Abstract

Background

   Recent studies suggest that intratumoral microbiome and altered
   metabolic networks play crucial roles in pancreatic cancer (PC)
   progression. However, the precise interplay between microbial
   communities and tumor metabolism in PC remains poorly understood. This
   study aims to investigate the impact of the intratumoral microbiome,
   the metabolic landscape, and their interactions on PC development.

Methods

   16S rDNA sequencing and Untargeted metabolomic profiling were performed
   on 47 paired pancreatic cancer and adjacent normal tissues to analyze
   their intratumoral microbiome and metabolic landscapes. Bioinformatics
   tools were used to conduct differential microbiome abundance analysis
   and pathway enrichment. A correlation analysis was performed to
   identify key microbiota-metabolite interactions.

Results

   16S rDNA sequencing revealed significant differences in the abundance
   and diversity (α-diversity and β-diversity) of the intratumoral
   microbiome in PC. The predominant species in pancreatic cancer were
   Pseudomonas. Enrichment analysis showed that amino acid metabolic
   pathways, including Arginine and Proline Metabolism, Arginine
   Biosynthesis, were significantly enriched in PC. Untargeted
   metabolomics identified 298 metabolites that were significantly altered
   in PC (fold change > 1.5, P-value < 0.05). These included amino acid
   metabolites such as Lys-Leu, Pro-Leu, Arg-Leu, Lys-Val, His-Lys, and
   others. Functional enrichment analysis highlighted several metabolic
   pathways that play important roles in pancreatic cancer, including
   Glycine, Serine, and Threonine Metabolism, Amino Acid Biosynthesis,
   Metabolic Pathways and Cysteine and Methionine Metabolism. Correlation
   analysis between microbiome and metabolic data revealed significant
   associations between Pseudomonas and several metabolites, including
   Alpha-ketoisovaleric acid, 16-hydroxyhexadecanoic acid, Myristic acid,
   Nonanoic acid (the Spearman correlation coefficient r, 0.5 ≤|r|≤ 1 and
   P-value < 0.05).

Conclusion

   This study suggests a relationship between the microbiome and
   metabolism in pancreatic cancer. We observed that Pseudomonas
   contributes to altered amino acid metabolism, but whether this
   interaction is causal and the mechanisms underlying it remain unclear.
   Further experimental validation is required before considering
   microbiome-targeted metabolic interventions as viable therapeutic
   strategies.

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s12943-025-02458-9.

   Keywords: Pancreatic cancer, Intratumoral microbiome, Metabolic
   landscape, Microbiota-metabolite interactions

Introduction

   Pancreatic cancer (PC) has a rising incidence and is one of the most
   lethal human malignancies [[36]1]. PC is projected to become the
   second-leading cause of cancer-related deaths by 2040 in the United
   States [[37]2], with its 5-year relative survival rate at only 13%
   [[38]1], which is the lowest among all types of cancer.

   The current standard of care for pancreatic cancer is a systemic
   treatment paradigm that comprising surgery, chemotherapy, immunotherapy
   and targeted therapy. Early symptoms of pancreatic cancer are often
   subtle, and the majority of patients are diagnosed at a locally
   advanced or metastatic stage. Surgical resection is the sole effective
   method for achieving cure and long-term survival in pancreatic cancer
   patients; however, less than 20% of patients have the opportunity for
   curative surgical resection [[39]3], approximately 80% of who undergo
   curative surgery will experience recurrence within 2 years
   postoperatively [[40]4]. Chemotherapy is the standard of care for
   pancreatic cancer patients, but the response rate to chemotherapy is
   only 30% and is prone to drug resistance [[41]5]. Only a very small
   number of pancreatic cancer patients benefit from immunotherapies and
   targeted therapy approaches [[42]4, [43]6]. Pancreatic cancer has a
   very poor prognosis due to the difficulty of early diagnosis, low
   resection rate, and poor therapeutic efficacy. Therefore, there is an
   urgent need to find novel molecular mechanisms and potential targets.

   A defining feature of the tumor microenvironment (TME) in pancreatic
   cancer is the intricate interplay between the internal and external
   mechanisms of cancer cells, which gives rise to a TME characterized by
   a dense extracellular matrix, metabolic reprogramming, and pronounced
   immunosuppression [[44]7, [45]8]. To adapt to the severe hypoxia
   environment, cells have changed their metabolic phenotypes to maintain
   their survival and proliferation, including enhanced glycolysis
   metabolism, glutamine metabolism and fatty acids metabolism [[46]9].
   Adaptive metabolic changes of cancer cells promote their survival and
   affect tumor-infiltrating immune cells in the TME, which contributes to
   the immunosuppressive microenvironment and induces tumor immunotherapy
   resistance [[47]9, [48]10].

   There is a close link between the intratumoral microbiota and host
   metabolic state, and this interplay may influence tumor metabolic
   reprogramming. The intratumoral microbiome exerts tumor-promoting or
   tumor-suppressive effects by engaging in metabolic reactions within the
   body, regulating cancer-related signaling pathways, and impacting both
   host cell function and immune system [[49]11]. Microbiota-derived
   metabolites represent important mediators of host-microbe interactions
   and play a profound role in physiological functions and the immune
   environment. These microbes are involved in regulating host mechanisms
   such as detoxification and nutritional pathways, as well as influencing
   tumor metastasis and the efficacy of chemotherapy [[50]12].

   Accumulating evidence suggests that microbiota (specifically specific
   well-studied oral, gastrointestinal, and intrapancreatic microbes, as
   well as certain hepatotropic viruses and bactibilia) and their
   metabolites may have potential etiological roles in altering the
   microenvironment of pancreatic tumors, impacting the functionality of
   the immune system, and regulating the biological behavior of pancreatic
   cancer cells, those actions can promote or inhibit the progression of
   pancreatic tumors [[51]13, [52]14]. Concrete mechanisms include
   perpetuating inflammation, regulating the immune system-microbe-tumor
   axis, affecting metabolism, and altering the tumor microenvironment
   [[53]14]. Intratumoral microbiota and their products can regulate the
   microenvironment of pancreatic tumors, the biological behavior of
   pancreatic cancer cells, and the functionality of the immune system
   [[54]15]. Three potential mechanisms of intratumoral microbiota
   promoting pancreatic tumorigenesis: secretion of metabolites to cause
   DNA damage, activation of oncogenic pathways and alteration the tumor
   immune microenvironment [[55]15].

   Metabolomics reveals the metabolic characteristics of PC. Despite
   existing in a relatively hypoxic and nutrient-poor environment, PC
   achieves its metabolic advantage through three mechanisms: (i)
   reprogramming intracellular energy metabolism, (ii) enhancing nutrient
   acquisition via scavenging and recycling (autophagy), and (iii)
   facilitating metabolic interactions with other components in its
   microenvironment [[56]16]. PC can be stratified into glycolytic and
   lipogenic subtypes [[57]17]. These subtypes can interconvert (e.g.,
   transitioning from glycolysis to oxidative phosphorylation) under
   varying metabolic demands. Such plasticity may be closely linked to
   tumor aggressiveness, therapy resistance, and microenvironmental
   adaptation. Distinct metabolic profiles may influence the development
   of PC through metabolites and metabolic pathways [[58]18], underscoring
   the pivotal role of amino acid metabolism in the progression of PC
   [[59]19].

   Investigating the role of the intratumoral microbiome in the diagnosis
   and treatment of pancreatic cancer using 16S sequencing has emerged as
   an active area of research. The tumor microbiome composition correlates
   with distinct metabolic pathways in PC patients with short-term
   survival (STS) versus long-term survival (LTS) [[60]20], highlighting
   that intraltumor microbiome influences the prognosis of PC patients
   through lipid metabolism and protein anabolic pathways. Studies have
   revealed that the microbiota of pancreatic cancer can reshape the
   immune microenvironment through their metabolites; however, the roles
   of specific metabolites (such as amino acids) and their metabolic
   pathways remain insufficiently characterized [[61]21, [62]22].

