Abstract

Background

   Molecular events that drive endometriosis (EM) and cause accompanying
   immune deregulation remain elusive. Our purpose was to identify key
   pathways involved in lesion formation across diverse populations and to
   detect transcriptomic changes in eutopic endometrium that accompany EM.

Methods

   We searched Gene Expression Omnibus and ArrayExpress and performed
   differential gene expression analysis and a network meta-analysis on
   nine qualifying datasets. Those contained transcriptomic data on 114
   ectopic endometrium samples (EL), 138 eutopic endometrium samples from
   women with endometriosis (EEM), and 79 eutopic endometrium samples from
   women without endometriosis (EH). Gene ontology and enrichment analysis
   were performed in DAVID, Metascape, and Cytoscape, and drug repurposing
   was done in CMap.

Results

   EEM compared to EH upregulated CCL21 and downregulated BIRC3, CEL, and
   LEFTY1 genes (|log[2]FC| > 0.5, p < 0.05). EL showed increased
   expression of complement and serpin genes (EL vs. EEM: C7, logFC =
   3.38, p < 0.0001; C3, logFC = 2.40, p < 0.0001; SERPINE1, logFC = 1.02,
   p < 0.05; SERPINE2, logFC = 1.54, p < 0.001) and mast cell markers (EL
   vs. EEM: CPA3, logFC = 1.54, p < 0.0001; KIT, logFC = 0.74, p < 0.001).
   Functional enrichment analysis highlighted complement and coagulation,
   inflammation, angiogenesis, and extracellular matrix remodeling as
   drivers of endometriosis. Pharmacogenomic analysis indicated Janus
   kinase (JAK), cyclin-dependent kinase (CDK), and topoisomerase
   inhibitors as therapy targets.

Conclusion

   Our results suggest an interplay between complement and coagulation,
   mast cells, extracellular matrix remodeling, and the JAK/STAT3 pathway
   in endometriosis. We underscore the significance of complement C3 and
   propose JAK inhibitors as therapy candidates. Detected expression
   differences between EEM and EH are important for the development of
   diagnosis via endometrial biopsy.

   Keywords: endometriosis, eutopic and ectopic endometrium, network
   meta-analysis, complement and coagulation, mast cells, Janus kinase
   (JAK) inhibitors

1. Introduction

   Immune system deregulation is a well-accepted phenomenon in
   endometriosis, and various inflammatory phenotypes have been associated
   with increased risk of this condition ([33]1–[34]3). To date, pathways
   contributing to the immune imbalance remain elusive. Lack of knowledge
   on the key processes that drive endometriosis hinders its early
   detection and therapy development. There is a need to define those
   molecular events and to understand how they interact to foster
   peritoneal inflammation.

   Endometriosis is a chronic and complex disease currently showing a
   median diagnostic delay of 7–9 years ([35]4–[36]7). There has been
   significant progress in the development of endometriosis imaging
   protocols ([37]8); however, laparoscopy remains a gold standard for
   final diagnosis. There is a need to explore the less invasive
   endometrial biopsy option. To consider this strategy, in-depth
   knowledge of the molecular differences in the eutopic endometrium of
   healthy controls and women with endometriosis is needed.

   There has been significant progress in the development of endometriosis
   imaging protocols ([38]8); however, laparoscopy remains a gold standard
   for final diagnosis. There is a need to explore the less invasive
   endometrial biopsy option. To consider this strategy, in-depth
   knowledge of the molecular differences in the eutopic endometrium of
   healthy controls and women with endometriosis needs to be obtained.
   Recent evidence shows the potential of stromal cells from menstrual
   discharge of women with endometriosis to initiate lesion growth in
   mice, thus highlighting the role of endometrial seeding ([39]9). It was
   shown that eutopic endometria of women with and without endometriosis
   displayed different profiles of infiltrating immune cells. Ectopic
   endometrium tissue from women with endometriosis showed increased
   amounts of CD8+ T cells and CD56+ NK cells but decreased numbers of
   CD163+ macrophages, and those findings were correlated with increased
   risk of infertility ([40]10). A detailed understanding of the
   alterations in eutopic endometrium that occur during endometriosis
   development will further shed light on the process of endometrial
   seeding and contribute to the development of endometrial biopsy as a
   diagnostic tool.

   Several attempts have been made at delineating disease biomarkers, but
   to date this has not yet proven successful. Various omics technologies
   enabled identification of key genes related to the pathophysiology of
   endometriosis. However, a consensus has not yet been reached, and we
   are still missing the focal points on which to concentrate the
   therapeutic endeavors. A multi-cohort analysis is needed to address the
   issue in an unbiased and comprehensive manner.

