Pseudomonas-associated bacteria play a key role in obtaining nutrition from bamboo for the giant panda (Ailuropoda melanoleuca)

ABSTRACT Gut microbiota plays a vital role in obtaining nutrition from bamboo for giant pandas. However, low cellulase activity has been observed in the panda’s gut. Besides, no specific pathway has been implicated in lignin digestion by gut microbiota of pandas. Therefore, the mechanism by which they obtain nutrients is still controversial. It is necessary to elucidate the precise pathways employed by gut microbiota of pandas to degrade lignin. Here, the metabolic pathways for lignin degradation in pandas were explored by comparing 209 metagenomic sequencing data from wild species with different feeding habits. Lignin degradation central pathways, including beta-ketoadipate and homogentisate pathway, were enriched in the gut of wild bamboo-eating pandas. The gut microbiome of wild bamboo-eating specialists was enriched with genes from pathways implicated in degrading ferulate and p-coumarate into acetyl-CoA and succinyl-CoA, which can potentially provide the raw materials for metabolism in pandas. Specifically, Pseudomonas, as the most dominant gut bacteria genus, was found to be the main bacteria to provide genes involved in lignin or lignin derivative degradation. Herein, three Pseudomonas-associated strains isolated from the feces of wild pandas showed the laccase, lignin peroxidase, and manganese peroxidase activity and extracellular lignin degradation ability in vitro. A potential mechanism for pandas to obtain nutrition from bamboo was proposed based on the results. This study provides novel insights into the adaptive evolution of pandas from the perspective of lignin metabolism. IMPORTANCE Although giant pandas only feed on bamboo, the mechanism of lignin digestion in pandas is unclear. Here, the metabolic pathways for lignin degradation in wild pandas were explored by comparing gut metagenomic from species with different feeding habits. Results showed that lignin degradation central pathways, including beta-ketoadipate and homogentisate pathway, were enriched in the gut of wild bamboo-eating pandas. Genes from pathways involved in degrading ferulate and p-coumarate via beta-ketoadipate pathway were also enriched in bamboo-eating pandas. The final products of the above process, such as acetyl-CoA, can potentially provide the raw materials for metabolism in pandas. Specifically, Pseudomonas, as the most dominant gut bacteria genus, mainly provides genes involved in lignin degradation. Herein, Pseudomonas-associated strains isolated from the feces of pandas could degrade extracellular lignin. These findings suggest that gut microbiome of pandas is crucial in obtaining nutrition from lignin via Pseudomonas, as the main lignin-degrading bacteria.

P hylogenetic studies have shown that the giant panda (Ailuropoda melanoleuca) is an herbivorous mammal belonging to the bear family (1,2).Giant pandas have undergone a series of evolution to adapt to the special bamboo diet, including ecological (3)(4)(5), morphological (6)(7)(8), and genetic (9)(10)(11) adaptations (4,12).Notably, giant pandas lack genes for lignin, cellulose, and hemicellulose degrading pathways in their genome (9).As a result, conservation biologists have been curious to know how giant pandas obtain nutrition from bamboo.A complete metabolic pathway for cellulose and hemicellulose degradation has been detected in the intestine of giant pandas (13).Therefore, the gut microbiota is considered the main route through which giant pandas obtain nutrition from bamboo diet (9).However, the gut microbiota of giant pandas has not shown an obvious convergent evolution phenomenon according to their feeding habits, and they still retain a gut microbiota structure like that of carnivores (14)(15)(16).Although the abundance of hemicellulose-degrading bacteria in the gut of giant pandas increases after eating bamboo (17), the abundance of genes related to cellulose and hemicellulose degradation in their gut microbiome is significantly lower than that in other herbivores (13,14,17).Many previous studies have confirmed that the gut microbiota of giant pandas is distinct from that of red pandas and clusters closer to those of the black bears and carnivores (14)(15)(16).However, Huang et al. found a close similarity in gut microbiota structure among bamboo-eating pandas (giant and red pandas), confirming that food factors drove the convergent evolution of gut microbiota of pandas (18).Therefore, it remains controversial whether the gut microbiota of giant pandas can adapt to their highly specialized diet (14)(15)(16).Indeed, studies on the adaptation of gut microbiota of giant pandas to the bamboo diet have mainly focused on captive populations.Considering the obvious difference in gut microbiota composition between captive and wild giant pandas (13,19,20), more studies should be performed to uncover mechanisms behind the adaptation of gut microbiota of wild giant pandas to the bamboo diet.
Bamboo is mainly composed of lignin, cellulose, and hemicellulose (21).Lignin attaches to cellulose and hemicellulose in the cell wall and protects them from degradation.Therefore, the gut microbiota of giant pandas needs to first oxidize and decompose lignin before digesting the other components of the bamboo diet.To date, the precise mechanism of lignin degradation by giant pandas is unclear.Perenniporia medulla-panis is a fungus found in the gut of giant pandas and exhibits lignin peroxidase activity (22).Lac51 is a gene-encoding laccase (Lac), an enzyme that can degrade a variety of lignin monomer phenols via oxidation.The sequence similarity alignment of Lac51 cloned from the fecal microbiotas of giant pandas is closely related to multicop per oxidase gene of Pseudomonas sp.(23).However, only a few studies have assessed the potential role of gut microbiota of giant pandas in lignin degradation (17).The community structure of microbiota in wild giant pandas is significantly different from that of captive giant pandas (19).Specifically, Streptococcus and Enterobateriaceae are the most abundant genera in captive pandas (16,19,24), while Pseudomonas is the most dominant genus in the gut of wild pandas (19,20,25).Similarly, Pseudomonas is the most dominant bacteria in the gut of the wild population of red pandas that feed on bamboo (26).Most Streptococcus genus and Enterobateriaceae family have been reported as primarily pathogens or opportunistic pathogens (27)(28)(29).Pseudomonas is one of the few bacteria that can efficiently degrade cellulose and lignin (30,31).For instance, Pseudomonas sp.strain ys-1p, Pseudomonas fluorescens, and Pseudomonas PKE117 can degrade several lignin monomers and dimers (32)(33)(34)(35).Davinia et al. revealed that Pseudomonas putida KT2440 secretes vesicles carrying enzymes that can degrade lignin and its derivatives (36).Whether the Pseudomonas-associated bacteria in wild giant pandas can potentially degrade lignin is still unclear.
A recent study on the adaptation of gut microbiome of giant pandas to different diets and habitats before and after reintroduction identified four genes encoding enzymes, including catalase peroxidase, dioxygenase, quinone reductase, and triacylglycerol lipase, in the metagenome of fecal microbiotas of the giant pandas (37).Pseudomonas, Enterococcus, and Lactococcus are the bacteria enriching these genes.The abundance of the three bacterial genera significantly increased after the release of giant pandas into the wild (37).Yao et al. (20) and Tang et al. (25) also reported that Pseudomonas abundance in the gut of captive giant pandas significantly increased after their release into the wild, suggesting that the diversity of gut microbiota of giant pandas increased after reintroduction in the wild.Notably, Pseudomonas is the main cyanide-degrading bacteria in the gut of bamboo-eating pandas (wild giant and red pandas), revealing their adaptation to the bamboo diet (38).Collectively, these studies suggest that Pseudomo nas is the main bacteria responsible for the adaptation of giant pandas to the bamboo diet.However, it is still not clear whether the strain of Pseudomonas in the gut of giant pandas can degrade lignin.Also, the specific metabolic pathway of degrading lignin has not been elucidated.In this study, the potential role of the gut microbiome of giant pandas in lignin degradation was explored via metagenomic sequencing to reveal the metabolic function of gut microbiota involved in lignin degradation in bamboo-eating pandas (giant and red pandas), herbivorous, omnivorous, and carnivorous.High-quality individual draft genomes (bins) were produced based on the shotgun metagenomic assemblies.The genomes were then used to reveal the metabolic pathways related to lignin degradation.In addition, Pseudomonas-associated strain were isolated from the feces of giant pandas.Lignin degradation and extracellular secretion ability of lignino lytic enzymes were also verified in vitro.
This study reveals the specific metabolic pathway of lignin degradation in giant pandas and provides new evidence for adaptive evolution of giant pandas to the bamboo diet.Therefore, these findings provide a basis for future studies on giant pandas, which is important for conservation.