   As seen from the above discussion, current research predominantly
   focuses on the association of individual microbiota or metabolites with
   PC; however, the causal mechanisms governing their dynamic interplay
   (e.g., how specific microbial communities mediate oncogenic metabolite
   synthesis) remain systematically underexplored. Traditional
   single-omics approaches fail to capture complex interactions within
   multifaceted biomolecular networks, resulting in fragmented insights
   into host-microbiota metabolic crosstalk. Therefore, systematically
   mapping the interactions between intratumoral microbiota and metabolism
   in PC is critical for advancing early diagnosis and targeted therapy of
   the disease.

   Microbiome-induced alterations in human metabolism may contribute to
   diverse metabolic diseases and cancers [[63]23]. The intratumoural
   microbiota modulate cancer initiation and progression through metabolic
   regulation [[64]24], including in hepatocellular carcinoma [[65]25],
   breast cancer [[66]26], gastric cancer [[67]27], and cervical cancer
   [[68]28]. Most current studies have focused on the effects of gut
   microbes on PC [[69]13]. Changes in gut microbiota may affect host
   metabolic status, tumor metabolic reprogramming and influence
   pancreatic cancer progression by modulating immune responses [[70]13,
   [71]29]. Although studies have performed microbiome and metabolome
   analyses to investigate how smoking influences PC progression via
   alterations in microbial communities and metabolic pathways, the
   relationship between specific microbes and amino acid metabolism
   remains unexplored [[72]30]. Therefore, studies on the integration of
   microbes and metabolites in intratumoral tissues are still
   insufficient. The relevance of specific intratumoral microbiota to
   metabolites and metabolic pathways in PC tissues and their potential
   mechanisms of action remain unknown.

   The integration of multi-omics approaches—leveraging data-level
   cross-validation and interaction model construction—can elucidate key
   regulatory hubs in tumor progression and holds promise for advancing
   precision oncology [[73]31]. Multi-omics approaches, which are novel
   frameworks, integrate multiple omics datasets generated from the same
   patients to better understand the molecular and clinical features of
   cancer [[74]32]. Zhu et al.’s study, employing microbiome and spatially
   resolved metabolomics analyses, revealed how Akkermansia muciniphila
   exerts anticancer effects in a murine model through interactions with
   intratumoral microbiota and metabolic reprogramming [[75]33], providing
   a new potential target for the prevention and treatment of gastric
   cancer. Another study integrated multi-omics data (microbiome,
   transcriptome, and metabolome) to reveal that Streptococcus anginosus
   modulates the arginine metabolic pathway, elevating ornithine levels
   and thereby remodeling the tumor immune microenvironment (TIME) to
   promote gastric cancer progression [[76]27]. Multi-omics approach
   elucidated the impact of intratumoral microbes on neuron-related gene
   expression in glioma, mediated by microbial metabolites, suggesting
   that Fusobacterium nucleatum may serve as a potential diagnostic and
   therapeutic target for glioma [[77]34]. Furthermore, flora analysis,
   mass spectrometry (MS), and integrated metabolomics demonstrate that
   coordinated microbiota modulation alters the abundance and diversity of
   intratumoral microbiota and disrupts metabolic pathways within both the
   microbiota and tumor microenvironment [[78]35]. Through these
   approaches, distinct signaling pathways and biosynthetic demands or
   metabolites influencing tumor progression have been identified. The
   clear underlying principles of microbiome-regulated tumorigenesis and
   the established microbiome metabolism regulation methods provide
   distinctive insights into tumor therapy [[79]35].

   Therefore, we proposed the scientific hypothesis of this project:
   specific intratumoral microbiota within pancreatic cancer tissues are
   involved in pancreatic cancer progression by regulating
   metabolism-related products (e.g., amino acids). The aim of this
   article is to combine 16S sequencing (microbiome) and untargeted
   metabolomics data to systematically reveal the microbial-metabolite
   interaction network in pancreatic cancer tissues and to elucidate its
   role in tumor progression.

Methods

Pancreatic samples sources

   All human samples (including 47 cases of pancreatic cancer and adjacent
   tissues), as well as the clinical data, were obtained from our
   hospital. This study was conducted in accordance with the regulations
   of the tissue biobank and approval by the Ethics Committee of our
   hospital. All patients provided written informed consent.

Experimental techniques and procedures

Untargeted metabolomic analysis(LC–MS/MS)

   PC samples were thawed slowly at 4℃ and metabolites were extracted
   according to the experimental procedure. LC–MS/MS analysis was
   performed using an UHPLC (1290 Infinity LC, Agilent Technologies)
   coupled to a quadrupole time-of-flight (AB Sciex TripleTOF 6600), in
   both positive and negative ion modes. The raw MS data were converted to
   MzXML files using ProteoWizard MSConvert before importing into freely
   available XCMS software. After sum-normalization, the processed data
   were analyzed by R package (ropls). Data analysis includes univariate
   statistical analysis, multidimensional statistical analysis,
   differential metabolite screening, differential metabolite metabolic
   pathway enrichment, and other bioinformatics analyses. Differential
   analysis between two sample groups often utilizes univariate
   statistical methods, including Fold Change Analysis (FC Analysis) and
   non-parametric tests (e.g., t-test). Based on the results of univariate
   analysis, differential analysis is performed on all detected
   metabolites, including those that are unidentified. Metabolites were
   classified as differential if they meet the criteria of FC > 1.5 or
   FC < 0.67 and have a P value < 0.05 (t-test). These differential
   metabolites were then visualized using a volcano plot. Multivariate
   data analysis was performed with orthogonal partial least-squares
   discriminant analysis (OPLS-DA). The variable weight values obtained
   from the OPLS-DA model, known as Variable Importance for the Projection
   (VIP), can be used to measure the strength of influence and explanatory
   power of the expression patterns of metabolites on the classification
   and discrimination of samples between groups. Metabolites with a VIP
   value > 1 were generally considered to have a significant contribution
   to the model interpretation. In this experiment, metabolites with an
   OPLS-DA VIP value > 1 and a T-test P value < 0.05 were selected as
   significantly different metabolites for subsequent bioinformatics
   analysis. Functional analysis of differential metabolites encompasses
   KEGG pathway annotation, MetPA enrichment, and metabolite set
   enrichment analysis (MSEA).

Microbiome analysis (16S rDNA sequencing)

   DNA extraction, detection, PCR amplification, purification of
   amplification products, fluorescence quantification, library
   preparation, and V3-V4 variable region amplification and sequencing
   were performed using the Illumina NovaSeq 6000 sequencing platform. Raw
   data obtained from sequencing were processed using the default
   parameters of QIIME2 for quality control analysis to obtain Amplicon
   Sequence Variants (ASVs) for ASV/Operational Taxonomic Units (OUT)
   clustering and annotation. Subsequently, the data analysis was carried
   out, including community structure analysis, alpha (α) diversity
   analysis, beta (β) diversity analysis, species difference analysis and
   function prediction. Specifically, α and β diversity indices were
   calculated using QIIME2 software (with default parameters). α diversity
   (including Chao 1, Shannon and Simpson indices) was used to analyze the
   diversity of microbial communities within the sample. β Diversity
   compared the microbial community composition of different samples.
   Anosim analysis was used to examines whether the differences between
   groups were significantly greater than the differences within groups.
   The Wilcoxon rank-sum test was used to determine whether differences in
   species diversity between groups were statistically significant
   (P < 0.05 indicates a significant difference). Linear discriminant
   analysis effect size analysis (LEfSe, LDA analysis) was used to
   identify differentially abundant species. With only those having an LDA
   Score > 2 (default set at 2) displayed in the figure. The length of the
   bar represents the magnitude of the LDA value. Microbial function and
   pathway enrichment were analyzed based on KEGG database by the PICRUSt
   software (with the default parameters) to infer the Functional gene
   composition of samples by comparing the species composition information
   obtained from 16S sequencing data, this allowed for the analysis of
   Functional differences between different samples or groups. Due to the
   wide variation in the expression of individual abundance of
   intratumoral microbiota in pancreatic adenomas, only 16 patients in the
   microbiome were included in the subsequent data analysis based on PCoA
   component analysis.