   In this article, we aimed to better understand complex events that
   underlie endometriotic lesion formation and progression. To achieve
   this, we systematically reviewed endometriosis data and performed
   network meta-analysis on chosen datasets. We generated a transcriptomic
   profile of endometriosis, determined the key pathways involved in
   lesion formation, and explored possible drug candidates for
   endometriosis therapy.

2. Materials and methods

2.1. Search strategy and study selection

   The dataset search was conducted in public repositories Gene Expression
   Omnibus (GEO) and ArrayExpress to ensure that no relevant studies were
   missed. A systematic approach was employed to assess the risk of bias
   throughout both the dataset selection process and subsequent analyses.

   Datasets were retrieved from the GEO using search terms “endometriosis”
   and “Homo sapiens” and filtered with terms “expression profiling by
   array” or “expression profiling by high-throughput sequencing.” MEDLINE
   and PubMed were searched for publications that correspond to those
   publicly deposited datasets.

   The studies were identified using the following PICOS principle:
   Patients = patients with or without endometriosis, Intervention = bulk
   RNA performed on excised tissue, Comparison = dividing patients into
   those with endometriosis or without based on laparoscopic findings,
   Outcome = differential gene expression, and Study design =
   transcriptomic bulk RNA studies. Studies included in the analysis had
   to contain at least two tissues of interest: ectopic
   endometrium—endometrial lesion (EL), eutopic endometrium from women
   without endometriosis (EH), or eutopic endometrium from women with
   endometriosis (EEM).

   Inclusion criteria were predefined and stringent to minimize the risk
   of bias, focusing on datasets using RNA sequencing (RNA-seq) or
   microarray technologies with available raw data. Transcriptomic
   analysis had to be performed directly on human endometrial tissue that
   had not been subjected to any manipulation or cell isolation prior to
   RNA extraction. Samples had to be taken from patients not on hormonal
   treatment in the 3 months preceding tissue collection. The presence or
   absence of endometriosis had to be confirmed with laparoscopy for
   samples to be included in our study. Only datasets with accompanying
   publications were considered to ensure all information about samples
   was available. Datasets with incomplete information were excluded to
   reduce variability and minimize errors. The database selection
   flowchart and a full list of inclusion/exclusion criteria are
   summarized in [41]Figure 1 and [42]Table 1 . Two independent reviewers
   screened datasets for relevance, and any discrepancies were resolved in
   discussion with a third reviewer. The study protocol was registered in
   PROSPERO (ID CRD42024548098).

Figure 1.

   [43]Flowchart illustrating the dataset selection process. Initially,
   122 databases were identified through GEO searching. Eighteen
   duplicates were removed, leaving 104 for screening. Twelve were
   excluded for lacking abstracts or articles, leaving 92 original
   datasets. Further exclusions were made: animal datasets (7), lack of
   endometriosis group (18), only adenomyosis (7), no control group (6),
   single cell (6), no raw data (5), no mRNA (4), only cells/organoids
   (13), and serum (1). Therefore, 9 datasets were assessed for
   eligibility.
   [44]Open in a new tab

   Flow diagram showing the selection process of databases retrieved from
   the Gene Expression Omnibus search for transcriptomic data comparing
   eutopic and ectopic endometrial tissue. All datasets from ArrayExpress
   were also deposited in Gene Expression Omnibus thus they were not
   further considered in the selection process. Created in
   [45]https://BioRender.com.

Table 1.

   Inclusion and exclusion criteria for selection of GEO datasets.
   Inclusion criteria
   Transcriptomic datasets containing endometriosis and a control group
   Accompanied by the corresponding publication
   The presence or absence of endometriosis confirmed via laparoscopic
   procedure
   Patients not undergoing hormonal treatment three months prior to tissue
   collection
   Transcriptomic analysis performed directly on endometrial human tissue
   Exclusion criteria
   Tissue processing post excision (for further cell isolation or single
   cell analysis)
   Adenomyosis included as endometriosis group
   ScRNA-seq analysis
   Lack of whole genome mRNA sequencing and or profiling
   Sequencing or profiling performed on cells or organoids isolated from
   tissue
   Lack of raw data deposited in GEO
   [46]Open in a new tab

