Chemicals for cultivation of Escherichia coli and Thermococcus sp. strain 2319x1E, such as yeast extract, lysogeny broth (LB), beechwood xylan, D-glucose and D-xylose were obtained from Carl Roth (Germany). Avicel^® cellulose, carboxymethyl cellulose (CMC), maltodextrin and soluble starch were purchased from Sigma-Aldrich (United States), xylobiose, ortho-nitrophenyl-β-D-galactopyranoside (ONPG), para-nitrophenyl acetate (PNPA), para-nitrophenyl-β-D-xylopyranoside (PNPX) and para-nitrophenyl- β-D-glucopyranoside (PNPG) from Megazyme (Ireland). Other chemicals which were used in this work are 4’,6-diamidino-2-phenylindole (DAPI; Thermo Fisher Scientific, United States), urea (GE Healthcare Life Sciences, Germany), ammonium bicarbonate (ABC; Sigma-Aldrich, United States), dithiothreitol (DTT; Sigma-Aldrich, United States), iodoacetamide (IAM; Sigma-Aldrich, United States), formic acid (FA; Fischer Chemical, Germany) and bovine serum albumin (BSA; VWR Chemicals, United States).

Thermococcus sp. strain 2319x1E was sequenced based on short-read and long-read sequencing methods. In a first step, whole-genome-shotgun PCR-free libraries for Illumina sequencing were constructed from 2 μg of gDNA with the Illumina TruSeq DNA PCR-free Sample Preparation Kit (Illumina, United States) based on the manufacturer’s protocol. The libraries were quality controlled by analysis on an Agilent 2000 Bioanalyzer with Agilent High Sensitivity DNA Kit (Agilent Technologies, United States) for fragment sizes of 550 bp. Sequencing was performed on a Illumina MiSeq (2 × 300 bp/v3 chemistry) in paired-end mode. Adapters and low-quality reads were removed by an in-house software pipeline prior to polishing as recently described (Wibberg et al., 2016).

In a second step, the Oxford Nanopore sequencing library from Thermococcus sp. strain 2319x1E gDNA was prepared using the Nanopore Native barcoding genomic DNA kit with native barcodes (SQK-LSK109 with EXP-NBD104) according to the manufacturer’s instructions. The resulting libraries were sequenced on an Oxford Nanopore GridION Mk1 sequencer using an R9.4.1 flow cell, which was prepared according to the manufacturer’s instructions. MinKNOW was used to control the run using the 24 h sequencing run protocol; live base calling was performed using guppy v3.2.6. The assembly was performed as described recently with smaller modifications - Unicycler v0.4.6. was used for the hybrid-assembly and no additional polishing was performed (Wibberg et al., 2020). The finished genome sequence was imported into the annotation platforms Prokka (Seemann, 2014) and GenDB (Meyer et al., 2003) for automatic prediction of genes and functional annotation as described previously (Tomazetto et al., 2018).

For an extensive phylogenic analysis, all available genomes of cultivated species of Thermococcus, Pyrococcus, Palaeococcus, and Methanothermobacter thermautotrophicus Delta H (used as an outgroup) were downloaded from GenBank. In total, 122 archaeal-specific conserved proteins were identified from 34 genomes and aligned using gtdb-tk v.1.0.2 (Parks et al., 2018; Chaumeil et al., 2019). The concatenated alignment was refined using Gblock v.0.91b with the most gentle parameters and complete gap elimination option (Castresana, 2000). The final alignment contained 27,083 amino acids. A phylogenetic tree was constructed in RAxML v.8.2.12 (Stamatakis, 2014) with the PROTGAMMAILG model of amino acid substitutions. Local support values were 1000 rapid bootstrap replications.

Carbohydrate-active enzyme (CAZyme) genes were searched with dbCAN2 (Zhang et al., 2018) using HMMER (Potter et al., 2018) with default parameters. For selected CAZymes, phylogenetic analysis was performed as follows: the top 5 BLAST homologs for each sequence and all proteins from the Swiss-Prot database (evidence at protein level only) affiliated to the corresponding family were aligned in Muscle (Edgar, 2004). Phylogenetic trees were constructed in RAxML v.8.2.12, as described above (Phylogenetic analysis section). Manual curation of protein function prediction was conducted using BlastP (McGinnis and Madden, 2004), HHpred (Zimmermann et al., 2018), HHMMER (phmmer and hmmscan) (Potter et al., 2018), and SUPERFAMILY 2 (Pandurangan et al., 2019). Secretion signal peptides and transmembrane domains were predicted using SignalP (Almagro Armenteros et al., 2019), Phobius (Käll et al., 2007), and TMHMM (Krogh et al., 2001). Search for orthologous gene clusters in Thermococcus sp. strains 2319x1 and 2319x1E was performed with OrthoVenn2 (Xu et al., 2019).

Thermococcus sp. strain 2319x1E was isolated from the same in situ enrichment as the previously described Thermococcus sp. strain 2319x1 (Gavrilov et al., 2016). Cultivation as well as growth experiments were conducted in a strictly anaerobic modified Pfennig medium, consisting of 0.33 g l^–1 MgCl₂, CaCl₂, KCl, NH₄Cl and KH₂PO₄, 9 g l^–1 NaCl, 2 g l^–1 NaHCO₃, 16 g l^–1 Na₂S ⋅ 9 H₂O, 0.5 ml l^–1 trace elements (Kevbrin and Zavarzin, 1992), 1 ml l^–1 vitamin solution (Wolin et al., 1963), 1 g l^–1 sulfur, 0.1 g l^–1 yeast extract and 1 g l^–1 of the respective growth substrate (beechwood xylan, Avicel^® cellulose, D-glucose, D-xylose). All growth experiments were performed at 85°C and pH 7.0 under a N₂-atmosphere, either in 500 ml Schott bottles without shaking or in a 10 l bioreactor (Bioengineering, Switzerland) with constant stirring (100 rpm) and N₂ infusion (0.5 l min^–1). Growth was determined by counting cells in a Neubauer improved counting chamber (Brand, Germany) on a Zeiss Axioscope microscope (Zeiss, Germany). Cells that were further used for assays, ABPP experiments or proteomic analysis, were grown on the respective substrate for at least three consecutive transfers to the same fresh medium.