Multi-omics correlation analysis

   Microbiota with significant differences identified in the 16S
   microbiome were integrated with the metabolites that showed significant
   differences in the metabolome. Spearman correlation analysis was used
   to identify associations between significantly different flora (LEfSe
   LDA > 2 and P value < 0.05) and significantly different metabolites
   (VIP > 1 and P value < 0.05). A microbial-metabolite interaction
   network was then constructed, including Spearman rank correlation
   analysis and Mantel test network graphs), focusing on analyzing the
   functional roles of core microbiota and metabolic nodes.

Correlation Analysis of Clinical Data

   ①Correlation analysis of clinical staging: The correlation analysis
   between alpha diversity indices (ACE, Chao1, Shannon, and Simpson
   indices) and clinical staging was conducted. Additionally, the relative
   abundance of Pseudomonas was analyzed in relation to clinical staging.
   The analysis was performed using R software, employing t-tests, with a
   P-value less than 0.05 indicating a statistically significant
   difference. ②Survival analysis of Pseudomonas: Survival analysis was
   first performed using public data from the TCMbio database
   ([80]https://microbiomex.sdu.edu.cn/). To conduct the analysis,
   navigate to "Survival Analysis—Microbe," select "PAAD" as the Cancer
   Type, and enter "Pseudomonas" in the "Enter microbe name" field. For
   overall survival (OS) analysis, set the Group Cutoff value to 70%, and
   for disease-free interval (DFI), use the default Median (50%). Keep all
   other settings as default and click "Plot" to obtain the survival
   curves. In our cohort, Kaplan–Meier survival analysis was performed
   using the median abundance of Pseudomonas as the grouping criterion.

Statistical analysis

   Data were analyzed by R (version 3.4.2). Differences between two groups
   were considered statistically significant at P < 0.05.

Results

Significant differences between the microbiomes of pancreatic cancer and
adjacent tissues

   Differences in the number of species detected were observed between the
   pancreatic cancer tissues (T group) and the adjacent tissues (N group)
   (Fig. [81]1A). Analysis of α-diversity using the Shannon and Simpson
   indices (Fig. [82]1B-D) and β-diversity using ANOSIM analysis
   (Fig. [83]1E), and the PCOA analysis (Fig. [84]1F) based on the
   Braycurtis distance revealed that significant differences in the
   α-diversity and β-diversity between cancerous tissues (T) and adjacent
   (N). These results revealed the significant differences in species
   richness and community composition between cancerous tissues (T) and
   adjacent tissues (N), suggesting the distinct intratumoral microbiota
   compositions between the two groups (Fig. [85]1). LEfSe analysis was
   used to determine the dominant species between two groups. Using an LDA
   score > 2 as the screening criterion, the dominant species in the
   intratumoral tissues of pancreatic cancer were Pseudomonadales,
   Pseudomonadacea, and Pseudomonas (Fig. [86]2A), while those in the
   paracancerous tissues were Alphaproteobacteria, Moraxellaceae,
   Acinetobacter, Aeromonadaceae, Aeromonadales,Aeromonas. KEGG analysis
   (based on computer calculations) highlighted several key metabolic
   pathways in pancreatic cancer: arginine and proline metabolism,
   arginine biosynthesis, cationic antimicrobial peptide (CAMP)
   resistance, geraniol degradation, glutamatergic synapse,
   GABAergic_synapse, necroptosis and (Fig. [87]2B).

Fig. 1.

   [88]Fig. 1
   [89]Open in a new tab

   Analysis of intra-tissue bacterial diversity between PC (T) and
   adjacent (N) group. A Venn diagram of detected between PC and adjacent
   group; B Chao 1 index; C Simpson index; D Shannon index; E anosim
   index; F PCoA showing significant differences between two groups (the
   horizontal axis shows the first principal component (PCoA1), and the
   vertical axis shows the second principal component (PCoA2). The
   percentages show how much each component contributes to the differences
   between the samples). Note: group: T; adjacent group: N. A-E The
   horizontal axis represents different groups; the vertical axis
   represents the alpha and beta diversity index (Chao 1, P = 0.84;
   Simpson, P = 0.021; Shannon, P = 0.026; and anosim index, P = 0.001).
   P < 0.05 indicates statistical significance

Fig. 2.

   [90]Fig. 2
   [91]Open in a new tab

   Functional analysis of dominant species and differential intratumoral
   microbiota between PC (T) and adjacent (N) groups. A Significantly
   different intratumoral microbiota between PC and adjacent (N)
   group; B KEGG pathway enrichment bar graphs of different species
   between two groups. Note: The LDA value distribution bar chart features
   a red area for the adjacent tissue group (N) and a green area for the
   tumor group (T), indicating distinct groups. In the tree branches, red
   nodes represent important microbial taxa in the adjacent tissue group
   (N), while green nodes represent those in the tumor group (T). The
   figure shows species with an LDA Score > 2 (The default cutoff value
   for LDA is set at 2), with the length of each bar reflecting the
   magnitude of the LDA value

Metabolomics results showed significantly different metabolites and their
pathways in pancreatic cancer

   Untargeted metabonomics was used to compare metabolites difference
   between pancreatic cancer group (T) and adjacent group (N). The model
   demonstrated good stability and effectiveness (Fig. [92]3A-D). A
   volcano map identified 298 significantly different metabolites between
   two groups, including both up-regulated and downregulated metabolites
   (Fig. [93]3E-F).

Fig. 3.

   [94]Fig. 3
   [95]Open in a new tab

   Analysis of model stability and metabolite changes in PC (T) and
   adjacent group (N). A-B metabolite OPLS-DA score chart (positive and
   negative ions). The figure illustrates the first principal component on
   the horizontal axis and the second principal component on the vertical
   axis, with ellipses indicating the 95% confidence intervals. Points of
   the same color represent biological replicates within a group, and
   their distribution demonstrates the differences both between and within
   groups; C-D Histogram of permutation of metabolites (positive and
   negative ions). In the figure, the horizontal axis represents the
   retention rate of permutations, which is the proportion consistent with
   the original order of the Y variable in the model. The vertical axis
   indicates the values of R^2 and Q^2. The green points denote R^2, while
   the blue points denote Q^2. The two dashed lines represent the
   regression lines for R^2 and Q^2, respectively. The R^2 and Q^2 values
   in the upper right corner correspond to a retention rate of 1, which
   means they are the R^2 and Q^2 values of the original model; E–F.
   Volcanic diagram of differential metabolites (positive and negative
   ions). The horizontal axis shows the log2 value of the fold change
   (log2 Fold Change), while the vertical axis shows the negative log10
   value of the significance p-value (-log10 p-value). Metabolites with a
   fold change > 1.5 and P < 0.05 are highlighted in red, while those with
   a fold change < 0.67 and P < 0.05 are highlighted in blue.
   Non-significantly different metabolites are indicated in grey

   Analysis of metabolite differences revealed that the following
   metabolites in positive ion mode were significantly upregulated in
   cancer tissues: Zaleplon, Derythromimidazolyl Glycol Phosphate,
   Dobutamine, Pro − hyp, 1 − methyl − 3 − isobutylxanthine,
   N − methylnicotinamide, Terbutaline, Octanoylcarnitine and Tryptophan
   betaine. Conversely, the following metabolites were downregulated:
   Lys − Leu, Pro − phe, Pro − leu, Pro − gln, methyltyrosinate,
   Pro − ser, Arg − Leu, Phe − Ile, Lys − Val, His − Lys,
   Dl − Leu − dl − Val, Met − Pro − Lys and Asp − Leu − Lys (Fig. [96]4A).
   In negative ion mode, the following metabolites were significantly
   upregulated: α- Hydroxy3oxo4cholestronic acid,
   S − carboxymethyl − l − cysteine, N − acetyl − d − galactosaminitol and
   D − fructose 6 − phosphate. Meanwhile, the following metabolites were
   downregulated: Trp − Leu, Trp − Val, Thr − Val − Leu, Val − Gly − Va,
   Pro − Ala, Thr − Leu, Val − Glu, Glu − Arg, Val − Ile and Ile − Thr
   (Fig. [97]4B). These results demonstrate significant differences in
   amino acid metabolism in pancreatic cancer tissues (T) compared to
   adjacent tissues (N).