2.2. Data extraction and differential gene expression

   Each dataset was analyzed individually to ensure that data-specific
   preprocessing and normalization steps were applied appropriately. For
   microarray datasets, the raw data files were retrieved from the GEO
   repository using the R package “GEOquery” ([47]11). The preprocessing
   of microarray data was conducted following the manufacturer’s
   protocols. Background correction and quantile normalization were
   applied for all array data. To adjust for differences in library size
   and transcript length, the raw read counts from RNA-seq datasets were
   normalized and scaled using the average transcript length for each
   sample. Following this, library size normalization was performed using
   the trimmed mean of M-values (TMM). After preprocessing, group
   comparisons were conducted on the normalized datasets to identify
   differentially expressed genes (DEGs) between experimental groups. For
   this purpose, the “limma” package in R was utilized. In the
   meta-analysis, we used logFC values and standard errors estimated with
   the limma package to calculate combined effect sizes. The estimation of
   standard errors incorporates empirical Bayes moderation, which provides
   more stable variance estimates compared to standard t-tests. The
   analysis was exploratory in nature, with the primary goal of
   identifying potentially differentially expressed genes that could be
   candidates for further investigation. Therefore, at this stage, no
   correction for multiple testing was applied. The results from this step
   served as a basis for subsequent functional analyses (pathway and
   pharmacogenomics analysis), in which appropriate methods accounting for
   multiple testing correction were employed. Additionally, the expression
   of significantly differentially expressed genes (defined as logFC >
   |0.5| and p-value < 0.05) showed consistent patterns across the
   individual datasets.

2.3. Network meta-analysis

   Network meta-analysis on gene expression was performed using the
   “netmeta” package ([48]12). This approach allows for the integration of
   information from multiple comparisons, even when direct comparisons
   between all datasets are not available, providing a comprehensive and
   robust analysis of the data. In the context of this study, network
   meta-analysis was used to compare gene expression levels across
   datasets systematically. Although meta-analysis allows for the
   determination of both direct and indirect effects, in our subsequent
   analyses, we focused on the combined effect to maximize the quality of
   the analyzed data and reduce the influence of less reliable direct or
   indirect effects. For investigated difference measurement, we used
   logFC and its corresponding standard error. These were interpreted as
   the mean difference and the standard error of the mean difference,
   respectively, which are widely used metrics in comparative gene
   expression studies. This standardization ensures that the results are
   both interpretable and comparable across datasets.

   We performed 7,664 network meta-analysis for genes that occurred in
   each of the datasets included in the study. Genes with a p-value < 0.05
   were considered statistically significant, and a |logFC| > 0.5 was used
   to filter genes with biologically meaningful changes in expression.
   This relatively low threshold was chosen to avoid missing potentially
   important differences. Choosing |logFC| > 0.5 may potentially result in
   the inclusion of genes that were influenced by random or systematic
   errors and biopsy quality if only one study was taken into
   consideration. Here, the biological significance of obtained
   differentially expressed genes is validated by the comparison across
   nine different datasets.

2.4. Risk of bias

   To reduce the risk of bias, we included studies with raw data deposited
   and results published in peer-reviewed journals. Information from the
   accompanying publication was used to ascertain the quality of the study
   and to identify if the absence of endometriosis was properly determined
   and to confirm that tissue did not undergo any manipulation prior to
   RNA isolation. Heterogeneity was evaluated using I² statistics and
   Cochran’s Q-test, while sensitivity analyses validated the robustness
   of findings. Funnel plots were generated to assess bias. These measures
   ensured a thorough evaluation of potential biases, enhancing the
   reliability and validity of the meta-analytic findings. I² statistics
   and Cochran’s Q-test values are included in the [49]Supplementary
   Dataset .

2.5. Gene ontology and pathway analysis

   The list of DEGs obtained from the network meta-analysis was submitted
   to DAVID ([50]13) for gene ontology and KEGG and Reactome pathways
   analysis. For functional clustering, we applied a cutoff enrichment
   score of >2.5, p < 0.05, and medium classification stringency. The same
   list of DEGs was analyzed in Metascape ([51]14) v3.5.2024.0101, and the
   most enriched terms were visualized in Cytoscape ([52]15) v3.10.2.

2.6. Computational pharmacogenomics

   To identify pharmacological compounds likely to reverse the
   endometriosis gene signature, we queried the drug repurposing reference
   database, CMap ([53]16). We submitted a list of 150 upregulated and 150
   downregulated genes that had the highest combined logFC for EL versus
   EH and EL versus EEM comparisons and p < 0.05. The CMap analysis
   reports a median tau score, which represents connectivity strength
   between the submitted list of DEGs and thousands of compounds and
   perturbagens tested on human cell lines. A median tau score of above 90
   or below −90 is considered to represent a high connectivity.

3. Results

3.1. Characteristics of chosen studies

   Nine datasets met the inclusion criteria ([54] Table 1 ; [55]Figure 1 )
   and were included in the analysis ([56] Table 2 ). Those contained
   transcriptomic data on 114 ectopic endometrium samples (EL), 138
   eutopic endometrium samples from women with endometriosis (EEM), and 79
   eutopic endometrium samples from women without endometriosis (EH). The
   absence of endometriosis in the healthy (EH) group had to be confirmed
   during the laparoscopic procedure. EEM and EL samples were either
   obtained from the same person (in studies: [57]GSE25628, [58]GSE37837,
   and [59]GSE7305) or from different individuals (in study
   [60]GSE141549). Tissues were collected on three different continents
   and encompassed all types and stages of endometriosis (clinical data in
   [61]Table 3 ).