Thermococcus sp. strain 2319x1E cells were collected from grown cultures by centrifugation at up to 8,000 × g for 10 min at 25°C, washed two times with 1× PBS (8 g l^–1 NaCl, 0.2 g l^–1 KCl, 1.44 g l^–1 Na₂HPO4, 0.24 g l^–1 KH₂PO₄, pH 7.0) and stored at –80°C. After thawing, the cell pellets were resuspended in 1× PBS pH 7.0 (2 ml per gram wet weight) and sonicated with a UP 200S sonicator (Hielscher Ultrasonics GmbH, Germany) for 3× 5 min (cycle 0.5, amplitude 50) followed by centrifugation at 12,000 × g for 30 min at room temperature to remove intact cells and cell debris. To separate the cytosolic protein fraction from the membrane protein fraction, ultracentrifugation of the crude protein extract was done (30 min, 100,000 × g, 4°C). To retrieve the respective membrane fraction, the pellet was resuspended in 1 × PBS pH 7.0 after collecting the cytosolic protein-containing supernatant. Protein concentrations were determined using the Bradford assay (Bio-Rad, United States) with bovine serum albumin (BSA) as standard (Bradford, 1976).

For determining glucosidase activities of Thermococcus sp. strain 2319x1E from either crude extracts or cytosolic and membrane fractions, the 3,5-dinitrosalicylic acid (DNSA) assay was applied (Miller, 1959). The enzyme assays were performed in 50 mM Tris-HCl pH 7.0 with a total reaction volume of 550 μl, containing 250 μl substrate solution in H₂O (1% w/v for xylan and CMC, 0.04% w/v for xylobiose). The reactions were started by the addition of 7.5 μg of protein from the respective sample followed by incubation for 30 min at 85°C. The reaction mixture was cooled on ice for 5 min, mixed with 700 μl DNSA reagent (Miller, 1959) and subsequently incubated at 100°C for 10 min. The concentration of reducing sugars was then calculated from the absorption of the samples at 575 nm by using a D-xylose or D-glucose calibration curve. Background absorption of cell extracts and substrate solution was subtracted from the absorption of the samples. One unit of enzyme activity was defined as the amount of enzyme required to release 1 μmol D-xylose equivalents per minute. All absorbance measurements were done in 96-well plates using a Tecan infinite M200 plate reader (Tecan trading AG, Switzerland).

The Thermococcus sp. strain 2319x1E genes EGIDFPO_00674 (1800 nt) and EGIDFPO_00532 (2196 nt) were amplified from gDNA extracted with the DNeasy Blood & Tissue Kit (Qiagen, Germany) using Q5^® polymerase (NEB, United States) and the following gene specific primers (Eurofins Genomics, Germany): for EGIDFPOO_00674 5′-GCGGCCCATGGGCTACCAGAAGTTTGGATATCATTTTCA TGC-3′ and 5′-ATTAGAATTCCTAGTGGTGGTGGTGGTGGT GAACCCTGGCCCTATGCTCGCACCATT-3′ (NcoI and EcoRI restriction sites underlined); for EGIDFPOO_00532 5′-CTGGTGCTAGCCTGATGCTTGTCTTTCCCGATTCCTTCC TCT-3′ and 5′-CTATCCTCGAGTCATAGCTCCGGAAGTCCAT ACTTTG-3′ (NheI and XhoI restriction sites underlined). PCR products were afterwards purified using the Wizard^® SV Gel and PCR clean-up kit (Promega, United States).

The genes were cloned into the pET-28b(+) vector (Novagen, United States) for subsequent expression, providing the recombinant proteins with a C-terminal (EGIDFPOO_00674) or N-terminal 6× His-tag. After restriction digest of the purified PCR products and the empty vector with the respective restriction enzymes (NEB), PCR product and vector were used in a molar ratio of 1:6 for sticky end ligation using T4 DNA ligase (NEB) at 16°C over night. E. coli DH5α cells (Novagene, United States) were transformed with the obtained constructs and the presence of successfully cloned genes was confirmed by colony PCR with the transformed cells as template and T7 promotor and terminator primers. The correctness of the sequence was confirmed by Sanger sequencing of both strands (Eurofins Genomics, Germany). The codon optimized gene (E. coli) EGIDFPOO_00563 was synthesized and cloned into a pET28b(+) vector with N-terminal 6× His-tag (BioCat GmbH, Germany).