Fig. 4.

   [98]Fig. 4
   [99]Open in a new tab

   Screening of metabolites with significant differences between two group
   (T vs. N). A Metabolites (positive ions) with significant differences
   between PC and N group; B Metabolites (negative ions) with significant
   differences between PC and N group. Note: Significant differential
   metabolites are intuitively displayed using a butterfly plot. The
   horizontal axis shows the log2 fold change (FC) values of differential
   metabolites, derived from base 2 logarithms of their fold changes. The
   vertical axis displays the significantly different metabolites, with
   red indicating upregulated and blue indicating downregulated
   metabolites

   KEGG analysis identified several metabolic pathways that play important
   roles in pancreatic cancer: Glycine, serine and threonine metabolism,
   Nucleotide metabolism, Pantothenate and CoA biosynthesis、Biosynthesis
   of amino acids, ABC transporters, Metabolic pathways, Protein digestion
   and absorption、Alanine, aspartate and glutamate metabolism, Cholesterol
   metabolism, Insulin resistance, Choline metabolism in cancer, D − Amino
   acid metabolism, Bile secretion, beta − Alanine metabolism, Insulin
   secretion, Primary bile acid biosynthesis, Taurine and hypotaurine
   metabolism, Aminoacyl − tRNA biosynthesis、Cysteine and methionine
   metabolism (Fig. [100]5A). Met PA analysis highlighted the most
   relevant metabolic pathways, including alanine, aspartate and glutamate
   metabolism, taurine and hypo taurine metabolism, glycine, serine and
   threonine metabolism, pantothenate and CoA biosynthesis and
   biosynthesis of unsecured fatty acids (Fig. [101]5B). MSEA analysis, a
   supplementary mode of KEGG analysis, further identified highly
   correlated pathways such as: Neuroactive Ligand Receiver Interaction,
   Sulfur metabolism, Taurine and Hypo Taurine Metabolism, Pyrimidine
   metabolism, Linoleic acid metabolism、Biosynthesis of nucleotide
   sugars、Primary bile acid biosynthesis、Galactose metabolism、Metabolic
   pathways、 Histidine metabolism and biosynthesis of unstructured fatty
   acids (Fig. [102]5C). Therefore, we will focus on amino acid
   metabolism, particularly glutamate metabolism (involved in glutamine
   metabolism pathway) and its intermediates, in pancreatic cancer.

Fig. 5.

   [103]Fig. 5
   [104]Open in a new tab

   Enrichment pathway analysis of significantly different metabolites
   between PC group (T) and adjacent group (N). A Bar graph of
   significantly different metabolite KEGG enrichment analysis pathway
   between the two groups (The bar chart illustrates KEGG metabolic
   pathways on the vertical axis and the p-value of the enrichment
   analysis on the horizontal axis, expressed as the negative common
   logarithm (-log10(p-value)). Each bar's height indicates the number of
   differentially abundant metabolites enriched in the corresponding
   pathway. Additionally, the color of the bars represents the rich factor
   for each pathway, with darker shades of red indicating higher values.
   The rich factor shows the proportion of differentially abundant
   metabolites relative to the total annotated metabolites in the
   pathway); B Bubble graph of significantly different metabolite Met PA
   analysis pathway between the two groups (Each circle represents a
   distinct metabolic pathway. The horizontal axis indicates how much the
   pathway is impacted, referred to as Pathway Impact. The size of each
   circle correlates with Pathway Impact; larger impact values produce
   larger circles. The vertical axis shows the negative logarithm of the
   P-value from the pathway enrichment analysis. The color gradient,
   ranging from blue to red, indicates a positive correlation with the
   negative logarithm of the P-value for pathway changes. The redder the
   color, the more significantly enriched the pathway is.); C Bar chart of
   significantly different metabolite MSEA analysis pathway between two
   groups (The vertical axis shows the names of the metabolite sets and
   their annotated p-values. In contrast, the horizontal axis illustrates
   the degree of enrichment, where the color corresponds to the p-value
   from the enrichment analysis. As the color darkens, the p-value
   decreases, indicating a more significant enrichment analysis)

Interaction between significantly different intratumoral microbiota and
significantly different metabolites

   The association analysis between 16S sequencing data and metabolomics
   primarily utilized statistical algorithms to identify significant
   correlations between differentially abundant intratumoral microbiota
   and metabolites. Spearman correlation analysis revealed significant
   associations between differentially abundant microbiota and
   metabolites, as well as among the metabolites themselves and within the
   intratumoral microbiota. These correlations were visualized as a
   heatmap of the correlation coefficient matrix (Supplementary Fig. 1).
   Spearman and Mantel test correlation network plots showed that
   Pseudomonas may negatively regulate tumor progression by downregulating
   metabolites such as Alpha-ketoisovaleric acid (a metabolic intermediate
   of valine, a branched-chain amino acid [BCAA]), 16-hydroxyhexadecanoic
   acid, and Myristic acid. Conversely, Pseudomonas may promote tumor
   progression by upregulating metabolites like Acetylcarnitine, Nonanoic
   acid, and Melamine (Fig. [105]6). Combined analysis of metabolomes and
   microbiomes highlighted the prominent role of amino acid metabolism in
   PC. Therefore, the interaction among Pseudomonas, amino acids, and
   their intermediate metabolites may regulate the initiation and
   progression of pancreatic cancer.

Fig. 6.

   [106]Fig. 6
   [107]Open in a new tab

   Network diagram of correlation between significantly differential
   intratumoral microbiota and significant differential metabolites in PC
   and adjacent group A Mantel Test analysis of correlation for the top 25
   differential metabolites and differential flora with the smallest
   significant p-value. (Note: The thickness of the connecting lines
   indicates the correlation between the two matrices, while the color of
   the lines represents the significance of the correlation's p-value; the
   matrix plot illustrates the correlation among differentially abundant
   metabolites) B Spearman correlation analysis network plot of
   significant differential intratumoral microbiota with significant
   differential metabolites (Note: In the figure, circles represent
   significantly different microbial genera, and rectangles represent
   significantly different metabolites. The color of the lines indicates
   the sign of the correlation coefficient between them: blue for negative
   correlation, red for positive correlation. While the thickness of the
   lines is proportional to the absolute value of the correlation
   coefficient. Node size is positively correlated with its degree;
   higher-degree nodes are larger

The clinical significance of differential abundance of Pseudomonas

   No significant correlation between alpha diversity (ACE, Chao 1,
   Shannon, and Simpson indices) and the stages (stages I, II, and III)
   was found in pancreatic cancer (Supplementary Fig. 2). Similarly, there
   was no significant correlation between the relative abundance of
   Pseudomonas and the T stage, N stage, or clinical stages (Supplementary
   Fig. 3). Overall survival (OS) analysis by the Cancer Microbiome
   Database (TCMbio) showed that higher levels of Pseudomonas are linked
   to shorter overall survival (OS, P = 0.049) and disease-free intervals
   (DFI, P = 0.015) in patients with pancreatic cancer (Fig. [108]7 A-B).
   Kaplan–Meier curves revealed that high abundance of Pseudomonas in
   pancreatic cancer is associated with reduced overall survival in our
   cohort (log-rank P = 0.037, Fig. [109]7C).

Fig. 7.