Sample collection and individual information
Fecal samples were collected from both wild and captive animals.Fresh fecal samples were collected from wild giant pandas (Ailuropoda melanoleuca) (n = 7) and red pandas (Ailurus fulgens) (n = 5) in Fengtongzhai Nature Reserve (Ya'an, Sichuan Province, China).In addition, fecal samples from captive giant pandas (n = 7) and red panda (n = 5) were collected from the China Conservation and Research Center and Bifengxia Ecological Zoo (Ya'an, Sichuan Province, China), respectively.The feces were aseptically collected by ranger staff as part of their daily monitoring.We assessed the freshness of the feces by considering both color and surface sheen.After the contaminated part of the surface is removed, the samples were immediately frozen in liquid nitrogen, then transferred to −80°C refrigerator for further utilization.
Previous studies showed that the seven fecal samples from wild giant pandas were from seven different animals (19,39).The five wild red panda samples have also been confirmed to be from different individuals via GPS collars.The detailed sample informa tion, including age, gender, and population, is shown in Table S1.

DNA extraction and sequencing
Total microbial genomic DNA was extracted from the fecal samples using the PowerFecal DNA isolation kit (QIAGEN, Inc., Valencia, CA, USA), according to the manufacturer's instructions.The DNA concentration and purity were measured using the Qubit (Thermo Fisher Scientific, Waltham, MA, USA).Agarose gel electrophoresis was performed to assess the DNA quality.The DNA samples that met the criteria of metagenomic sequencing were used for library preparation: (i) DNA concentration of >15 ng/µL; (ii) total amount of DNA >6 µg; (iii) non-contaminated and intact DNA fragment.
Shotgun metagenomic DNA libraries were constructed based on the Illumina TruSeq DNA Sample Prep V2 Guide (Illumina, Inc.; San Diego, CA, USA), with shearing to 300to 400-bp fragments.Shotgun metagenomic sequencing was performed on Illumina platform using paired-end 2 × 150 bp chemistry (Novogene, Beijing, China).Ultimately, we obtained gut metagenome data via shotgun metagenomic sequencing from seven wild giant pandas, five wild red pandas, seven captive giant pandas, and five captive red pandas.

Downloading of gut metagenomic data
The composition of gut microbiotas is significantly different between captive and wild animals.In this study, only data sets from wild mammals were included for joint analysis, except for bamboo-eating animals.Meanwhile, only the metagenome data generated by Illumina platform were downloaded to avoid the possible bias caused by different sequencing platforms.Gut metagenomics data of six red pandas and eight giant pandas from the study of Zhu et al. (38) were retrieved from the National Centre for Biotechnology Information's Sequence Read Archive (38).The previously published mammalian metagenome data set representing herbivores (n = 97), omnivores (n = 10), and carnivores (n = 14) were downloaded from Levin et al. 's study (40).In addition, the published data of tiger (n = 6) (41), black rhinoceros (n = 17) (42), David's deer (n = 30) (38), Gazella subgutturosa (n = 4) (43), Chinese pangolin (n = 5), and Malayan pangolin (n = 5) were also downloaded from the Genome Sequence Archive (44).The published raw metagenome sequences used in this study were showed in Table S2.

