified RNA was then treated with Turbo DNase (Thermo Fisher), according to the manufacturer’s instructions, and reverse transcribed using the iScript reverse transcription supermix (Bio-Rad). The cDNA was used as a template for PCR to amplify the intergenic regions using primers anchored to the reading frame (see Table 1 for primer sequences). C. bescii genomic DNA and no-RT iScript supermix were used as PCR controls. PCR was performed with Q5 polymerase using standard buffer (New England BioLabs). In silico terminator prediction was performed using the ARNold Web server by taking the sequence spanning 200 bp before the first coding sequence in a target genomic locus and ending 200 bp after the last coding sequence. Promoter prediction was performed using PePPER with the same genomic regions as those used for terminator prediction.

Bioinformatic analyses.

Full genome sequences for the following species of Caldicellulosiruptor were downloaded from NCBI Reference Sequence (RefSeq) database: C. bescii DSM 6725, C. kronotskyensis 2002, C. saccharolyticus DSM 8903, C. owensensis OL, C. acetigenus DSM 7040, C. lactoaceticus 6A, C. kristjanssonii I77R1B, C. changbaiensis CBS-Z, C. diazotrophicus YA01, C. obsidiansis OB47, C. naganoensis NA10, C. danielii (Caldicellulosiruptor sp. strain Wai35.B1), C. morganii (Caldicellulosiruptor sp. Rt8.B8), C. hydrothermalis 108, and Caldicellulosiruptor sp. F32. Genome-wide carbohydrate-active enzyme (CAZyme) family annotations were assigned using dbCAN to ensure consistency between genomes deposited in the CAZy database and those not in the CAZy database (42). Outputs from dbCAN were manually curated to include only GH-containing proteins. These outputs were further manually curated against GH families that have hemicellulase activity, consisting of GH families 2, 3, 10, 11, 39, 43, 51, and 67, to create FASTA files of all putative hemicellulases. The core and paninventories of hemicellulases were determined by running the curated hemicellulase FASTA files through get_homologues (38) using the available COGtriangles algorithm. Transcriptional rank order and fold induction analyses of the xylanolytic clusters were performed using previously deposited RNA sequencing data (NCBI Gene Expression Omnibus database accession number GSE163475) from Rodionov et al. (25).