   [110]Fig. 7
   [111]Open in a new tab

   Kaplan–Meier survival analysis stratified by Pseudomonas abundance. A
   Overall survival analysis of Pseudomonas based on TCMbio database, B
   Disease-Free Interval analysis of Pseudomonas based on TCMbio database,
   C Overall survival analysis of Pseudomonas from our cohort (n = 42)

Discussion

   Microbiome analysis revealed that the predominant intratumoral
   microbiota in pancreatic cancer tissues is Pseudomonas, while
   metabolomic profiling identified significant alterations in amino acid
   metabolism particularly in glutamate metabolism, which is involved in
   glutamine metabolism pathway. Multi-omics correlation analysis further
   demonstrated a strong association between Pseudomonas and amino acid
   metabolic pathways, including alpha − ketoisovaleric acid (a metabolite
   derived from the branched -chain amino acid valine [[112]36]),
   suggesting that the interaction may serve as a key regulatory mechanism
   within the pancreatic cancer microenvironment. The results of this
   study not only uncover a specific regulatory mechanism in the
   pancreatic cancer microenvironment (i.e., the enrichment of Pseudomonas
   and its significant correlation with amino acid metabolic pathways),
   but also expand the frontier of research on metabolism-microbe
   interactions. Specifically, differences in amino acid metabolism are a
   critical feature of metabolic reprogramming in pancreatic cancer cells.
   Pseudomonas may influence tumor cell growth and invasive potential by
   modulating specific amino acid metabolic pathways.

   Our findings indicated that Pseudomonas was significantly enriched in
   pancreatic cancer tissues, where it serves as the predominant
   intratumoral microbiota. The severe hypoxia, immunosuppressive
   microenvironment, and disorganized vascular system within tumors
   collectively create favorable conditions that facilitate rapid
   bacterial colonization, proliferation, and replication within tumor
   tissues [[113]11, [114]37]. Furthermore, the distribution of microbiota
   within a tumor is not random; rather, it is highly organized into
   microniches that promote cancer progression through interactions with
   immune and epithelial cells [[115]38]. The intratumor bacteria are
   mostly intracellular and are present in both cancer and immune cells.
   Therefore, bacteria are more preferentially present in tumor tissues
   rather than in the surrounding normal tissues [[116]39]. More
   importantly, studies found that the cancerous pancreas harbors a
   markedly more abundant microbiome compared with normal pancreas in both
   mice and humans [[117]21, [118]22]. Multi-omics analysis further
   suggested that Pseudomonas plays a pivotal role in shaping the
   pancreatic cancer microenvironment, particularly through the regulation
   of amino acid metabolism. Studies have shown that the intratumoral
   microbiota in pancreatic cancer is closely linked to tumorigenesis,
   progression, immune response, drug resistance, and prognosis [[119]14,
   [120]15, [121]22, [122]40, [123]41]. While Pseudomonas has been
   implicated in inflammation and metabolic disorders in some tumors
   [[124]42, [125]43], and Pseudomonas has been identified as the most
   abundant genus of bacteria transported from the gut to the pancreas in
   patients with PC [[126]21] and is significantly associated with
   prognosis [[127]20]. However, its specific functions and underlying
   mechanisms in pancreatic cancer remain poorly understood. In the
   present study, we demonstrated that Pseudomonas acts as a dominant
   bacterial species within the pancreatic cancer microenvironment,
   regulating key metabolic pathways. Our findings underscore the critical
   role of microbe-metabolite interactions in the pathogenesis of
   pancreatic cancer by integrating microbiomic and metabolomic analyses.

   Untargeted metabolomic analysis revealed significant abnormalities in
   amino acid metabolism in pancreatic cancer tissues. Pathway analysis
   identified several amino acid metabolic pathways (such as those
   involving glycine, serine, and aspartic acid) as being prominently
   altered in pancreatic cancer, with a particular emphasis on glutamate
   metabolism, which is involved in glutamine metabolism pathway.
   Metabolic reprogramming in pancreatic cancer cells, which enables
   adaptation to increased energy demands, includes enhanced glucose
   metabolism, glutamine metabolism, and fatty acid oxidation. These
   changes have been extensively documented [[128]44, [129]45].
   Dysregulation of amino acid metabolism in pancreatic cancer
   significantly influences cancer cell behavior, systemic metabolism, and
   various biological processes, including stemness, proliferation,
   invasion, migration, autophagy, and apoptosis. These metabolic
   alterations also modulate interactions with the tumor microenvironment,
   highlighting their potential as diagnostic, prognostic, and therapeutic
   targets [[130]19]. Previous studies have primarily focused on the
   metabolic reprogramming occurring within tumor cells. In contrast, the
   present study integrated metabolic alterations with the microbiome,
   emphasizing the potential role of microbes in mediating changes to the
   tumor microenvironment through metabolic regulation. This approach
   offers a novel perspective on the underlying pathological mechanisms.

   Microbiomic and metabolomic association analyses revealed significant
   correlations between Pseudomonas and the metabolism of specific amino
   acids. Our functional analysis results highlighted the glutamate
   metabolism pathway in PC, where glutamate participates in the synthesis
   and catabolism of glutamine. And the correlation analysis also
   underscored the significant association between Pseudomonas and
   alpha − ketoisovaleric acid (a metabolite derived from the
   BCAA-valine). PC cells utilize glutamine to support proliferation and
   redox balance [[131]46]. Glutamine and BCAA are crucial in regulating
   immune cell functions and tumor growth in PC [[132]44]. Amino acid
   metabolism, including glutamine, leucine (a branched-chain amino acid),
   glutamate, and asparagine, regulates the mammalian target of rapamycin
   (mTOR) signaling pathway in pancreatic cancer cells [[133]19, [134]44].
   The phosphatidylinositol 3-kinase/protein kinase B/mammalian target of
   rapamycin (PI3K/AKT/mTOR) signaling pathway is critical in the
   pathogenesis of PC, affecting both its initiation and progression
   [[135]47]. Glutamine supports pancreatic cancer growth through a
   KRAS-regulated metabolic pathway [[136]48]. Moreover, amino acid
   metabolism modulates autophagy in pancreatic cancer by acting both
   upstream and downstream [[137]19]. Kras-mutant PC cells require
   autophagy for their growth [[138]49, [139]50]. mTORC1 blocks autophagy,
   and glutamine metabolism is important for regulating this process
   [[140]19]. Therefore, we proposed the hypothesis between Pseudomonas
   and amino acid metabolism: our analysis suggests that Pseudomonas
   species within pancreatic tumors are associated with BCAA metabolism,
   potentially linked to mTOR signaling pathway activity. However, this
   association requires functional validation to establish causality. On
   the other hand, study have revealed that administration of Pseudomonas
   aeruginosa to the cancer patients induces autophagy [[141]51], and
   autophagy activation is associated with tumor cells becoming resistant
   to chemotherapy and radiation therapy [[142]52]. More importantly,
   Pseudomonas aeruginosa induces autophagy in human breast cancer cells
   through endoplasmic reticulum stress-mediated activation of the IRE1
   signaling pathway [[143]53]. We therefore propose the following
   hypothesis: our analysis indicates that Pseudomonas species within
   pancreatic tumors show potential association with IRE1 pathway-induced
   autophagy. However, functional validation (e.g., bacterial
   colonization/eradication experiments) is required to establish
   causality. However, the aforementioned mechanistic hypothesis requires
   further experimental validation. Collectively, these two hypotheses are
   hypothesized pathways based on literature and remain experimentally
   unverified. However, this microbial-metabolic interaction may represent
   a critical mechanism driving the progression of pancreatic cancer. In
   recent years, multi-omics analyses have become increasingly prominent
   in pancreatic cancer research, offering valuable insights into the
   systems biology mechanisms that drive tumor progression [[144]54].
   Despite existing studies on the characterization and role of the
   microbiome and metabolome in diseases such as inflammatory bowel
   disease (IBD) and uterine cancer [[145]55, [146]56], research on
   microbe-metabolism interactions in pancreatic cancer remains limited.
   This study was the first to establish a clear association between
   Pseudomonas and amino acid metabolism in pancreatic cancer, suggesting
   its potential role as a key regulator. These findings offer new
   insights into the systemic mechanisms underlying the pancreatic cancer
   microenvironment. This study aligns with previous research in
   emphasizing the central role of microenvironmental regulatory
   mechanisms and metabolic reprogramming in the progression of pancreatic
   cancer. However, its uniqueness lies in identifying Pseudomonas as a
   dominant intratumoral microbiota and elucidating the mechanism by which
   it influences pancreatic cancer through amino acid metabolic pathways.
   The study presentes a novel hypothesis and provides preliminary
   evidence, supported by integrated multi-omics analysis, to substantiate
   this hypothesis.