Shotgun metagenomic sequence analysis
Adapter sequences were removed from all the sequence data using Cutadapt v1.9.1 software (45).Raw paired-end reads were processed by Trimmomatic software to filter out low-quality sequences using a sliding window (5-bp bases) (46).The criteria for quality control were as follows: (i) a sequence was removed once its average quality within the window fell below Q20; (ii) sequences containing any N-bases were filtered out; (iii) reads that were below 50 bp in length were dropped; (iv) only paired-end reads were retained.Animal feces often contain cells shed from their intestines.Therefore, total bacterial DNA extracted from animal feces usually contains the DNA of the host, indicating that the metagenomic sequencing data may contain part of the host genome sequences.After quality control, Bowtie2 (47) was employed to blast the data set of each species to their genome sequences.Sequences with >90% similarity to the genome sequences of the host were removed.
High-quality reads were assembled into contigs using metaSPAdes with the default parameters (48).Assembled contigs with more than 500 bp in length were retained for subsequent analyses.Open reading frames (ORFs) of gut microbiome of each species were predicted from the assembled contigs using MetaGeneMark ORFs ark v 2.8 (49).The non-redundant gut microbial gene set of each species was generated with a 90% identity cutoff using CD-HIT v4.8.1 (50).Data were randomly sub-sampled to the minimum number of sequences in all samples using seqtk-master(https://github.com/lh3/seqtk)before aligning the sequences against the non-redundant panda gut microbiome gene set to reduce biases caused by different sequence depths.Bowtie2 (47) was used to align the paired-end clean reads of each sample against their non-redundant gut microbial gene set.Gene abundance A(g) in each sample was determined according to Qin et al. (51) as follows: Non-redundant gut microbial gene set of each species was aligned against the Kyoto Encyclopedia of Genes and Genomes (KEGG) online database to annotate the gene functions (52).GHOSTX searches were selected as the assignment method, and other parameters were set as default (53).
Kraken2 was used for taxonomic classification on the clean reads to explore the relative abundance of Pseudomonas-associated bacteria in the gut of bamboo-eating pandas and other mammals.

Extraction of individual draft genomes (bins) from metagenomic data of wild giant pandas
High-quality individual draft genomes (bins) of gut bacteria of wild giant pandas were obtained from shotgun metagenomic assemblies using MetaWRAP (v1.2) to explore the gut strain of giant pandas with the potential to degrade lignin (54).Metagenomic assemblies were binned into draft genomes using the binning module in three metagenomic binning software (MaxBin2, metaBAT2, and Concoct).All bins recovered from the assembly were combined to obtain a single complete bin set using Bin_refinement module.Furthermore, Reassemble_bins module was used to improve N50, completion, and reduce contamination of the bins.Reassembled bins with completeness >70% and contamination <5% were considered as high-quality individ ual draft genomes (bins) for subsequent analysis.Estimation of bin abundances across samples was performed using the metaWRAP-Quant_bins module.Bins were uploaded to RAST (Rapid Annotations using Subsystems Technology) web server for the identifica tion of species and potential functions of the genome.All parameter options were set at default, and the closest neighbors of bins (with the highest "Identify Score") were considered as the possible species.A phylogenetic tree was constructed based on all bins using PhyloPhlAn 3.0 tool (55).Furthermore, 40 complete Pseudomonas genomes representing the four closest neighbors of Pseudomonas-associated strains recovered by binning were downloaded for phylogenetic analysis.
PCR products were detected using 2.0% agarose gel electrophoresis, and fragments (~150 bp) were obtained for subsequent analysis.Bands with the expected fragment size were purified using Axygen Gel Extraction Kit (Axygen, Silicon Valley, USA) accord ing to the manufacturer's instructions.The purified target fragment was ligated into the pMD19-T vector system (TaKaRa, Dalian, China) and transformed in Escherichia coli DH5α cell according to the manufacturer's instructions.A total of 50-85 positive clones per fecal sample were selected.The size of the inserted bacterial laccase-like gene fragment was evaluated via PCR, and sequencing was performed by Beijing Genomics Institute (BGI, Shenzhen, China) using primer RV-M.Nucleotide sequences were manually proofread, and similar sequences were obtained through BLAST search in GenBank.Nucleotide sequences were translated into amino acid sequences and were aligned using ClustalW program (57).Sequence analysis and phylogenetic reconstruction were performed using Mega 11 software (58).The neighbor-joining tree was reconstructed based on the maximum composite likelihood method.The strains encoding laccase-like multicopper oxidase enzyme were identified as different species.

Isolation and identification of Pseudomonas-associated bacteria
Pseudomonas-associated bacteria were screened in fresh fecal samples of wild giant pandas.Briefly, 1-g fecal sample was dissolved in 100-mL sterile saline solution to form a fecal solution.The fecal solution was diluted 10 −7 times using sterile saline solution for single colony plotting on an agar plate.The diluted fecal solution (200 µL) was inocula ted into Pseudomonas-associated bacteria solid screening medium containing 16.0 g/L gelatin peptone, 10.0 g/L acid hydrolyzed casein, 10.0 g/L K₂SO₄, 1.4 g/L Mg 2 Cl, 0.2 g/L cetyltrimethylammonium bromide, 10 mL/L glycerol, 15 mg/L nalidixic acid, and 12.0 g/L agar.Single colonies were obtained after incubating the sample at 37°C for 24 h.The single colonies were re-streaked thrice to obtain a pure colony for strain identification.
Single colonies were incubated with Luria broth (LB) liquid medium at 37°C, 200 rpm for 24 h.The LB liquid medium was then centrifuged at 12,000 × g for 5 min, and the supernatant was discarded.DNA isolation kit (Tiangen Biotech, Beijing, China) was used to extract genomic DNA of the bacterial cells following the manufacturer's instructions.The 16S rDNA sequence was amplified with bacterial universal primers 27F (AGAGTTT GATCCTGGCTCAG) and 1492R (GGTTACCTTGTTACGACTT).The amplified fragment was sequenced by Tsingke Biotechnology Co., Ltd.The obtained sequences were compared with the NCBI sequences using the BLAST website.