Table 2.

   Characteristics of GEO datasets chosen for the network meta-analysis.
   GEO accession number Method Number of detected genes Ectopic
   endometrium (EL, n = 114) Eutopic endometrium from patients with
   endometriosis (EEM, n = 138) Eutopic endometrium from patients without
   endometriosis/ healthy control (EH, n = 79)
   [62]GSE232713 ([63]90) High-
   throughput
   sequencing 17488 – 7 7
   [64]GSE153740 &
   [65]GSE153739 ([66]91) High-
   throughput
   sequencing 16840 &
   17449 – 4 & 4 4 & 3
   [67]GSE141549 ([68]92) Microarrays 19746 79 49 21
   [69]GSE134056 ([70]93) High-
   throughput
   sequencing 18885 – 16 22
   [71]GSE25628 ([72]94) Microarrays 12644 7 9 6
   [73]GSE37837 ([74]95) Microarrays 21094 18 18 –
   [75]GSE6364 ([76]96) Microarrays 20857 – 21 16
   [77]GSE7305 ([78]97) Microarrays 20857 10 10 –
   [79]Open in a new tab

Table 3.

   Clinical characteristics of patients per dataset.
   GEO accession number Author Age: mean (range) Type of endometriosis
   Stage of endometriosis Cycle phase Fertility Ethnicity Collection site
   [80]GSE232713 Hiao-Yan Li 33.93
   (28–38) Not provided Not provided Secretory (14) Infertile (14) Not
   provided China
   [81]GSE153739 &
   [82]GSE153740 *** Kashmira Bane 31.70
   (Not provided) OE (4)
   PE (3)
   OE, PE (1) II*
   III*
   IV* Secretory (8)
   Proliferative (7) Fertile*
   Infertile* Not provided India
   [83]GSE141549 Michael Gabriel 33.43
   (21–48) OE (15)
   PE (32)
   DE (17)
   RE (6)
   SLE (7) I (18)
   II (19)
   III (21)
   IV (61)
   Missing (2) Proliferative (53)
   Secretory (81)
   Menstrual (14)
   Missing (1) Not provided Not provided Finland
   [84]GSE134056 Sadia Akter Not provided
   (18–49) Not provided Not provided Not provided Not provided Not
   provided USA
   [85]GSE25628
   ** Stefania Crispi 31.5
   (22–42) Not provided II (2)
   IV (6) Proliferative (8) Not provided Not provided Italy
   [86]GSE37837 Meraj A Khan 33.17
   (24–40) OE (18) III (8)
   IV (10) Secretory (5)
   Proliferative (13) Fertile (18) Not provided India
   [87]GSE6364 Burney RO 36.23
   (22–50) OE, PE (7), PE (2), OE, PE, RE (5)
   RE, PE (7) III*
   IV* Secretory (26)
   Proliferative (11) Fertile*
   Infertile* Caucasian (26), Asian (4),
   Black (3), Asian, Indian (1)
   Unknown (3) USA
   [88]GSE7305 Aniko Hever Not provided OE (10) Not provided Secretory (6)
   Proliferative (2)
   Unknown (2) Not provided Caucasian (10) USA
   [89]Open in a new tab

   Abbreviations in the “Type of endometriosis” columns denote ovarian
   endometriosis, (OE); peritoneal endometriosis, (PE); Deep infiltrating
   endometriosis (DIE); rectovaginal endometriosis, (RE); sacrouterine
   ligament endometriosis, (SLE). *No specific information regarding
   number of patients. **No clinical data for healthy patients (n = 6);
   clinical data available for patients with endometriosis (n = 8); for
   each affected woman, two biopsies were obtained for the eutopic and
   ectopic endometrium, respectively, while only one biopsy from the
   healthy eutopic endometrium was collected. ***For [90]GSE153739 and
   [91]GSE153740, data of 15 samples is deposited in GEO while clinical
   data in the corresponding article is provided for all 66 patients.

3.2. Transcriptomic profile of eutopic and ectopic endometrium.

   Differential expression analysis was performed for each of the three
   comparisons: EL versus EEM, EL versus EH, and EEM versus EH. Using p <
   0.05 and |logFC| > 0.5, we identified 1,109 DEGs between EL and EEM,
   1,267 DEGs between EL and EH, and four DEGs between EEM and EH ([92]
   Figure 2 ). The heatmap of the top 40 upregulated and downregulated
   genes for all comparisons per dataset is presented in [93]Figure 2D .
   The full list of network meta-analysis results is deposited in
   [94]Supplementary Dataset .