Recombinant expression of the glycoside hydrolases EGIDFPOO_00674, EGIDFPOO_00532 and EGIDFPOO_00563 was performed in E. coli Rosetta cells (Novagen, United States) which were transformed with the respective plasmids. A freshly inoculated 1 l culture in LB medium, supplemented with 50 μg ml^–1 kanamycin and 50 μg ml^–1 chloramphenicol, was grown to an OD₆₀₀ of 0.4 at 37°C with constant shaking (180 rpm) upon subsequent induction of protein expression with 500 μM isopropyl-β-D-thiogalactopyranoside (IPTG). The cells were incubated at 18°C for a further 16 h, harvested by centrifugation (8,000 × g for 20 min at 4°C) and resuspended in 3–5 mL of buffer (50 mM Tris-HCl pH 7.0) per gram wet weight of the pellet for further protein purification. Cell lysis was performed by sonication as described above, followed by centrifugation at 12,000 × g for 45 min at 4°C. For further purification of EGIDFPOO_00674 or EGIDFPOO_00563, the cleared lysate was passed through a 0.45 μm filter, followed by an affinity purification using a 5 ml Ni-IDA column (GenScript, GenScript Biotech, United States) equilibrated with buffer A (50 mM Tris-HCl pH 7.8, 250 mM KCl, 10 mM imidazole). Protein elution was performed using a linear gradient of 10–350 mM imidazole in buffer A. The elution buffer was exchanged for storage buffer (50 mM Tris-HCl pH 8.0, 20 mM KCl, 5 mM MgCl₂) using Amicon^® centrifugal filter devices (30 or 50 kDa cutoff, Merck, Germany). For long-time storage at –80°C, 50% (v/v) glycerol was added and the protein solution was flash frozen in liquid N₂. EGIDFPOO_00532 was purified from inclusion bodies using a high pH-buffer purification strategy (Singh et al., 2015). The pellet obtained after sonication was washed with 50 mM Tris-HCl pH 8.0, taken up in 7.5 ml resuspension buffer (50 mM Tris-HCl pH 12.5, 2 M urea) per gram (wet weight) and subsequently incubated for 30 min at room temperature. The solubilized pellet was then centrifuged (8,000 × g, 10 min, 4°C) and the supernatant was diluted stepwise 1:20 into refolding buffer [50 mM Tris-HCl pH 8.0, 2 M urea, 10% (w/v) sucrose] on ice. Afterwards, a heat precipitation (20 min, 80°C) with subsequent centrifugation (8,000 × g, 10 min, 4°C) was performed. The obtained supernatant was directly used for further enzyme assays.

The activity of the heterologously expressed proteins EGIDFPOO_00532 and EGIDFPOO_00674 was determined using a discontinuous assay with the colorimetric nitrophenyl substrates PNPA, PNPX, PNPG, OPNG. Upon enzymatic cleavage of the nitrophenyl substrates, the absorbance of the free nitrophenol was measured at 400 nm in 96-well plates using a plate reader. Absorbance values of controls lacking the enzyme were subtracted from the absorbance values of the samples. Enzyme activity was calculated from a calibration curve with para-nitrophenol (PNP). The pH-dependent activity was measured at 348 nm in McIlvaine buffers with the respective pH values (McIlvaine, 1921) as described previously (Kallnik et al., 2014). Deacetylase activity of EGIDFPOO_00674 was measured in 50 mM Tris-HCl pH 8.0 using 2.5 μg of the purified enzyme and up to 8.8 mM PNPA in a total sample volume of 250 μl. α-amylase activity of EGIDFPOO_00674 was determined in 50 mM Tris-HCl pH 8.0 using the DNSA assay as described above, with 10 μg of the purified enzyme and 3 mM 6-O-α-maltosyl-β-cyclodextrin (MβCD) per 550 μl reaction volume. The β-glycosidase activity of EGIDFPOO_00532 was assessed in refolding buffer containing 25 μg of the purified enzyme and up to 26.7 mM of the respective PNP-substrate. Kinetics were determined at the optimal pH and temperature which was pH 8.0 and 100°C for both enzymes. To assess the thermostability of the enzymes, the PNP-substrate assay was performed as described above, but the enzymes were preincubated at 60, 80, and 100°C for up to 4 h. The glycogen phosphorylase activity of EGIDFPOO_00563 was assayed using a discontinuous enzyme assay; First, 500 μl samples containing 50 mM Tris-HCl pH 7, 20 μg purified EGIDFPOO_00563, 10 mM NaH₂P0₄ and 0.05% (w/v) substrate (glycogen, maltodextrin or starch) were incubated for 5 min at 100°C. Subsequently, the samples were stored on ice for 5 min. The amount of glucose 1-phosphate formed in the glycogen phosphorylase reaction was then determined by the indicator reaction, containing 4 U phosphoglucomutase (Sigma-Aldrich, United States), 3 U glucose-6-phosphate dehydrogenase (Roche, Switzerland), 10 mM MgCl₂, and 0.5 mM NADP⁺ in 50 mM Tris-HCl pH 7 (total volume of 500 μl). The assay was incubated for 3 min at 30°C and the amount of glucose 1-phosphate was calculated from the absorption at 340 nm using a Specord 210^® photometer (Analytik Jena, Germany).

Protein solutions were diluted with their respective buffer to a final protein concentration of 0.02 mg protein ml^–1 for native crude extract and 0.03 mg protein ml^–1 for heterologously expressed purified protein. To examine activity inhibition upon covalent binding of the glycosidase probes JJB384 or JJB111, the samples (containing heterologously expressed proteins or Thermococcus sp. Strain 2319x1E crude extract) were incubated with 2 and 4 μM JJB384 or JJB111, respectively, for 1 h at 85 °C and subsequently used for activity determination assays with DNSA or nitrophenyl substrates.

Cells for proteomics studies were grown in Schott bottles as described above (Cultivation and Growth Experiments). For every carbon source under evaluation (beechwood xylan, Avicel^® cellulose, D-glucose, D-xylose), four biological replicates have been prepared. Therefore, 200 ml growth medium supplemented with the respective substrate were inoculated from a preculture which has been grown on that substrate for at least three consecutive transfers. Cells were harvested from cultures by centrifugation up to 8,000 × g for 20 min at 4°C after 10–12 h of growth during early mid-log phase. For storage at –80°C, the pellets were resuspended in 2 ml of the culture supernatant, transferred to fresh 15 ml Falcon tubes and centrifuged 6,500 × g for 10 min at 4°C. After removing the supernatant, the pellets were directly flash frozen in liquid N₂.