Determination of lignin-degrading capability of Pseudomonas-associated bacteria
The isolated Pseudomonas-associated strains were inoculated into modified M9 solid medium containing 6.78 g/L Na2HPO4, 3 g/L KH2PO4, 0.5 g/L NaCl, 1 g/L NH4CL, 2 mM MgSO4, 100 µM CaCl2, 100 µM MnSO4, 100 µM CuSO4, 100 µM FeSO4, 50 µM ZnSO4, 2 g/L kraft lignin, and 12.0 g/L agar (with lignin as the sole carbon source; pH 7.0).The cell growth and lignin degradation were determined to test the ligninolytic activity of the Pseudomonas-associated strains.Pseudomonas-associated strains were inoculated in LB medium and cultured at 37°C and 200 rpm for 24 h while shaking.The culture broth was centrifuged at 12,000 × g for 5 min to obtain the bacterial cells.The bacterial cells were inoculated into 100 mL of modified M9 medium (pH 7.0) containing 2 g/L kraft lignin (initial OD600 = 0.1), then cultured at 37°C, 200 rpm for 7 days.Fermentation broth (1 mL) was taken daily to measure the growth curve and lignin-degrading capability of the Pseudomonas-associated strains.Cell growth was determined at OD600; then, the growth curve of Pseudomonas-associated strains was plotted.The lignin degradation rate was evaluated by monitoring the decrease in A280.Degradation ratio was calculated as follows: degradation% = (A control − A sample ) / A control * 100%.All the experiments were performed in triplicates.

The decolorization capability for aromatic dyes and enzyme assays
The capability of the screened isolates to decolorize the aromatic dyes with almost similar structures to lignin fragments was determined to further evaluate whether Pseudomonas-associated strains can secrete extracellular ligninolytic enzymes to degrade lignin.The Pseudomonas-associated strains were inoculated into aniline blue solid medium (pH 7.0) containing 10 g/L glucose, 10 g/L peptone, 0.04 g/L water-solu ble aniline blue, and 12 g/L agar.The decolorization circle of the Pseudomonas-asso ciated strains on aniline blue solid medium was confirmed; then, the bacterial cells with OD600 = 0.1 were inoculated into 50-mL aniline blue liquid medium (pH 7.0).The samples were cultured at 37°C and 200 rpm for 96 h while shaking.The aniline blue liquid medium without bacteria was used as the blank control.The culture medium was centrifuged at 12,000 rpm for 5 min every 24 h, and the absorbance of supernatants was measured at 600 nm.The decolorization ratio was calculated as follows: decolorization% = (A control − A sample ) / A control * 100%.All the experiments were performed in triplicates.

Determination of degradation products of lignin by Pseudomonas-associated strains
The centrifugal supernatants of lignin fermentation medium of Pseudomonas-associ ated strains were filtered using a 0.2-µm filter membrane, and quality control sam ples were prepared.Metabolites of the supernatants were extracted before liquid chromatography-mass spectrometry (LC-MS) detection, as previously described (59).Each sample was run in triplicate, and the stability of mass spectrometry was eval uated.LC-MS detection was performed as previously described (60,61).The raw data were first converted to mzXML format via MSConvert in ProteoWizard software package (v3.0.8789) (62) and processed using XCMS (63) for feature detection, retention time correction, and alignment.The metabolites were identified via accuracy mass (<30ppm) and LC-MS data, which were matched with Human Metabolome Data base (http://www.hmdb.ca/)(64), MassBank (http://www.massbank.jp/(65), LipidMaps (https://www.lipidmaps.org)(66), mzcloud (https://www.mzcloud.org)(67), and KEGG (https://www.kegg.jp)(68).The robust LOESS signal correction (QC-RLSC) was applied for data normalization to correct any systematic bias (69).Ion peaks with relative standard deviations less than 30% in QC were kept ensuring proper metabolite identification.Relative quantification was estimated based on the ratio of the respective peak areas to the total peak areas.

Statistical analysis
Significant differences in the relative abundance of genes between different diet groups were determined using the Kruskal-Wallis test followed by Dunn's multiple-comparison post-hoc test.All statistical analyses were performed using GraphPad Prism 7 software (GraphPad Software, Inc., USA).Figures were generated using "pheatmap" packages (70) and different functions ("boxplot, " "barplot, " "pie, " and "plot") in R base (version 3.6.1)(71).Multivariate statistical analysis and modeling were conducted using the resultant data sets from LC-MS procedures via the Ropls software (72).Models were built based on principal component analysis (PCA) to explore data set variations.The P-value and variable importance projection (VIP) produced by OPLS-DA were applied to discover the significance of metabolites (P value <0.05 and VIP values >1) in different groups.

Metagenome sequencing data
A total of 319,158,629 paired-end metagenome reads were generated in this study.In addition, 5,347,754,253 published raw paired-end reads from 209 wild individual samples (representing bamboo-eating pandas, herbivores, omnivores, carnivores) were retrieved from the public database for comprehensive analysis.A total of 5,117,016,424 clean paired-end reads were retained for subsequent analysis after filtering out the low-quality reads (Q < 20) and host sequences.The final contigs of each species were obtained through de novo assembly.Non-redundant gut microbiome gene set of each species was created from contigs.