   From a clinical perspective, the results of our study offer several
   potential applications for the diagnosis and treatment of pancreatic
   cancer: (i) Discovery of potential diagnostic and therapeutic targets:
   The abundance of Pseudomonas and its regulation of amino acid
   metabolites (e.g., glutamine, and branched-chain amino acid
   metabolites) may serve as potential diagnostic and prognostic
   biomarkers of pancreatic cancer: ①Intratumoral microbial biomarkers:
   investigating the correlation of Pseudomonas abundance with early
   diagnosis and prognosis of PC. A significant difference in the
   abundance of microbiota was observed between adjacent tissues and
   pancreatic tumors and the intratumoral microbiota may serve as a
   biomarker for both early detection and prognosis prediction in PC
   [[147]15]. Early diagnosis of pancreatic cancer is crucial for
   improving patient survival rates. Intra-tumoral microbes can
   significantly increase the risk of early-onset cancers, including
   early-onset pancreatic cancer [[148]57]. Our study found that
   Pseudomonas was the dominant bacterial group in PC, and high abundance
   of Pseudomonas was associated with poor prognosis. Therefore,
   Pseudomonas may be a potential marker of diagnosis and prognosis for
   patients with PC. ②Biomarkers of amino acid metabolite: Investigating
   the spatial distribution and enrichment levels of glutamate, glutamine,
   branched-chain amino acid metabolism, and intermediate products (e.g.,
   α-ketoisovaleric acid) in pancreatic cancer tissues using spatial
   metabolomics imaging, and their association with early diagnosis and
   prognosis. The alterations of amino acids and their metabolism have
   great potential in PC early detection [[149]19, [150]58, [151]59]. Our
   study found that glutamate metabolism, glutamine metabolism and BCAA
   metabolism were significantly enriched in pancreatic cancer, and thus
   may serve as potential diagnostic and prognostic biomarkers.
   ③Microbiota-metabolite combined biomarker panel: The ratio of
   Pseudomonas abundance to branched-chain amino acid metabolites (e.g.,
   alpha-ketoisovaleric acid) could also be a promising direction for
   diagnosis and prognosis. More importantly, the abundance of Pseudomonas
   and its regulation of amino acid metabolites (e.g., glutamine, and
   branched-chain amino acid metabolites) can also be promising
   therapeutic targets: ① Direct targeting of Pseudomonas (e.g., using
   sensitive antibiotics such as amikacin): Intra-tumoral microbes can
   impact carcinogenesis and therapy responses via several mechanisms.
   Intra-tumoral bacteria may negatively impact anti-tumor immunity by
   promoting a tolerogenic immune microenvironment, characterized by the
   immunosuppressive activity of tumor-associated macrophages [[152]21].
   They may also induce chemoresistance by breaking down chemotherapeutic
   agents into inactive forms. For example, Gammaproteobacteria can
   metabolize the chemotherapeutic drug gemcitabine
   (2',2'-difluorodeoxycytidine) into its inactive form,
   2',2'-difluorodeoxyuridine, thereby contributing to gemcitabine
   resistance in colon cancer and pancreatic cancer [[153]40]. Moreover,
   the combined use of antibiotics has been shown to enhance the antitumor
   response to gemcitabine. Antibiotics can suppress pancreatic cancer
   progression by eradicating pro-tumorigenic intratumoral microbiota,
   thereby improving survival rates in patients with advanced disease
   [[154]60]. Therefore, targeting the microbiota in pancreatic tumors
   increases the effectiveness of immune therapy and chemotherapy
   [[155]15, [156]61]. Therefore, Pseudomonas may serve as a potential
   therapeutic target, with the potential to improve treatment outcomes by
   modulating its role within the tumor microenvironment.②Blocking the
   microbiota-metabolite-host axis (glutaminase inhibitors): Devimistat
   (CPI-613®), a novel mitochondrial metabolism inhibitor, disrupts the
   tricarboxylic acid (TCA) cycle by targeting two critical carbon entry
   points. Specifically, it inhibits the pyruvate dehydrogenase and
   α-ketoglutarate dehydrogenase complexes, thereby blocking the
   incorporation of glucose- and glutamine-derived carbon units into
   mitochondrial energy production. Furthermore, amino acid metabolism
   (e.g., glutamine, branched-chain amino acid-leucine and glutamate),
   regulates the mammalian target of rapamycin (mTOR) signaling pathway in
   pancreatic cancer cells [[157]19, [158]44]. Thus, inhibiting
   Pseudomonas-induced glutamate (glutamine) catabolism may
   synergistically enhance the efficacy of mTOR inhibitors (e.g.,
   everolimus) and chemotherapy regimens (e.g., gemcitabine or
   FOLFIRINOX). Therefore, intervention strategies targeting the
   Pseudomonas-amino acid metabolism axis could offer novel therapeutic
   approaches for the treatment of pancreatic cancer. In summary,
   Pseudomonas, glutamate, glutamine, and branched-chain amino acid
   metabolites (e.g., alpha-ketoisovaleric acid) may emerge as novel
   biomarkers and therapeutic targets for pancreatic cancer.
   (ii)Postoperative recurrence monitoring and prognostic assessment:
   Pseudomonas and its associated metabolites could serve as dynamic
   biomarkers for monitoring postoperative recurrence in pancreatic
   cancer. By tracking changes in microbial abundance and metabolite
   levels, the risk of recurrence may be detected earlier. This approach
   could complement traditional imaging-based monitoring, offering a more
   sensitive and timely follow-up strategy. (iii) Promoting the
   development of precision medicine: The heterogeneity in microbiota and
   metabolite profiles among pancreatic cancer patients provides a
   foundation for personalized treatment strategies. By analyzing the
   microbiome and metabolic profiles of individual patients, treatment
   responsiveness can be predicted, enabling the development of
   personalized interventions, such as microbial therapies or
   metabolism-modifying drugs. These findings not only enhance our
   understanding of the mechanisms underlying pancreatic cancer but also
   provide a foundation for the development of innovative diagnostic and
   therapeutic strategies targeting microbial and metabolic pathways. Such
   approaches hold the potential to significantly improve the prognosis
   and treatment outcomes for pancreatic cancer.