Cell pellets were resuspended in an adequate volume (approx. 2000–4000× less than the original culture volume) of 50 mM Na₂HPO₄ pH 8.0 supplemented with 1× MS-SAFE Protease and Phosphatase Inhibitor (Sigma-Aldrich, United States) and transferred to 1.5 ml reaction tubes. Cell lysis was performed by conducting a two-step sonication protocol. First, the cells were sonicated 7× 1 min in an ultrasonic bath filled with ice with short breaks in between to vortex the samples, followed by sonication with a Bioruptor UCD-200 (Diagenode, Belgium) using the following conditions: 1 min pulse and 30 s break in 10 cycles with the intensity set to “high” while cooled with ice. The extracts were cleared twice by centrifugation (21,000 × g, 20 min and 5 min, 4 °C) to remove cell debris and the protein concentration of the cleared extracts was determined by a modified Bradford assay with ROTI^®Nanoquant (Carl Roth, Germany).

For each sample, 15 μg of total protein were subjected to a methanol-chloroform precipitation (Wessel and Flügge, 1984). The pellet was washed with methanol twice, dried on air and resuspended in 25 μl 8 M urea in 100 mM ABC, followed by reduction of disulfide bonds with 5 mM DTT in 100 mM ABC and incubation at 23°C for 30 min shaking at 1,000 rpm. Subsequently, alkylation of cysteine residues was done using 20 mM IAM in 100 mM ABC and incubation at 37°C for 30 min shaking in the dark at 1,000 rpm. The reaction was quenched by adding DTT to a final concentration of 25 mM. Afterwards, the proteins were digested with 500 ng Lys-C (FUJIFILM Wako Chemicals, Japan) dissolved in 100 mM ABC for 3 h at 37°C shaking at 1,000 rpm, followed by overnight digestion (∼16 h) with 500 ng trypsin (Thermo Scientific, United States) dissolved in 50 mM acetic acid at 37°C shaking at 1,000 rpm. Prior to sample clean-up for LC-MS/MS, formic acid (FA) was added to the samples to a final concentration of 5% (v/v).

In variation, the pellet of the samples from cells grown on xylan was washed once with 10% DMSO/10% ethanol and thrice with methanol after precipitation. Furthermore, the samples were dried in a vacuum concentrator (Eppendorf, Germany) after acidification and the peptides were dissolved in 0.5% (v/v) FA prior to desalting.

All chemical probes were dissolved in DMSO. Labeling of active glycoside hydrolases was performed using a 2 ml liquid culture which was obtained by concentration of the growth culture (50× for samples analyzed on gel; 500× for samples dedicated for MS-based target identification) by centrifugation at up to 8,800 × g for 20 min at 25°C. For labeling, 2 μM of JJB384 or JJB111 were added to the liquid culture, followed by incubation at 78°C for 2 h shaking at 180 rpm. Afterwards, the cells were harvested by centrifugation at 8,800 × g for 10 min at 25°C. Protein extraction was performed as described above. For MS-based target identification, each sample was prepared in triplicates.

Labeled protein extracts (10–15 μg protein) were mixed with 4× lithium dodecyl sulfate (LDS) gel loading dye [423 mM Tris-HCl, 563 mM Tris base, 8% (w/v) LDS, 40% (w/v) glycerol, 2 mM EDTA, 0.075% (w/v) SERVA Blue G250; freshly supplemented with 100 mM DTT], incubated at 70°C for 15 min and separated by denaturing polyacrylamide gel electrophoresis on 11% Bis-Tris resolving gels. The separated proteins were transferred on a PVDF membrane (Merck, United States) using a tank blot setup (Bio-Rad, United States) and the membrane was washed thrice with TBS-T [20 mM Tris base pH 7.5, 150 mM NaCl, 0.2% (w/v) Tween20]. Blocking with 3% BSA (w/v) in TBS-T was done overnight at 4°C, followed by incubation with a streptavidin-horse radish peroxidase (HRP) conjugate (Sigma-Aldrich, United States) directly added into the blocking solution (1:25,000) for 2.5 h at room temperature. The membrane was washed 6× with TBS-T and the labeled proteins were detected by enhanced chemiluminescence (ECL) with a mix of the SuperSignal^® West Pico Chemiluminescent Substrate and the SuperSignal^® West Femto Maximum Sensitivity Substrate (4:1; Thermo Scientific, United States) using an Amersham Imager 600 (GE Healthcare Life Sciences, Germany).