Composition gut microbiotas of bamboo-eating pandas
Principle coordinate analysis of predicted metagenomic function (KO genes) based on Bray-Curtis distance showed that wild bamboo-eating pandas formed distinct clusters from other animals (Adonis R 2 = 0.473; P-value = 0.001), implying that they have a distinct function (Fig. S1).Moreover, wild bamboo-eating pandas harbored a distinct gut microbiome from captive bamboo-eating pandas.Furthermore, gut microbiome in carnivores was distinct from that of herbivores and omnivores.Overall, the functions of the gut microbiome of different animals clustered according to their food habits, indicating their various roles.Analysis at the genus level showed that Pseudomonas and Enterobacteriaceae were the most abundant genus in the gut of wild giant pandas (Fig. S2A) and captive giant pandas (Fig. S2B), respectively.
Taxonomic assignment of enzymes revealed that a large proportion (26.08% on average) of the predicted genes were present in species within the Pseudomonas genus bacteria.For example, 8 out of 12 vdh, 16 out of 54 vanA, and 38 out of 163 pcaD genes in this study were homologous to genes from Pseudomonas-associated bacteria (Table S3).This finding indicated that predicted genes coding for lignin derivatives or monomer-digesting enzymes were mainly contributed by Pseudomonas genus.
Heatmap analysis showed that the average abundance of genes involved in the metabolism of central aromatic intermediates, p-coumarate, and ferulate was highest in wild bamboo-eating pandas compared with the abundance in other animals (carni vores, omnivores, herbivores, and captive bamboo-eating pandas) (Fig. 1D).The relative abundance of these genes was significantly higher in wild bamboo-eating pandas than in other animals (Fig. 2).

Phylogenetic analysis of Pseudomonas genomes in wild giant pandas
The four closest neighbors of bin 1, bin 17, bin 33, and bin 52 were Pseudomonas fluorescens PfO-1, Pseudomonas syringae pv.phaseolicola 1448A, Pseudomonas putida KT2440, and Pseudomonas aeruginosa PAO1, respectively, as identified through RAST.A total of 40 complete genomes of different strains (10 per species), belonging to Pseudo monas fluorescens, Pseudomonas syringae, Pseudomonas putida, and Pseudomonas aeruginosa, were randomly selected to construct phylogenetic trees for further classifica tion of Pseudomonas strains in the wild giant pandas.Phylogenetic analysis showed that bin 52 (Pseudomonas fluorescens) clustered in the Pseudomonas fluorescens clade and that bin 1, bin 17, and bin 33 (Pseudomonas) clustered into a single clade in the phylogenetic tree (Fig. 4).Similar to Pseudomonas genomes in this study (Fig. 3C), the genome of Pseudomonas fluorescens PfO-1 (Fig. S3A), Pseudomonas syringae pv.phaseolicola 1448A (Fig. S3B), Pseudomonas putida KT2440 (Fig. S3C), and Pseudomonas aeruginosa PAO1 (Fig. S3D) encoded all enzymes implicated in the metabolism of central aromatic intermediates.

Laccase-like multicopper oxidase genes were present in the gut of wild giant pandas
Approximately 150-bp fragments between regions I and II of laccase-like multicopper oxidase genes, which are crucial in lignin degradation, were obtained from the DNA derived from feces of wild giant pandas through PCR amplification.A total of 228 colonies from three samples (76 per sample) of wild giant pandas were chosen at random for sequencing.Nucleotide sequences obtained were used as query sequences for BLAST search to obtain the sequences deposited in GenBank nucleotide database.The BLAST results indicated that most (except for colony 2) PCR products highly corre sponded to the laccase-like multicopper oxidase genes of bacteria.The results showed that 76% of products of colonies had a high similarity (>90%) to gene sequences retrieved from GenBank nucleotide database.Notably, only two clones had less than 80%   similarity compared with nucleotide sequences deposited in the GenBank (Table S6).The most consistent alignment results obtained from GenBank sequences were considered as the possible species of laccase fragments.A total of 228 colony products were identified as Verrucomicrobiaceae HC12, Verrucomicrobiaceae ONA9, Pseudomonas antarctica, Janthinobacterium svalbardensis, Pseudomonas gingeri, Agrobacterium, Flavobacterium sp., Acinetobacter sp., Pseudomonas azotoformans, Pseudomonas fluorescens, Pseudomo nas putida, Klebsiella pneumonia, Stenotrophomonas sp., Caulobacter sp., Brevundimonas diminuta, Brevundimonas vancanneytii, Brevundimonas naejangsanensis, and Sphingosini cella sp.(Table S6).
Eighteen published laccase-like multicopper oxidase gene sequences were retrieved for phylogenetic analysis.Phylogenetic results indicated that the laccase-like multicop per oxidase genes in the feces of wild giant pandas clustered into 19 clades.Notably, each representative clone fragment showed a close evolutionary relationship with the reference sequences obtained from GenBank (Fig. 5A).Pseudomonas-associated sequences were abundant in feces libraries of the wild giant pandas, where they accounted for 83.33% of the total number of sequences (Fig. 5B).The proportion of Pseudomonas fluorescens-associated sequences was the highest (about 46%), followed by the abundance of Pseudomonas putidaand Pseudomonas azotoformans-associated sequences, accounting for 19% and 15%, respectively (Fig. 5B; Table S6).In addition, predicted genes coding for multicopper oxidase enzymes were identified in the genome of Pseudomonas-associated bin (Table S5).These results imply that laccase-like

Lignin degradation capability of single Pseudomonas-associated bacteria
Three most abundant Pseudomonas-associated bacteria were obtained based on colonial morphology difference in the Pseudomonas-associated bacteria solid screening medium after three repeated cultivations.These Pseudomonas-associated isolates were closely matched to Pseudomonas putida strain cqsH1 (99%), Pseudomonas sp.strain QW16-14 (99%), and Pseudomonas oryzihabitans strain h-2 (99%) based on 16S rRNA gene sequence homology.
Three Pseudomonas-associated isolates grew well on a solid medium with lignin as the sole carbon source (Fig. 6A).The growth and lignin-degrading curves of Pseudomo nas putida, Pseudomonas sp., and Pseudomonas oryzihabitans are shown in Fig. 6B.The lag phase, exponential growth phase, stationary phase, and decline phase occurred at 0-12 h, 24-72 h, 84-120 h, and 132-168 h, respectively.The initial average absorbance of the lignin culture medium was 4.47 at 280 nm, which decreased to 3.21 after 7 days of incubation.The degradation rates for Pseudomonas putida, Pseudomonas sp., and Pseudomonas oryzihabitans after 7 days were 19.66%, 19.14%, and 18.17%, respectively (Fig. 6C).