   Limitations of this study and strategies to address are as follow: (i)
   Sample size: The sample size in this study is relatively small,
   especially the reduction of samples for 16S sequencing, which may limit
   the ability to fully capture the diversity and heterogeneity of the
   microbiome and metabolome in pancreatic cancer patients. The
   implications of this reduction on the generalizability of results:
   ①Insufficient Representativeness: Given the substantial
   inter-individual variability in microbiomes, analyzing only 16 samples
   may not adequately reflect the microbiome characteristics of the entire
   study population. This reduction in sample size may lead to biased
   research outcomes, failing to accurately capture the comprehensive
   relationship between the microbiome and the study variables. ②Reduced
   Statistical Power: A smaller sample size may in result insufficient
   statistical power, thereby increasing the risk of Type I errors (false
   positives) and Type II errors (false negatives). This can make it
   difficult to detect biologically significant differences. Consequently,
   some potential microbiome features or biomarkers may be overlooked,
   thereby affecting the reliability of the study conclusions. ③Limited
   Extrapolation Capability: Conclusions drawn from a smaller sample may
   be restricted when extrapolated to or larger more complex populations.
   The composition and function of the microbiome are influenced by a
   variety of factors, such as diet, environment, and genetic background.
   For instance, the alpha diversity and abundance of Pseudomonas were
   likely not significantly influenced by tumor staging, based on our
   results. (ii) The absence of stage IV (M1) patients: this finding may
   be limited because all included samples were M0, and no stage IV (M1)
   cases were included. We acknowledge the limitation of lacking stage IV
   (M1) samples. Advanced-stage microbiome profiles likely diverge
   significantly from early/mid-stage patients. This sampling gap
   constrains comprehensive understanding of microbiome evolution
   throughout pancreatic cancer progression, thereby constraining clinical
   generalizability across disease stages. To address this issue, it has
   to explicitly states: ①conclusions are generalizable to resectable
   pancreatic cancer cohorts, with no direct extrapolation to stage IV or
   non-surgical candidates. ②Multicenter trials with expanded cohorts
   (incorporating stage IV patients) will evaluate population-wide
   generalizability and strengthen conclusion validity. Therefore, it is
   necessary to validate these findings in a larger cohort. Our
   explanations and proposed solutions are as follows: ①The heterogeneity
   observed within the sample cohort is driven by tumor microenvironment
   heterogeneity in pancreatic cancer [[159]62], variability in
   intratumoral microbiome distribution [[160]38], and the random
   selection of the 47 samples included. ②We acknowledge that the reduced
   sample size for 16S sequencing is relatively small, which may limit the
   generalizability of the results. However. To ensure the quality of the
   intergroup difference analysis (tumor vs. adjacent tissue), we employed
   strict selection criteria to exclude samples with high dispersion in a
   paired manner. This approach ensured a relatively ideal consistency
   within the samples of each group, thereby enhancing the reliability of
   the intergroup comparison results from a statistical and methodological
   rigor. ③The validity of using small sample cohorts in exploratory
   research is well-recognized(sample size varying from 12 to 20 case)
   [[161]20, [162]21, [163]63–[164]65]. And small sample cohorts in
   exploratory research are beneficial as they provide rapid preliminary
   results that can guide future large-scale studies. ④Validation in a
   larger, more diverse cohort (To enhance the generalizability of our
   research findings, we plan to expand the sample size to better capture
   the diversity of the microbiome in future study. We intend to initiate
   a multicenter study cohort, incorporating pancreatic cancer patients
   from five high-volume pancreatic centers across three regions in China
   in subsequent studies.) Furthermore, integrating and analyzing existing
   large-scale microbiome datasets (e.g., TCMbio) may help identify more
   robust microbiome features. These will include patients with different
   stages (I- IV) and pathological subtypes of PC, as well as healthy
   controls, would strengthen comparisons and provide more robust
   validation. In summary, although a reduced sample size may have certain
   impacts on the generalizability of the research findings, expanding the
   sample size and conducting validation studies can significantly enhance
   the reliability and applicability of the research conclusions. (iii)
   Bias in microbiome data: Microbiome analysis based on 16S rDNA
   sequencing may introduce biases. This technique does not
   comprehensively characterize all bacterial species, and non-bacterial
   components of the microbiome, such as Fungi and viruses, are often
   overlooked. Additionally, 16S rDNA sequencing has Limitations in
   identifying species at the strain level. However, the primary objective
   was to identify bacterial community structure shifts associated with
   host metabolic phenotypes. 16S sequencing remains the gold standard for
   such species and strain-level microbiome analysis, with proven
   reproducibility across large cohorts [[165]66]. Given the need for
   paired multi-omics data (metabolomics) from clinical samples, 16S
   sequencing provided cost-effective bacterial profiling while preserving
   budget for functional assays. This study also employed metabolomic
   sequencing and analysis. We correlated PICRUSt2-predicted metabolic
   functions (based on computer calculations) from 16S sequencing data
   with metabolomic data, including pathways such as arginine and proline
   metabolism, arginine biosynthesis, and glutamate metabolism. This
   approach validated the reliability of our conclusions. Therefore,
   although PICRUSt2 functional prediction has inherent biases, our study
   can enhance the reliability and scientific rigor of the research
   results to some extent by cautiously interpreting the results,
   combining with other methods (such as metabolomics), and keeping track
   of database updates. While less precise than metagenomics, this
   approach predicts the metabolic functions of the microbial community
   and correlates them with relevant modules in metabolomics. However,
   complementary approaches, such as metagenomics or metatranscriptomics,
   and the inclusion of other omics technologies (e.g., proteomics), can
   provide comprehensive genomic information about microbial communities,
   enabling strain-level identification and functional gene analysis and
   reveal dynamic expression profiles of these communities
   [[166]67–[167]69]. These new technologies could explore the structure
   and function of intratumoral microbiota more thoroughly and
   comprehensively in pancreatic cancer [[168]69]. Therefore, we will
   validate our findings by metagenomics or metatranscriptomics in a
   larger cohort in subsequent research. (iv) Deficiencies in causality:
   Current data indicate potential associations between intra-tumoral
   Pseudomonas species and amino acid (BCAA) metabolic reprogramming.
   However, mechanistic inferences tracing microbiota-metabolism-signaling
   cascades (specifically Pseudomonas → IRE1 → autophagy and
   Pseudomonas → BCAA → mTOR pathways) remain hypothetical statements,
   requiring causal validation through future experiments investigating
   these pathways' regulatory roles in pancreatic cancer development. Such
   causal validation will employ in vivo models (animal models) or
   pancreatic cancer organoid/Pseudomonas co-cultures with independent
   BCAA metabolism and IRE1 signaling interventions. Although a
   significant correlation between Pseudomonas and amino acid metabolism
   was identified, the causal relationship remains unclear. Further in
   vitro and in vivo experiments, including animal models, are necessary
   to validate whether Pseudomonas influences pancreatic cancer
   progression through amino acid metabolism. The proposed study will
   utilize germ-free KPC mice to investigate the changes in metabolic
   profiles following colonization with Pseudomonas (n = 10 per group).
   The experimental design was divided into four groups, namely
   Experimental Group 1 (Colonization with Pseudomonas, clinically
   isolated), Control Groups 1 (Colonization with non-pathogenic
   Escherichia coli, negative control), Control Group 2(Germ-free control
   group) and Control Group 3(Colonization with Pseudomonas + treatment
   with a sensitive antibiotic: amikacin). Then metabolite detection at
   multiple time points within 2 weeks (e.g., 1 day, 3 days, 7 days) to
   elucidate the distribution and changes of amino acid metabolites by
   spatial metabolite imaging. By comparing the metabolomic data between
   the colonization group and the control groups, the causal effects of
   Pseudomonas on host metabolism can be directly assessed. Additionally,
   advanced tools such as CRISPR could be used to investigate the
   functional role of these metabolic pathways within the intratumoral
   microbiota. CRISPR knockout strains of key amino acid metabolism genes
   in Pseudomonas (e.g., genes encoding key enzymes in branched-chain
   amino acid metabolism or glutamine metabolism) will be constructed and
   compared with wild-type (WT) strains. These strains will be inoculated
   into germ-free C57BL/6 mice (n = 10 per group) to assess changes in
   amino acid metabolism levels by spatial metabolite imaging between the
   knockout and control groups. (v) Hypothesis-Verification Gap. This
   study proposes two Pseudomonas-focused research hypotheses with
   experimental verification gaps: (1) BCAA-mTOR Axis Hypothesis:
   Pseudomonas species may activate mTOR signaling by regulating BCAA
   metabolites within the tumor microenvironment. Supported by: ① observed
   correlations between Pseudomonas and BCAA metabolites; Verification
   gaps: ① Pseudomonas colonization/eradication functional validation; ②
   BCAA metabolic flux tracing; ③ mTOR pathway activity modulation assays.
   (2) IRE1-Autophagy Axis Hypothesis: Pseudomonas species may trigger
   endoplasmic reticulum stress → IRE1 activation → autophagy. Supported
   by: ①literature-documented interactions between microbiota (P.
   aeruginosa), endoplasmic reticulum stress, IRE1, and autophagy.
   Verification gaps: ① Pseudomonas colonization/eradication studies; ②
   IRE1 inhibitor interventions; ③ functional assays (e.g., TEM autophagy
   flux quantification). This work advances novel hypotheses regarding
   Pseudomonas-mediated regulatory mechanisms in pancreatic cancer
   progression, necessitating validation through follow-up studies such as
   (but not limited to) pancreatic cancer organoid/Pseudomonas co-cultures
   and germ-free murine models. (vi) Biases introduced by sample handling
   or experimental variability and corresponding processing strategies:
   ①Sample handling biases include sample collection and preservation (the
   timing, location, and methods of sample collection, as well as the
   conditions of sample preservation, can influence the composition of
   microbial communities and metabolites), DNA or metabolite extraction
   (different methods for DNA or metabolite extraction can lead to
   variations in extraction efficiency and quality, thereby affecting the
   results of 16S sequencing and metabolomics analysis) and sample
   contamination (contamination during sample collection, storage, and
   extraction of DNA can interfere with the analysis of target
   intratumoral microbiota).②Experimental variation biases include PCR
   amplification (differences in PCR primers, reaction conditions, or
   operators can lead to variations in amplification efficiency, affecting
   the accuracy of 16S sequencing), sequencing platforms (different
   sequencing platforms or batches can result in variations in data
   quality, impacting the analysis of microbial communities), metabolomics
   analysis (different instruments, analytical methods, or operators can
   cause systematic biases in metabolite detection, affecting the results
   of metabolomics analysis) and data processing (different data
   processing methods or analytical software can lead to differences in
   results, affecting the final biological interpretation). ③Specific
   strategies to reduce biases about sample handling include standardized
   operating procedures (develop and strictly adhere to standardized
   protocols for sample collection and preservation), standardized sample
   reservation conditions (samples are processed or stored under
   appropriate conditions to prevent changes in the composition of
   microbial communities and metabolites), and validated DNA and
   metabolite extraction methods (select validated methods and use the
   same kits and operators in experiments to reduce differences in
   extraction efficiency). ④ Addressing experimental variation, we can
   minimize the possibility of batch effects in the experimental design
   phase, regularly calibrate instruments to ensure consistent
   performance, and standardize and homogenize training for operators to
   ensure consistent experimental operations. For data processing and
   analysis, we can implement rigorous quality control during data
   processing and analysis to exclude low-quality data, use appropriate
   statistical methods (e.g., the ComBat algorithm) to detect and correct
   batch effects before differential analysis, and apply bioinformatics
   tools (e.g., Decontam) to filter out contaminants [[169]70]. In
   summary, these limitations highlight areas for improvement in future
   studies. Expanding the sample size (including all clinical stage),
   integrating multi-omics data, conducting mechanistic validation
   experiments, and addressing sample handling and experimental bias will
   enhance both the scientific understanding and clinical applicability of
   these findings.