Prior to affinity enrichment, the protein solution (225 μg protein in a final volume of 500 μl 50 mM Na₂HPO₄ pH 8.0 supplemented with 1× MS-SAFE Protease and Phosphatase Inhibitor) was cleaned-up by precipitation with a 4× volume of methanol overnight at –20°C. The proteins were collected by centrifugation at 21,000 × g for 10 min at 4°C and washed with a 2× volume of methanol. The air-dried pellets were dissolved in 850 μl 2% (w/v) SDS in 1× PBS (155 mM NaCl, 3 mM Na₂HPO₄, 1.06 mM KH₂PO₄, pH 7.4) by incubation at 37°C with agitation at 1,000 rpm and diluted with 1× PBS pH 7.4 to a final concentration of 0.2% (w/v) SDS. The resulting protein solutions were incubated with avidin agarose beads (Thermo Scientific, United States), equilibrated in the same buffer, for 1 h at room temperature while gently tumbling. The beads were collected by centrifugation (400 × g, 5 min), washed 5× with 1% (w/v) SDS in MS-quality water and finally 4× with pure MS-quality water. Afterwards, tryptic on-bead-digestion was conducted. Thereto, the beads were taken up in 100 μl 0.8 M urea in 50 mM ABC buffer. Reduction of disulphide bonds with 10 mM DTT in 50 mM ABC was done by incubation at room temperature for 1 h shaking at 1,500 rpm, followed by alkylation of cysteine residues with 25 mM IAM in 50 mM ABC and incubation at room temperature for 1 h in the dark with shaking at 1,500 rpm. The reaction was quenched by increasing the DTT concentration to 35 mM and further 10 min incubation. For digestion of proteins, 10 μl of a 100 ng μl^–1 trypsin stock solution in 50 mM acetic acid were added and the samples were incubated at 37°C for 16 h with shaking at 1,250 rpm. The beads were collected by centrifugation at 3,000 × g for 5 min and FA was added to the recovered supernatants to a final concentration of 5% (v/v). The beads were washed with 50 μl 1% (v/v) FA at room temperature for 15 min shaking at 1,500 rpm and collected by centrifugation at 3,000 × g for 5 min. Both supernatants were combined and subsequently cleared by passing over home-made tips containing two disks of glass microfiber (GE Healthcare, Life Sciences, Germany; poresize 1.2 μm; thickness 0.26 mm).

Peptides were desalted on home-made C₁₈ StageTips (Rappsilber et al., 2007) containing two layers of an octadecyl silica membrane (3M, United States). All centrifugation steps were carried out at room temperature. The StageTips were first activated and equilibrated by passing 50 μl of methanol (600 × g, 2 min), 80% (v/v) acetonitrile (ACN) with 0.5% (v/v) FA (600 × g, 2 min) and 0.5% (v/v) FA (800 × g, 3 min) over the tips. Next, the tryptic digests were passed over the tips (800 × g, 3–4 min). The flow through was collected and applied a second time (same settings). The immobilized peptides were then washed with 50 μl and 25 μl 0.5% (v/v) FA (800 × g, 3 min). Bound peptides were eluted from the StageTips by an application of two rounds of 25 μl 80% (v/v) ACN with 0.5% (v/v) FA (600 × g, 2 min). After elution from the StageTips, the peptide samples were dried using a vacuum concentrator (Eppendorf, Germany) and the peptides were dissolved in 15 μl 0.1% (v/v) FA prior to analysis by MS.

Experiments were performed on an Orbitrap Elite or Orbitrap Fusion Lumos mass spectrometer (Thermo Fischer Scientific, United States) that were coupled to an EASY-nLC 1000 or 1200 liquid chromatography (LC) system (Thermo Fischer scientific, United States). The LCs were operated in the one-column mode. The analytical column was a fused silica capillary (inner diameter 75 μm × 36 or 46 cm) with an integrated PicoFrit emitter (New Objective, United States) packed in-house with Reprosil-Pur 120 C18-AQ 1.9 μm (Dr. Maisch, Germany). The analytical column was encased by a column oven (Sonation, Germany) and attached to a nanospray flex ion source (Thermo Fischer scientific, United States). The column oven temperature was adjusted to 45 or 50°C during data acquisition. The LC was equipped with two mobile phases: solvent A (0.1% FA, in water) and solvent B (nLC 1000: 0.1% FA in ACN; nLC 1200: 0.1% FA, 20% H₂O, in ACN). All solvents were of UHPLC (ultra-high-performance liquid chromatography) grade (Honeywell, Germany). Peptides were directly loaded onto the analytical column with a maximum flow rate that would not exceed the set pressure limit of 980 bar (usually around 0.5 – 0.8 μl min^–1). Peptide solutions were subsequently separated on the analytical column using different gradients (for details see Supplementary Data Sheet 2).

The mass spectrometers were operated using Xcalibur software (Elite: v2.2 SP1.48; Lumos: v4.3.7.3.11). The mass spectrometers were set in the positive ion mode. Precursor ion scanning (MS¹) was performed in the Orbitrap analyzer (FTMS; Fourier Transform Mass Spectrometry with the internal lock mass option turned on (lock mass was 445.120025 m/z, polysiloxane) (Olsen et al., 2005). MS² Product ion spectra were recorded only from ions with a charge higher than +1 and in a data dependent fashion in the ion trap mass spectrometry. All relevant MS settings (Resolution, scan range, AGC, ion acquisition time, charge states isolation window, fragmentation type and details, cycle time, number of scans performed, and various other settings) for the individual experiments can be found in Supplementary Data Sheet 2.

RAW spectra were submitted to an Andromeda search (Cox et al., 2011) in MaxQuant (v.1.6.2.6) using the default settings (Cox and Mann, 2008). Label-free quantification (LFQ) (Cox et al., 2014) and match between runs was activated. MS/MS spectra data were searched against the self-constructed TspE_proteome_AA_new.fasta file (2182 entries; see above for reference). All searches included a contaminants database (as implemented in MaxQuant, 246 sequences). The contaminants database contains known MS contaminants and was included to estimate the level of contamination. Andromeda searches allowed oxidation of methionine residues (16 Da) and acetylation of the protein N-terminus (42 Da) as dynamic modifications while carbamidomethylation of cysteine residues (57 Da, alkylation with IAM) was selected as static modification. Enzyme specificity was set to “Trypsin/P.” The instrument type in Andromeda searches was set to Orbitrap and the precursor mass tolerance was set to ±20 ppm (first search) and ±4.5 ppm (main search). The MS/MS match tolerance was set to ±20 ppm. The peptide spectrum match FDR and the protein FDR were set to 0.01 (based on target-decoy approach). Minimum peptide length was 7 amino acids. For protein quantification, unique and razor peptides were allowed. In addition to unmodified peptides, modified peptides with dynamic modifications were allowed for quantification. The minimum score for modified peptides was set to 40.