Extracellular ligninolytic enzymes of Pseudomonas-associated isolates
The three Pseudomonas-associated isolates secreted extracellular ligninolytic enzymes and produced a transparent ring after 24 h of incubation on aniline blue solid medium (Fig. 7A).However, most aniline blue was decolorized by bacteria after 48 h (Fig. 7A).The aniline blue degrading curve of Pseudomonas putida, Pseudomonas sp., and Pseudomo nas oryzihabitans are shown in Fig. 7B.The degradation rates of Pseudomonas putida, Pseudomonas sp., and Pseudomonas oryzihabitans after 96 h were 90.74%, 88.97%, and 86.77%, respectively.The enzyme production curve of three Pseudomonas-associated isolates is shown in Fig. 8C through E. The results showed that the three Pseudomonasassociated isolates secreted considerable reactive Lac, LiP, and MnP.Furthermore, the three Pseudomonas-associated isolates showed a similar characteristic of secreting extracellular ligninolytic enzymes in the culture medium with bamboo powder as the sole carbon source  7C), while the average MnP activity was 300.08, 269.84, and 258.13 U/mL, respectively (Fig. 7E).The average LiP activity of Pseudomonas putida, Pseudomonas sp., and Pseudomonas oryzihabitans was 2130.5, 2083.12, and 2069.75U/mL, respectively (Fig. 7D).

Metabolites of degraded lignin by Pseudomonas-associated isolates
Compared with the control (0 day), LC-MS analysis detected new peaks in the culture medium after treatment with Pseudomonas-associated strain for 3 and 7 days (Fig. S4A).The PCA score plot displayed three clusters that corresponded to different cultivation time with Pseudomonas-associated strain.Metabolites in the culture medium after treatment with Pseudomonas-associated strain were significantly different compared with the control (0 day) (Fig. 8A).Distinctive clustering in the metabolite composition of the supernatant was detected between the control group (0 days) and the culture group (Fig. S4B).A total of 229 and 245 differential metabolites were detected between 0 day vs 3 days and 0 day vs 7 days, respectively, of which 201 were shared.Compared with the control (0 day), cluster analysis indicated that the 201 metabolites in the culture medium were significantly different after treatment with Pseudomonas-associated strains for 3 and 7 days (Fig. 8B).LC-MS analysis detected 14 of 25 substances for the products in the beta-ketoadipate and homogentisate pathway (Fig. 1).Among them, cis,cis-muconate, hydroxybenzaldehyde, vanillate, phenylpyruvate, 2-hydroxyphenylacetate, homogenti sate, fumarylacetoacetate, and fumaric acid were only identified in the supernatant of culture medium after treatment with Pseudomonas putida, Pseudomonas sp., and Pseudomonas oryzihabitans (separately) for 3 and 7 days (Table 2).Lignin derivatives, such as catechol, ferulate, 4-coumarate, and protocatechuate, were detected in the control group (Table 2).Nevertheless, the relative quantification of the above derivatives was higher on the third and seventh days (Fig. 8C).