Figure 2.

   [95]Three volcano plots (A, B, C) display gene expression data with
   color coding for significance and fold change. A heatmap (D) shows
   hierarchical clustering of gene expression in red and blue gradients. A
   Venn diagram (E) visualizes overlapping and distinct gene counts across
   three comparisons (EL vs EEM, EL vs EH, EEM vs EH).
   [96]Open in a new tab

   Differentially expressed genes identified by network meta-analysis.
   Volcano plots showing differentially expressed genes for the following
   comparisons (A) endometriotic lesions versus endometrium from women
   with endometriosis, (B) endometriotic lesions versus endometrium from
   women without endometriosis, (C) endometrium from women with and
   without endometriosis. For graphs A to C, red points signify genes with
   logFC less than −0.5 or more than 0.5 and p-value less than 0.05, blue
   points signifiy genes with logFC belonging to −0.5 to 0.5 range and
   p-value less than 0.05, green points signify genes with logFC less than
   −0.5 or more than 0.5 and p-value more than 0.05 and gray dots signify
   genes with with logFC belonging to −0.5 to 0.5 range and p-value more
   than 0.05. Genes with p-values < 1 × 10^−6 and log2FC>|2| (A, B) and
   p-values < 0.05 and log2FC>|0.5| (C) are labeled. Heatmap with top 40
   most differentially expressed genes per comparison per dataset (D).
   Expression pattern of differentially expressed genes per comparison
   type (E). EL, endometrial lesion; EEM, eutopic endometrium from women
   with endometriosis; EH, eutopic endometrium from women without
   endometriosis.

   The network meta-analysis revealed that the transcriptomic profile of
   lesions was profoundly different from that of eutopic endometrium ([97]
   Figures 2A, B ), while the eutopic endometrium from women with (EEM)
   and without endometriosis (EH) differed in the expression of four genes
   only ([98] Figure 2C ). BIRC3, CEL, and LEFTY1 were significantly less
   expressed in the endometrium of women with endometriosis than without
   (logFC = −0.79, p = 0.0051; logFC = −0.52, p = 0.0051; logFC = −0.61, p
   = 0.0099, respectively). CCL21 was significantly higher in EEM versus
   EH (logFC = 0.59, p = 0.0255) and even higher when EL with EEM was
   contrasted (logFC = 1.57, p < 0.0001, [99]Figure 2C ). C-C motif
   chemokine ligand 21 (CCL21) is an inflammatory mediator associated with
   moderate to severe endometriosis ([100]17); however, to-date, its use
   as a disease biomarker has failed ([101]18). Our results showed a
   directional increase of CCL21 from the endometrium of healthy patients
   through that of endometriosis sufferers to lesions themselves,
   indicating its role in the eutopic endometrium inflammation in patients
   with endometriosis.

3.3. Pathways contributing to lesion development

   In further analysis, we selected genes that showed differential
   expression in both EL versus EEM and EL versus EH comparisons (the
   intersection of the sets, [102]Figure 2E ) and exhibited the same
   direction of expression. For p < 0.05 and |logFC| > 0.5, we obtained a
   list of 989 DEGs: 536 upregulated and 453 downregulated, on which we
   performed functional annotation and enrichment analyses ([103]
   Figure 3A and detailed in [104]Supplementary Table S1A ). Results
   presented below satisfied a p-value below 0.0001. Those analyses
   revealed that most biological processes involved in the formation of
   endometriotic lesions were linked to cell adhesion (6.6%, p = 1.9 ×
   10^−10), inflammatory response (5.5%, p = 2.8 × 10^−10), and regulation
   of angiogenesis (2.9%, p = 7.7 × 10^−09). The gene ontology molecular
   functions analysis showed that the DEGs were significantly enriched in
   protein binding (76.7%, p = 1.9 × 10^−16), identical protein binding
   (15%, p = 2.6 × 10^−10), and extracellular matrix structural
   constituent (2.5%, p = 4.8 × 10^−9). In the cellular component, DEGs
   were mainly involved in extracellular exosome (21.2%, p = 2.1 ×
   10^−22), extracellular region (21.2%, p = 2.1 × 10^−22), and
   extracellular space (17.4%, p = 8.8 × 10^−16).

Figure 3.