Further data analysis and filtering of the MaxQuant output was done in Perseus (v.1.6.2.1) (Tyanova et al., 2016). LFQ intensities were loaded into the matrix from the proteinGroups.txt file and potential contaminants as well as reverse hits from the reverse database and hits only identified based on peptides with a modification site were removed. For all MS-based proteomics experiments, biological replicates were combined into categorical groups to allow comparison of the different treatment groups and the data was transformed to the log₂-scale. For the full proteome analysis, the data was filtered to keep only hits with a valid LFQ intensity for at least three out of four replicates for samples from cells grown on xylan. For the identification of glycoside hydrolases enriched with JJB384, the data was separately filtered for the two sugar substrates under investigation. Only hits with a valid LFQ intensity for a minimum of two out of three of the probe-labeled sample replicates were retained for further analysis. Prior to quantification, missing values were imputed from a normal distribution (width 0.3, down shift 1.8). Comparison of normalized protein group quantities (relative quantification) between different MS runs was solely based on the LFQ intensities as calculated by MaxQuant (MaxLFQ algorithm) (Cox et al., 2014). Briefly, label-free protein quantification was switched on and unique and razor peptides were considered for quantification with a minimum ratio count of 2. Retention times were recalibrated based on the built-in non-linear time-rescaling algorithm. MS/MS identifications were transferred between LC-MS/MS runs with the “Match between runs” option in which the match time window was set to 0.7 min and the alignment time window to 20 min. The quantification was based on the “value at maximum” of the extracted ion current. At least two quantitation events were required for a quantifiable protein. The log₂-fold enrichment of protein groups for samples from cells grown on xylan was calculated based on the mean LFQ intensity compared to the samples from cells grown on any of the other sugars. Statistical significance was calculated based on a two-sided Student’s t-test. Full MS data for the comparative full proteome analysis can be found in Supplementary Table 1. The log₂-fold enrichment of protein groups with JJB384 was calculated based on the mean LFQ intensity compared to the DMSO control. Protein groups with a negative fold enrichment were excluded from further analysis. Full MS data for the ABPP with JJB384 can be found in Supplementary Table 2.

Thermococcus sp. strain 2319x1E was isolated from the same enrichment culture as Thermococcus sp. strain 2319x1 (Gavrilov et al., 2016). The fully assembled genome of Thermococcus sp. strain 2319x1E consists of a single chromosome with the size of 1,989,851 bp and an average GC content of 44.39%. The sequence is deposited in the NCBI genome database under accession number LR778300. The average nucleotide identity (ANI) to the genome of Thermococcus sp. strain 2319x1 is 97.86%, thus being above the species level (95%) according to Goris et al. (2007), whereas the ANI to Thermococcus litoralis, which is considered the closest validly published Thermococcus species described, is only 90.01% and to the more distantly related Thermococcus kodakarensis 71.97%. The phylogenetic position of both Thermococcus isolates was determined using phylogenetic tree construction based on the “ar122” set of conserved proteins (Figure 2).

The predicted number of protein-coding sequences (CDS) in Thermococcus sp. strain 2319x1E is 2,195, and therefore is in a similar range as the 2319x1 genome (2,192 CDS). The comparison of both genomes revealed 1,886 single-copy orthologous gene clusters found in both strains and 266 and 360 unique genes in the genome of 2319x1E and 2319x1, respectively. Altogether, these numbers indicate that the two Thermococcus strains are relatively closely related to each other. With regard to enzymes involved in xylan degradation, the CAZymes present in both strains were compared. The Thermococcus sp. strain 2319x1E genome comprises 42 genes which putatively encode for CAZymes (Lombard et al., 2014), with 26 of them being predicted to encode for glycosyl transferases (GT), 13 for glycosyl hydrolases (GH) and 3 for carbohydrate esterases (CE) (Table 1). Remarkably, the genes encoding the unique multifunctional multidomain glycosidase described for the Thermococcus sp. strain 2319x1 isolate (Gavrilov et al., 2016) and 4 additional GHs, as well as 8 GTs are not present in the genome of the new isolate 2319x1E. Correspondingly, in the Thermococcus sp. strain 2319x1, genes encoding 9 GTs and 2 GHs (EGIDFPOO_01323 and EGIDFPOO_01324) that are present in strain 2319x1E are missing (Supplementary Table 1 in Supplementary Data Sheet 1).

Of the CAZymes identified in Thermococcus sp. strain 2319x1E, two putative GHs encoded by the genes EGIDFPOO_01845 and EGIDFPOO_01993 were predicted to be extracellular (Table 1), although it should be noted that the reliable prediction of archaeal secretion signals remains challenging (Bagos et al., 2009; Szabo and Pohlschroder, 2012). In addition, 11 putative GTs were predicted to be membrane-associated (Table 1) due to the presence of one or more predicted transmembrane domains. Among the 12 putative GHs, 7 are most likely involved in the hydrolysis of α-linked oligo- and polymers such as glycogen, starch, pullulan and maltooligosaccharides, as determined by BLAST against the Swiss-Prot database as well as phylogenetic analyses for some families (GH13 and GH57; Supplementary Figures 1, 2 in Supplementary Data Sheet 1). In addition, three enzymes were annotated as β-specific hydrolases: EGIDFPOO_00274 is predicted to act as a β-1,4-mannooligosaccharide phosphorylase (GH130), while the two GHs EGIDFPOO_00532 and EGIDFPOO_01324 are putative β-glucosidases (GH1). However, no homologs of enzymes known to be involved in xylan degradation, such as endo-xylanases, β-xylosidases, α-arabinofuranosidases, α-glucuronidases or acetylxylan esterases, were identified.