DISCUSSION
Numerous studies have identified genes encoding cellulase and hemicellulose in the gut microbiota genome of giant pandas (13).However, the absence of in vitro experiments to validate these findings has been notable.To date, only a limited number of studies have elucidated the degradation of lignin in giant pandas, particularly the specific metabolic pathway involved in giant panda lignin degradation.The present study is the first to reveal the complete metabolic pathway involved in the degradation of lignin derivatives by the gut microbiome of giant pandas.The β-ketoadipate pathway, comprising the catechol and protocatechuate branches, is the major microbial degradation pathway for lignin-derived aromatic compounds (73,74).Gut microbiota, such as Rhodococcus jostii RHA1 (75), Brevibacillus thermoruber (76), Bacillus ligniniphilus (77), and Pseudomonas putida KT2440 (78), degrade lignin mainly through β-KAP.In addition, the homogentisate pathway plays a key role in the degradation of lignin-derived aromatic compounds (79).Degradation pathways for 4-hydroxyphenylacetate (79), phenylalanine, and tyrosine eventually join the homogentisate pathway (80).Notably, genes involved in the β-KAP and homogentisate pathways were enriched in the gut microbiotas of wild giant pandas.The relative abundance of these genes was significantly higher in wild giant pandas than in captive giant pandas, herbivores, carnivores, and omnivores.However, the abundance of these genes in wild giant pandas was similar to those in another bamboo-eating panda (red panda).This indicates that the gut microbiome of wild giant pandas plays a key role in catalyzing cleavage of the benzene ring through the metabolic pathway involved in breakdown of central aromatic intermediates.Of note, the gut microbiotas of wild giant pandas showed significantly enriched genes encoding several enzymes that catalyze the degradation of ferulic and p-coumarate to protocatechuate, which subsequently join the Pbβ-KAP pathway.Ferulic and p-coumarate are lignin monomers (81,82) produced during the initial degradation reaction of lignin catalyzed by laccase (83).Laccase-like multicopper oxidase gene, which plays a similar role in degrading lignin (84), was also abundant in the feces of wild giant pandas and was mainly contributed by Pseudomonas-associated bacteria.Therefore, most genes implicated in lignin degrada tion were identified in the genome of Pseudomonas-associated bacteria of wild giant pandas.Notably, the final product of lignin degradation (fumarate, acetyl-CoA, and succinyl-CoA) can enter the TCA cycle to produce nutritionally important intermediates.The lignin in bamboo is mainly of the HGS type [p-hydroxyphenyl (H), vanillin (G), syringaldehyde (S)], containing a considerable amount of p-coumarate and ferulic (85).The β-ketoadipate pathway is the major lignin-degrading pathway in bacteria isolated from erosive bamboo slips (86,83).This explains why the gut microbiotas of wild giant pandas had significantly high expression levels of genes encoding enzymes implicated in p-coumarate and ferulic degradation through the β-ketoadipate pathway.Previous studies have shown that the genome of gut microbiotas of captive giant pandas lacks genes encoding enzymes involved in lignin degradation (17).The reason for these findings is that the gut microbiome composition of captive giant pandas is significantly different from that of wild giant pandas (19).For instance, the dominant gut bacteria in wild giant pandas is Pseudomonas, whereas the dominant gut bacteria of captive giant pandas are Enterobacteriaceae and Streptococcus (19,20).The wild giant pandas are exclusive bamboo specialists with almost 99% of its diet being bamboo (4, 12); however, except for bamboo, captive giant pandas are fed steamed grain mixture, fruits, and animal products (87).In addition, giant pandas can eat a wider variety of bamboo in the wild.These factors may be the possible reasons for the difference in the composition and metabolism of the microflora of giant pandas in the wild and in captivity.Of course, more research is needed in the future to explain why no lignin metabolity-related pathways have been found in captive pandas.Moreover, red pandas that exclusively eat bamboo exhibit a similar phenomenon as wild and captive population (26).The results of the present study suggest that the study of the adaptive evolution of animal intestinal flora to diet should be conducted using wild populations rather than captive populations.The genome of Pseudomonas-associated strains (bin 1, bin 17, bin 33, and bin 53) in the feces of giant pandas can encode all enzymes involved in the metabolism of central aromatic intermediates.Pseudomonas is the most efficient lignin degradation bacterium (88).The four closest neighbors of Pseudomonas-associated strains in the feces of giant pandas were Pseudomonas fluorescens PfO-1, Pseudomonas syringae pv.phaseolicola 1448A, Pseudomonas putida KT2440, and Pseudomonas aeruginosa PAO1.Pseudomonas fluorescens (89), Pseudomonas putida (36), and Pseudomonas aeruginosa (90) have high potential in lignin degradation.Although Pseudomonas syringae is generally considered a plant pathogen (91), its genome contains genes involved in the β-ketoadipate and homogentisate pathway (92).Furthermore, in vitro culture experiments showed that three Pseudomonas-associated strains isolated from the feces of wild giant pandas have a lignin degradation ability.Besides, multicopper oxidases gene was identified in Pseudomonas syringae (93), further confirming that Pseudomonas-associated bacteria in the feces of wild giant panda play a crucial role in lignin degradation.Pseudomonas is the most dominant bacterium in the gut of wild giant pandas (19,20).In this study, the abundance of Pseudomonas-associated bins (bin 1, bin 17, bin 33, and bin 53) was higher than other bins, implying that Pseudomonas is the main lignin-degrading bacteria in the gut of giant pandas.Laccase-like multicopper oxidase genes in the gut of giant pandas are mainly derived from Pseudomonas, consistent with the findings on microbiome abundance.Fang et al. reported that multicopper oxidase in the gut of giant pandas was derived from Pseudomonas sp., and the enzyme showed activity for oxidative degradation of lignin (23).The three Pseudomonas-associated strains could also secret extracellular Lac, Lip, and MnP in the culture medium with bamboo powder as the sole carbon source.This is the first study to show that Pseudomonas, the most dominant bacteria genus in the gut of giant pandas, can degrade lignin.
Based on these findings, we proposed a potential model for lignin degradation by gut microbiotas of giant pandas (Fig. 9).In this model, lignin is first depolymer ized by extracellular Lac, Lip, and MnP to form lignin derivatives or monomers (p-cou marate, ferulate, phenylpyruvate, etc.).Lignin monomers (p-coumarate and ferulate, etc.) are then degraded into aromatic compounds, such as vanillate, 4-hydroxyben zoate, and 2-hydroxyphenylacetate.The aromatic compounds are further metabolized to form fumarate, acetyl-CoA, and succinyl-CoA through the β-ketoadipate pathway and homogentisate pathway.Finally, fumarate, acetyl-CoA, and succinyl-CoA enter the TCA cycle of giant pandas for the biosynthesis of metabolically important products.Pseudomonas is the most dominant bacteria in the gut of wild giant pandas (19,20); thus, it plays a key role in this model.
Pseudomonas putida KT2440, a model strain for lignin degradation, can package enzymes implicated in lignin degradation in outer membrane vesicles (OMVs) and release them into the extracellular space to catalyze lignin degradation (36).Moreover, several Pseudomonas strains, including Pseudomonas aeruginosa (94) and Pseudomonas syringae (95), can release OMVs to facilitate the metabolism of lignin in vitro.This indicates that Pseudomonas-associated strains in the gut of giant pandas potentially release OMVs containing enzymes involved in lignin degradation into the extracellular space (in the gut of giant pandas) to degrade lignin.In this study, experiments of aniline blue degradation and enzyme activity determination in vitro confirmed that Pseudomo nas-associated strains can release extracellular ligninolytic enzymes to degrade lignin.In addition, Pseudomonas-associated strains can degrade lignin into some important raw materials, such as fumarate, acetyl-CoA, and succinyl-CoA, for the TCA in vitro.Notably, LC-MS did not detect acetyl-CoA and succinyl-CoA in the culture solution, indicating low content or other detection methods are needed.However, these results indicate that the gut of giant panda may assimilate the final products (fumarate, acetyl-CoA, and succinyl-CoA) of lignin degradation in the extracellular space into the TCA cycle to obtain nutritionally important compounds.This hypothesis effectively explains how giant pandas potentially obtain nutrition from a low-nutrition bamboo diet.However, additional studies are needed to verify if the intestinal tract of giant pandas can absorb lignin degradation products for biosynthesis metabolism.In summary, this study provides new insights into how giant pandas obtain nutrition from bamboo.