   [105]This image contains five panels depicting various data
   visualizations related to biological processes and molecular functions
   occuring in endometriosis lesions compared to eutopic endometrium.
   Panel A shows a bar graph categorizing processes like cell adhesion and
   inflammatory response across different compartments and functions.
   Panel B presents a bar graph with gray and red bars showing counts and
   significance levels for key processes in endometriosis. Panel C
   illustrates a network diagram highlighting interactions like
   inflammatory response, hormonal response and regulation of cell-cell
   adhesion. Panel D features a bar graph of enrichment scores for the top
   four most enriched terms. Panel E displays a heatmap with median tau
   scores for various inhibitors across different cell lines highlighting
   the most likely therapy candidates.
   [106]Open in a new tab

   Enriched pathways analysis, functional clustering and computational
   pharmacogenomics of DEGs between endometriosis lesions and eutopic
   endometrium. Gene ontology analysis using DAVID (A) reveals the
   importance of inflammation, cell adhesion, angiogenesis and ECM
   remodeling. Metascape enrichment analysis (B) and relationship network
   of enriched terms visualised in Cytoscape (C) show key events that
   contribute to endometriosis development. Those include inflammatory and
   hormonal response and proliferation and locomotion. Functional
   annotation clustering reports the highest enrichment score for
   complement and coagulation cascade, platelet activation, DNA remodeling
   and integrin mediated signaling respectively (D). Top 15 drug
   candidates identified using a drug repurposing reference database—CMap
   and showing median tau value above 95. JAK, CDK, and topoisomerase
   inhibitors are identified as potential pharmacological targets for
   endometriosis therapy (E).

   KEGG analysis showed enrichment in complement and coagulation cascades
   (2.5%, p = 5.7 × 10^−10), Staphylococcus aureus infection (2.3%, p =
   1.5 × 10^−07), and cell adhesion molecules (2.7%, p = 8.6 × 10^−06).
   The analysis against the Reactome database revealed a key role of
   extracellular matrix organization (5.2%, p = 3.2 × 10^−10), regulation
   of complement cascade (1.5%, p = 8.2 × 10^−07), and complement cascade
   (1.6%, p = 2.4 × 10^−06) ([107] Supplementary Table S1B ).

   A further pathway enrichment analysis was performed with Metascape
   ([108]14) ([109] Figure 3B ; [110]Supplementary Table S2 ) and
   visualized in Cytoscape ([111]15) ([112] Figure 3C ). Tube
   morphogenesis, which relates to vascular development, was the highest
   ranked result of enrichment analysis ([113] Figure 3B ). Inflammatory
   and hormonal response, as well as locomotion and proliferation, were
   among the top 10 most enriched pathways with count values above 75.
   Functional annotation clustering revealed that the complement cascade
   was the most enriched, with a score of 4.14, followed by platelet
   activation pathways, DNA remodeling, and regulation of transcription
   and cell/cell-matrix adhesion processes (enrichment scores of 3.77,
   3.02, and 2.71, respectively) ([114] Figure 3D ).

3.4. Altered expression of complement and coagulation pathway genes.

   Complement and coagulation cascade was the most enriched KEGG pathway
   for the ectopic versus eutopic endometrium comparison ([115] Figure 3D
   ). Genes including C7, C2, C3, A2M, and SERPIN superfamily genes
   involved in this pathway were among the most differentially expressed
   in endometrial tissue ([116] Figure 4 ; [117]Supplementary Dataset ).

Figure 4.

   [118]Six forest plots labeled A to F, each displaying log fold change
   (logFC) with 95% confidence intervals for comparison between EL and EEM
   across studies. Plots vary in logFC values and direction, with blue
   squares indicating estimated values and horizontal lines showing
   confidence intervals. Forest plot show the expression of C1QA
   (complement C1qA chain) - (A), C3 (complement C3) - (B), C7 (complement
   C7) - (C), SERPINE1 (serpin family E member 1) - (D), SERPINE2 (serpin
   family E member 2) - (E), and SERPINA5 (serpin family A member 5) -
   (F).
   [119]Open in a new tab

   Differential gene expression across studies for selected genes from the
   complement and coagulation pathway for the comparison between EL versus
   EEM. Forest plot showing the expression of C1QA (complement C1q A
   chain) - (A), C3 (complement C3) - (B), C7 (complement C7) - (C),
   SERPINE1 (serpin family E member 1) - (D), SERPINE2 (serpin family E
   member 2) - (E), and SERPINA5 (serpin family A member 5) - (F). Direct
   and indirect comparisons from meta-analysis are presented in row number
   five and six. The indirect comparisons had a low impact on the combined
   comparison outcome due to the analyses being performed on datasets
   containing comparisons between EL and EEM ([120] Table 2 ).

   Complement genes, including C1QA (logFC = 1.12; 95% CI = 0.77, 1.47),
   C3 (logFC = 2.40; 95% CI = 1.43, 3.37), and C7 (logFC = 3.36; 95% CI =
   2.50, 4.26), were upregulated in endometrial lesions and showed high
   logFC values. C7 was the gene that showed the highest level of
   upregulation in endometrial lesions among all examined genes.