Cells of Thermococcus sp. strain 2319x1E show a morphology typical for Thermococcales as they appear as irregularly shaped cocci with a size of 1–2 μm in diameter. This strain is an obligate anaerobe and grows optimal at 85°C and a pH of 7.0. When cultivated on xylan in batch mode, the generation time was 3.4 h (2.9 h for D-xylose, 3.7 h for Avicel^® cellulose, 3.2 h for D-glucose) and the final growth yield ranged 1–1.5 × 10⁷ cells ml^–1 for all substrates (see Supplementary Figure 3 in Supplementary Data Sheet 1). Notably, cell lysis was observed after 20 h of cultivation by DAPI staining, indicating a short stationary phase when the cells were grown in a closed bottle without stirring and gas exchange.

Thermococcus sp. strain 2319x1E crude extracts as well as the isolated cytosolic and membrane fractions obtained from cells grown on xylan possess hydrolytic activity on xylan and xylobiose, similar to strain 2319x1 (Figure 3A). The specific xylanolytic activity of the membrane fraction was roughly three times higher than for the cytosolic fraction, thus indicating the presence of membrane-bound hydrolase activity. Furthermore, the hydrolytic activity in crude extracts of cells grown on xylan, D-xylose and Avicel^® cellulose was compared using either xylan or CMC as substrate (Figure 3B). Whereas a higher xylanolytic activity is clearly linked to growth on xylan, only a slight effect of the growth substrate could be observed for the glucosidase activity on CMC compared to xylan. Of note, no activity was observed in the culture supernatant using the DNSA assay despite long incubation times. Additional efforts to concentrate the supernatant by means of centrifugation or filtration failed due to the viscosity of the residual xylan.

In order to first identify proteins from Thermococcus sp. strain 2319x1E that are regulated in response to the offered carbon source, we used a comparative proteomics approach. Thermococcus sp. strain 2319x1E cells were grown on either D-xylose, D-glucose, Avicel^® cellulose or beechwood xylan with subsequent MS-based full proteome analysis to identify differentially abundant glycoside hydrolases based on LFQ (Figure 4 and Supplementary Table 1).

Statistical evaluation of the identified proteins showed the strong alteration of protein abundances upon growth of Thermococcus sp. strain 2319x1E on beechwood xylan compared to D-glucose (Figure 4A), Avicel^® cellulose (Figure 4B), and D-xylose (Figure 4C). The abundance of 107 proteins was significantly upregulated with a log₂-fold change > 1 in the proteome of xylan-grown cells compared to cells grown on D-glucose (200 proteins compared to cellulose, 208 proteins compared to D-xylose) whereas 90 proteins were significantly downregulated with a log₂-fold change < -1 (114 proteins compared to cellulose and D-xylose). Among the proteins that showed a differential abundance in response to growth on the different sugars, we found several CAZymes (for overview see Figure 4D and Supplementary Table 2 in Supplementary Data Sheet 1). The α-specific CAZymes EGIDFPOO_00018 (GH13), EGIDFPOO_00375 (GH57), EGIDFPOO_00563 (GT35), EGIDFPOO_00674 (GH57) EGIDFPOO_01845 (GH57), and EGIDFPOO_01849 (GH13) were all found to be more abundant in cells grown on xylan compared to the other tested sugars, although this difference was not significant for EGIDFPOO_00674 compared to growth on D-xylose as well as for EGIDFPOO_01849 compared to growth on D-glucose. EGIDFPOO_00532, a predicted β-glucosidase, was less abundant in cells grown on xylan when compared to cells grown on D-glucose, but more abundant compared to cells grown on Avicel^® cellulose or D-xylose. Moreover, the GH122 family hydrolase EGIDFPOO_00753 was slightly upregulated in xylan-grown cells compared to D-glucose and Avicel^® cellulose, but slightly downregulated compared to D-xylose. Noteworthy, the genes for the two hydrolases EGIDFPOO_01845 (GH57) and EGIDFPOO_01849 (GH13) are in close genetic neighborhood to a predicted ATP-binding cassette (ABC) transporter comprising the genes EGIDFPOO_01846 and EGIDFPOO_01847 (both permeases), as well as to EGIDFPOO_01844 and EGIDFPOO_01848, which are predicted to encode for an ATP binding protein and a putative sugar binding protein, respectively. These proteins are homologous to subunits of an ABC transporter for maltose, trehalose and probably a number of other mono- and disaccharides (Greller et al., 1999). The abundance of this gene cluster is upregulated in cells grown on xylan compared to cells grown on Avicel^® cellulose, D-glucose or D-xylose. However, the putatively membrane bound sugar transport proteins EGIDFPOO_01846 and EGIDFPOO_01847 were removed from the data due to the filtering criteria applied (see methods; Supplementary Table 1). The log₂-fold changes for all candidate proteins are reported (Figure 4D).

To shift the focus on the detection of active glycosidases involved in xylan degradation in Thermococcus sp. strain 2319x1E, we screened the two well-established biotin-tagged cyclophellitol aziridine-based glycosidase probes JJB384 and JJB111 (Figure 5A) for their ability to label active enzymes in vivo upon growth on xylan. Thereto, Thermococcus sp. strain 2319x1E cells grown on xylan were incubated with 2 μM of the respective probe for 2 h, followed by the detection of labeled proteins via Western blot analysis (Figure 5B). This revealed a distinctive band pattern for the samples labeled with JJB384 with two prominent bands at a MW around 70 kDa. While the upper band is absent in the DMSO control, there is a band in the control sample at about the same height as the lower band in the labeled sample, but considerably less pronounced. In contrast, only faint bands are visible for the labeling with JJB111. Furthermore, the inhibitory effect of both ABPs on the xylanolytic activity was tested. Crude extract of Thermococcus sp. strain 2319x1E cells grown on xylan was incubated under the conditions previously described for the in vivo labeling, after which activity determination was performed via the DNSA assay. Despite only a faint labeling was obtained from the cells incubated with JJB111 in the Western blot analysis, the hydrolytic activity of the crude extract on xylan was decreased to 49% by preincubation with the probe. However, this effect was distinctly stronger for JJB384, with a remaining activity of about 23% (Figure 5C). Consequently, JJB384 was selected for further identification of target enzymes by MS-based proteomics.