Conclusion
The gut microbiome of wild giant pandas exhibits high expression levels of genes implicated in lignin derivative degradation pathways.These pathways include catechol branch of beta-ketoadipate pathway, protocatechuate branch of beta-ketoa dipate pathway, homogentisate pathway of aromatic compound degradation, and pathways for degradation of lignin monomers (p-coumarate and ferulate) into the important raw materials for TCA via beta-ketoadipate pathway, such as acetyl-CoA and succinyl-CoA.All these pathways can be found in the genome of the most dominant bacteria genus, Pseudomonas-associated strains.Furthermore, results showed that Pseudomonas-associated strains isolated from the feces of pandas can degrade extracellular lignin in vitro.In general, the predominant bacteria in giant pandas, particularly Pseudomonas, may play a crucial role as important lignin-degrad ing agents.Their presence may have been instrumental in facilitating the adaptation of giant pandas to a bamboo-centric diet.

FIG 1
FIG 1 Metabolic pathways involved in lignin derivative degradation enriched in the gut microbiome of wild giant pandas.(A) Beta-ketoadipate pathway, red arrows represent the catechol branch of beta-ketoadipate pathway, green arrows represent the protocatechuate branch of beta-ketoadipate pathway, and black arrows represent parts of the metabolic pathway shared by catechol and protocatechuate branch.(B) Homogentisate pathway for degradation of aromatic compounds.(C) The metabolic pathway for catabolism of lignin monomers p-coumarate and ferulate into protocatechuate, which can be further degraded into acetyl-CoA and succinyl-CoA through protocatechuate branch of beta-ketoadipate pathway.(D) Average abundance of genes coding for enzymes implicated in the beta-ketoadipate pathway, homogentisate pathway, and p-coumarate and ferulate degradation pathway in the gut of bamboo-eating pandas, terrestrial mammals, and wood-feeding insects presented as a heat map (log 10 abundance).

FIG 2
FIG 2Relative abundance of genes encoding enzymes for beta-ketoadipate pathway, homogentisate pathway, p-coumarate, and ferulate degradation pathway in the gut of bamboo-eating pandas, terrestrial mammals, and wood-feeding insects presented as box plots (P values were calculated by the Kruskal-Wallis test followed by Dunn's multiple-comparison post-hoc test: *P < 0.05, **P < 0.01, and ***P < 0.001).

FIG 3 (
FIG 3 (A) Phylogenetic analysis of the high-quality strain draft genomes (bins) from metagenome sequencing and the relative abundance of these strains in the gut of wild giant pandas.The left panel shows the neighbor-joining tree using high-quality strain draft genomes (bins) obtained from metagenome sequencing.The numbers in parentheses on the phylogenetic tree represent the bootstrap value of the nodes.The right panel represents the average abundance of each strain draft genomes (bins) in the gut of wild giant pandas.(B) Functional annotation of the genome of Pseudomonas-associated bin (bin 1, bin 17, bin 33, and bin 52) in the gut of wild giant panda.Annotation results of the four genomes were similar; thus, one bin (bin 1) was selected to present the functional annotation results.

FIG 4
FIG 4 Phylogenetic analysis of Pseudomonas-associated genome (bin) in the gut of wild giant panda and the published genomes of Pseudomonas fluorescens, Pseudomonas syringae pv.phaseolicola, Pseudomonas putida and Pseudomonas aeruginosa.Genomes of 10 different strains for each Pseudomonas species were included in the phylogenetic tree (neighborjoining) analysis.The numbers in parentheses on the phylogenetic tree represent the bootstrap value of the nodes.

FIG 5
FIG 5 Phylogenetic analysis and composition of bacterial species with laccase-like multicopper oxidase gene derived from the feces of wild giant pandas.(A) Neighbor-joining phylogenetic tree constructed using laccase-like multicopper oxidase gene sequences from the feces of wild giant pandas and the most similar alignment sequence obtained from GeneBank.One laccase-like multicopper oxidase gene sequence was randomly selected from each identified bacterial species.The numbers on the phylogenetic tree represent the bootstrap value of the node.(B) Pie plots showing the composition of bacterial species of laccase-like multicopper oxidase gene in the feces of wild giant pandas.
. The activities of Lac, LiP, and MnP reached the maximum values after 72 h.The maximum values of Lac, LiP, and MnP were 688.25 U/L, 3065.37 U/L, and 493.22 U/L, respectively, for Pseudomonas putida; 653.75 U/L, 2700.54U/L, and 439.53 U/L, respectively, for Pseudomonas sp.; and 459.12 U/L, 2668.54U/L, and 404.26U/L, respectively, for Pseudomonas oryzihabitans.The average Lac activity of Pseudomonas putida, Pseudomonas sp., and Pseudomonas oryzihabitans was 309.51, 251.58, and 209.52 U/L, respectively (Fig.

FIG 7
FIG 7 Analysis of extracellular enzyme activity of Pseudomonas-associated strain.(A) Decolorization efficiency of Pseudomonas-associated strain on aniline blue solid medium.(B) Decolorization rate curve of aniline blue B. Activities of the Lac (C), LiP (D), and MnP (E) during the incubation of Pseudomonas-associated strain.

FIG 9
FIG9 The putative model for lignin degradation by gut microbiotas of giant pandas based on the findings of study.Pseudomonas genus bacteria play a central role in the model.

TABLE 1
Species and potential functions identification of individual bacteria draft genomes in wild giant pandas (Continued on next page)

TABLE 1
Species and potential functions identification of individual bacteria draft genomes in wild giant pandas(Continued) a NA indicates the absence of genes involved in this pathway in the genome.

TABLE 2
Partial aromatic compounds identified in the control alkali lignin medium (0 day) and the alkali lignin medium degraded by Pseudomonas-associated strain for 3 and 7 days a