   Serpins regulate coagulation fibrinolysis processes ([121]19, [122]20)
   and were implicated in the development of endometriosis
   ([123]21–[124]23). Our network meta-analysis showed that serpin genes
   were differentially expressed between endometrial lesions and eutopic
   endometrium. In comparison with the above-presented complement genes,
   serpin family genes were characterized by more heterogeneous expression
   between investigated datasets. SERPINE1 and SERPINE2 were upregulated
   (logFC = 1.02; 95% CI = 0.15, 1.90 and logFC = 1.54; 95% CI = 0.80,
   2.27, respectively), while SERPINA5 was downregulated in lesions (logFC
   = −0.85; 95% CI = −1.52, −0.16, [125]Figures 4D–F ). Detailed
   comparisons for each of the subgroups can be found in
   [126]Supplementary Dataset .

3.5. Mast cell markers

   Our data showed an upregulation in the expression of mast cell markers,
   including CPA3 (logFC = 1.54; 95% CI = 0.96, 2.11), KIT (logFC = 0.74;
   95% CI = 0.30, 1.18), MS4A6A (logFC = 0.71; 95% CI = 0.30, 1.11), and
   markers of mast cell activation, FCGR2B (logFC = 0.78; 95% CI = 0.20,
   1.35) and S100A10 (logFC = 0.87; 95% CI = 0.38, 1.35, [127]Figures 5A–E
   ). The expression of MS4A4A and MS4A2 was also higher in lesions ([128]
   Supplementary Dataset ). Higher amounts of mast cells and their
   increased degranulation have been reported in endometrial tissue of
   animal models and humans ([129]24–[130]26); mast cells colocalized to
   the vasculature of ovarian endometriomas, and they were found to
   promote endometrial cell migration in in vitro assays ([131]27).
   Chromogranin and tryptase, other known markers for mast cells, were not
   significantly increased in lesions ([132] Supplementary Dataset ).
   Recent evidence has shown that chromogranin is not necessarily a
   biomarker of mast cell activation ([133]28). Tryptase is typically
   assessed in patients’ serum, and its level is dependent on many
   factors, including genetic features and comorbidities ([134]29), which
   could explain why in our comparison of tissue expression no difference
   was observed.

Figure 5.

   [135]Eight forest plots labelled A to H,each displaying log fold change
   (logFC) with 95% confidence intervals for comparison between EL and EEM
   across studies. Plots vary in logFC values and direction,with blue
   squares indicating estimated values and horizontal lines showing
   confidence intervals. Forest plot show the expression of CPA3 (mast
   cell carboxypeptidase A) - (A), KIT (tyrosine-protein kinase KIT) -
   (B), MS4A6A (Membrane Spanning 4-Domains A6A) - (C), FCGR2B (Fc Gamma
   Receptor IIb) - (D), S100A10 (S100 calcium-binding protein A10) - (E),
   MMP-9 (Matrix metalloproteinase-9) - (F), STAT5A (Signal Transducer And
   Activator Of Transcription 5A) - (G), and STAT5B (Signal Transducer And
   Activator Of Transcription 5B) - (H).
   [136]Open in a new tab

   Differential gene expression across studies for selected genes,
   including mast cells markers, JAK/STAT pathway, and extracellular
   matrix markers. Forest plot showing the expression of CPA3 (mast cell
   carboxypeptidase A) – (A), KIT (tyrosine-protein kinase KIT) – (B),
   MS4A6A (Membrane Spanning 4-Domains A6A) – (C), FCGR2B (Fc Gamma
   Receptor IIb) – (D), S100A10 (S100 calcium-binding protein A10) – (E),
   MMP-9 (Matrix metalloproteinase-9) – (F), STAT5A (Signal Transducer And
   Activator Of Transcription 5A) – (G), and STAT5B (Signal Transducer And
   Activator Of Transcription 5B) – (H). Direct and indirect comparisons
   from meta-analysis are presented in row number five and six.

3.6. Repurposing JAK and CDK inhibitors for endometriosis therapy.

   We used CMap drug repurposing software to find the most probable
   connections between therapeutic drugs and our network meta-analysis
   results. A median tau score value of 90 or above is considered the
   typical threshold for assessing meaningful drug-induced effects. We
   applied a median tau score cutoff at 95 and selected the top 15 hits.
   This analysis indicated that the candidates most likely to reverse the
   endometriosis mRNA profile were cyclin-dependent kinase (CDK)
   inhibitors, JAK, and topoisomerase inhibitors ([137] Figure 3E ).