To confirm that the band pattern obtained by labeling with JJB384 is specific for cells grown on xylan, a comparative in vivo ABPP of Thermococcus sp. strain 2319x1E grown on either D-xylose or beechwood xylan was conducted (Figure 6A). The Western blot analysis shows a differential labeling pattern of proteins from cells grown on the different carbon sources. In contrast to the labeling of cells grown on xylan, no differences in the band pattern between labeled cells and the control could be observed for cells grown on D-xylose. Thus, proteins that were not labeled with JJB384 after growth of cells on D-xylose show specific activity upon growth of cells on xylan and therefore may play a role in xylan degradation.

For identification of the JJB384 probe-labeled proteins, an affinity enrichment with subsequent tryptic on-bead digestion and LC-MS/MS analysis was performed using protein extracts from cells grown on D-xylose and beechwood xylan (Figures 6B,C and Supplementary Table 2). The initial data were filtered (see methods) to only keep protein groups that were identified in at least two out of three biological replicates of probe-labeled samples per carbon source. For the xylan samples, 122 protein groups were retained, of which 113 protein groups were enriched with the probe compared to the DMSO controls, whereas 68 out of 69 retained protein groups were enriched with JJB384 for the D-xylose samples. Among the xylan samples, 5 GHs (EGIDFPOO_00375, EGIDFPOO_00532, EGIDFPOO_00674, EGIDFPOO_01845, EGIDFPOO_01993) and 1 GT (EGIDFPOO_00563) were enriched with JJB384 (Figure 6C). Of these, only EGIDFPOO_01993 was not reported in the comparative full proteome analysis, as this protein did not pass the filtering criteria and was therefore removed from the initial dataset (see methods, Supplementary Table 1).

We next had a closer look at these active enzymes that likely participate in xylan degradation. EGIDFPOO_00532 is the only β-specific GH among the identified proteins and is closely related to β-glucanases and β-galactosidases from other thermophilic archaea (Supplementary Figure 4 in Supplementary Data Sheet 1 and Supplementary Table 3 in Supplementary Data Sheet 1). Notably, it is not only enriched in the xylan samples, but is also the only CAZyme enriched in the D-xylose samples (Figure 6B). The respective log₂-fold changes for all identified CAZymes are depicted in Table 2. Several α-specific CAZymes were thus only enriched in the xylan samples. The highest enrichment was observed for EGIDFPOO_00375 from the GH57 family (Supplementary Figure 2 in Supplementary Data Sheet 1) that is homologous to the characterized pullulan hydrolase TK-PUL from T. kodakarensis (Ahmad et al., 2014), 4-α-glucanotransferase from T. litoralis (Jeon et al., 1997) and α-amylase from P. furiosus (Laderman et al., 1993). These enzymes have been reported to hydrolyze multiple α-linked polysaccharides, including starch, glycogen, dextrin, amylose, amylopectin, and different cyclodextrins. According to phylogenetic analysis and BLAST with biochemically characterized enzymes, other enriched hydrolase candidates were predicted to function as amylopullulanase (EGIDFPOO_001845, Supplementary Figure 2 in Supplementary Data Sheet 1) or glucan-branching enzyme (EGIDFPOO_01993, Supplementary Figure 1 in Supplementary Data Sheet 1) based on PFAM analysis and sequence homology to biochemically described enzymes (Han et al., 2013; Sun et al., 2015) and should therefore also be specific for hydrolysis of α-linkages in polymers. EGIDFPOO_00563 is highly similar to the GT35 family maltodextrin phosphorylase from T. litoralis (97% sequence identity) (Xavier et al., 1999). Notably xylanase activity was reported for the characterized homolog from T. zilligii (76% sequence identity) (Uhl and Daniel, 1999), which was, however, later questioned and the respective protein is now annotated as a maltodextrin-phosphorylase in the PDB database (Rolland et al., 2002). EGIDFPOO_00674 is another putative α-specific glycosidase enriched in the xylan samples that belongs to the highly diverse GH57 family and, like EGIDFPOO_00375 and EGIDFPOO_01845, forms an isolated cluster within the phylogenetic tree (Supplementary Figure 2 in Supplementary Data Sheet 1). Unlike the other GH57 family proteins found with the ABPP approach, EGIDFPOO_00674 was not assigned a specific function by automated proteome annotation in Swiss-Prot (Gattiker et al., 2003) and was therefore selected for biochemical characterization along with the putative β-glucosidase EGIDFPOO_00532. The second predicted β-glucosidase EGIDFPOO_01324 (Table 1) was neither identified with the comparative full proteome survey nor with the ABPP approach.

In summary, of the 13 GHs identified in the genome of Thermococcus sp. strain 2319x1E (Table 1), 7 GHs were shown to be upregulated in response to xylan compared to any of the other sugars, and of these, 5 GHs were additionally shown to be active upon growth on xylan by labeling with JJB384 (Supplementary Table 2 in Supplementary Data Sheet 1). In addition, EGIDFPOO_01993 (predicted glucan-branching enzyme) was only confirmed via ABPP.