   The following directions deserve further exploration to deepen the
   understanding of the pancreatic cancer microenvironment and to
   facilitate the advancement of basic research to clinical translation:
   (i) Validation of the causal relationship between Pseudomonas and amino
   acid metabolism. To verify whether Pseudomonas directly regulates
   pancreatic cancer development and progression through amino acid
   metabolism, animal models or organoid models should be utilized. This
   would clarify the causal relationship between Pseudomonas and key
   metabolites (e.g., α-ketoisovaleric acid), providing a foundation for
   the subsequent development of therapeutic targets. (ii) Association
   between Pseudomonas and the efficacy of anticancer drugs. It is crucial
   to explore whether Pseudomonas and its metabolic activities influence
   the sensitivity of pancreatic cancer patients to conventional
   chemotherapies (e.g., gemcitabine or FOLFIRINOX) or targeted therapies.
   Understanding this relationship could inform the development of
   personalized treatment regimens and potentially reduce therapeutic
   resistance. (iii) Metabolite target development and validation.
   Developing inhibitors or mimetics targeting key metabolites (e.g.,
   α-ketoisovaleric acid) and investigating their therapeutic potential
   would offer valuable insights for drug development based on metabolic
   targets. This could open new avenues for the treatment of pancreatic
   cancer by modulating specific metabolic pathways. (iv) Mechanisms of
   the bacterial-metabolite-immunity axis. Future studies should focus on
   understanding how Pseudomonas and its associated metabolites contribute
   to pancreatic cancer progression by regulating the tumor immune
   microenvironment, including immunosuppressive cells and inflammatory
   factors. Identifying key nodes in the microbial-metabolic-immune
   network could lead to novel strategies for immunotherapy. (v) Promoting
   multidisciplinary integration in pancreatic cancer research.
   Integrating microbiomics, metabolomics, and cancer biology provides a
   comprehensive approach to studying the complex microenvironment of
   pancreatic cancer. This multidimensional analysis enhances our
   understanding of tumor biology and facilitate the development of more
   effective diagnostic and therapeutic strategies. (vi) Validation with a
   larger cohort is needed. We plan to initiate multicenter cooperation to
   build a multicenter cohort of pancreatic cancer patients, incorporating
   more samples to validate our findings, and enhance the generalizability
   of our research results. The multicenter cohort design is as follows:
   ①Cohort 1 (Retrospective Validation): Integrate microbiome, metabolome,
   and clinical data related to pancreatic cancer from three high-volume
   pancreatic centers of three large tertiary-A level hospitals affiliated
   with our university (n = 300) to verify the consistency of the research
   findings.② Cohort 2 (Prospective Multicenter Cohort): Collaborate with
   five large tertiary-A level hospitals in three regions in China to
   establish a multicenter pancreatic cancer cohort (n = 500) to assess
   the generalizability of the study's conclusions. Furthermore, we plan
   to collaborate with two existing partner-pancreatic cancer research
   institutions to share data and conduct integrated analyses to further
   validate our findings.

Conclusions

   Collectively, this study provides preliminary evidence of an
   interaction between Pseudomonas and amino acids in the pancreatic
   cancer microenvironment, suggesting a potential association with its
   regulation. It proposes a novel theoretical framework for the
   "microbe-metabolism-cancer" mechanism, which could enhance our
   understanding of pancreatic cancer pathogenesis. However, the causal
   relationship and underlying mechanisms remain unexplored experimentally
   and require further validation. These findings highlight future
   research possibilities for exploring diagnostic and therapeutic
   implications in pancreatic cancer, but they do not establish direct
   applications for early diagnosis, personalized treatment, or prognosis
   improvement.

Supplementary Information

   [170]12943_2025_2458_MOESM1_ESM.pdf^ (577.8KB, pdf)

   Supplementary Material 1. Supplementary Figure 1 Spearman correlation
   coefficient matrix heatmap of significantly different microbial
   communities and significantly different metabolites. Note: This matrix
   heatmap displayed the correlation between significantly different
   microbial communities and significantly different metabolites, as well
   as the correlation among significantly different metabolites and among
   significantly different microbial communities. The heatmap is divided
   into four quadrants by the blue dashed line (┄┄┄). The upper-left
   quadrant displays the correlation among significantly different
   microbial communities, while the lower-right quadrant shows the
   correlation among significantly different metabolites. The upper-right
   and lower-left quadrants display the correlation between significantly
   different microbial communities and significantly different
   metabolites, which are mirror-symmetrical. The Spearman correlation
   coefficient value (r) ranges from -1 to +1. The correlation coefficient
   (r) is represented by color, with an r value greater than 0 indicating
   a positive correlation (represented by red). The deeper the color, the
   stronger the correlation.
   [171]12943_2025_2458_MOESM2_ESM.pdf^ (145.9KB, pdf)

   Supplementary Material 2. Supplementary Figure 2 Correlation analysis
   between alpha diversity and stages (I-III). (A) Raincloud plots of the
   correlation analysis between ACE index and clinical staging, (B)
   Raincloud plots of the correlation analysis between Chao1 index and
   clinical staging, (C) Raincloud plots of the correlation analysis
   between Shannon index and clinical staging, (D) Raincloud plots of the
   correlation analysis between Simpson index and clinical staging.
   [172]12943_2025_2458_MOESM3_ESM.pdf^ (121.2KB, pdf)

   Supplementary Material 3. Supplementary Figure 3 Correlation analysis
   between relative abundance of Pseudomonas and clinical stages.(A)
   Raincloud plots of the correlation analysis between relative abundance
   of Pseudomonas and T stage (T1-3), (B) Raincloud plots of the
   correlation analysis between relative abundance of Pseudomonas and N
   stage (N0-2), (C) Raincloud plots of the correlation analysis between
   relative abundance of Pseudomonas and clinical stage(I-III).
   [173]Supplementary Material 4.^ (24.4KB, docx)
   [174]Supplementary Material 5.^ (20.8KB, docx)
   [175]Supplementary Material 6.^ (29.2KB, docx)
   [176]Supplementary Material 7.^ (29.9KB, docx)

Acknowledgements