   The JAK/STAT3 pathway is thought to govern migratory and invasive
   properties of cells. Its prolonged activation in breast cancer was
   linked with tumor development ([138]30) and resistance to taxane and
   platinum therapy ([139]31). Our results showed an increase in the
   expression of STAT5A (logFC = 0.83; 95% CI = 0.57, 1.08) and STAT5B
   (logFC = 0.57; 95% CI = 0.37, 0.77) in lesions compared with control
   tissue ([140] Figures 5G, H ). JAK3 significantly increased as well,
   but the logFC value was below 0.5 ([141] Supplementary Dataset ).

   JAK inhibitor Ruxolitinib reduced epithelial ovarian cancer cell
   viability ([142]32) and caused growth inhibition of Tam-resistant
   breast cancer cells in vitro. It was shown to lower mRNA VEGF
   expression and reduce the number of vessels and overall tumor weight in
   chorioallantoic assay ([143]31). Ruxotinilib is currently being tested
   in combination therapy for endometrial cancer ([144]33) but its use in
   vitro or in preclinical models of endometriosis has not been reported.
   Tofacitinib, another JAK inhibitor, showed a decrease in endometrial
   lesion size in mice and reduced proliferation of endometrial cancer
   cells in vitro ([145]34).

4. Discussion

   Understanding the main pathways involved in endometriosis development
   is necessary for successful biomarker discovery and improved therapy
   outcomes. Combining data in meta-analysis, we highlight pathogenetic
   mechanisms that are critical for lesion formation regardless of
   endometriosis subtypes and patients’ characteristics.

4.1. Endometrium of women with endometriosis differs from healthy controls

   We detected differences in gene expression between the endometrium of
   healthy women and those suffering from endometriosis, thus showing that
   endometriosis can also affect eutopic endometrium ([146] Figure 2C ).
   CCL21 was upregulated, while BIRC3, LEFTY1, and CEL were downregulated
   in EEM versus EH. Increased expression of CCL21 could suggest that this
   gene takes part in inducing early inflammatory changes in eutopic
   endometrium in women with endometriosis and that it continues its role
   in established lesions ([147] Supplementary Dataset ). Baculoviral IAP
   repeat containing 3 (BIRC3) has not been studied in the context of
   endometriosis. However, its mutations are often present in endometrioid
   adenocarcinoma and endometrial cancer ([148]35). In the latter, the
   lower protein levels of Birc3 correlate with worse patient survival
   ([149]36). One could speculate that the decreased Birc3 expression in
   EEM may contribute to the transformation of the endometrium into
   lesions. Endometrial bleeding associated factor (EBAF/LEFTY1) partakes
   in the regulation of cyclical exfoliation of endometrium and in
   decidualization ([150]37). Healthy endometrium does not express LEFTY1
   during the implantation window, while the endometrium of
   women suffering from endometriosis as well as infertility showed its
   expression ([151]38). Our results agree with that finding and suggest
   that the higher LEFTY1 expression in the EEM group could contribute to
   endometriosis-related infertility. The CEL gene encodes carboxyl ester
   lipase, which partakes in cholesterol and lipid-soluble vitamin ester
   hydrolysis. Its role so far is implicated in diabetes and hereditary
   pancreatitis ([152]39–[153]41) and progression of atherosclerosis. The
   CEL gene has not yet been studied in the context of endometriosis.

4.2. The complement and coagulation cascade in lesion formation.

   We further focused on delineating the expression profile that
   can differentiate ectopic endometrium from eutopic endometrium from
   women with and without endometriosis ([154] Figure 2 ). Gene ontology
   analyses highlighted crucial events accompanying lesion formation.
   Those were immune system activation, angiogenesis, regulation of
   transcription, response to hormones and cytokines, cell adhesion, and
   ECM-cell surface interactions ([155] Figure 3 ). The importance of
   immune system deregulation in endometriosis has been reported
   previously; various inflammatory phenotypes have been associated with
   increased risk of endometriosis ([156]1–[157]3). Our result showed that
   the complement system and platelet coagulation are the two most
   enriched pathways in endometriosis ([158] Figure 3D ). Both processes
   are essential in the natural endometrium growth and shedding cycle. The
   fact that both pathways are the most enriched agrees with the current
   theory that women prone to endometriosis are likely to have a
   different, dysregulated peritoneal microenvironment. The complement
   system is a mediator of tissue growth and regeneration ([159]42) and
   its activation has for a long time been implicated in the development
   of autoimmune disease ([160]43, [161]44) and in promoting tumor growth
   ([162]45). Its dysregulation could therefore provide means for
   immunosurveillance escape and facilitate the implantation of lesions.
   Its importance in the development of endometriosis has been suspected
   since the 80’s ([163]46) and confirmed more recently ([164]47,
   [165]48). Higher amounts of C1, C3, and C5 have been detected in serum
   ([166]49) and peritoneal fluid of women with endometriosis ([167]50,
   [168]51). Various complement proteins were shown to be present in
   epithelial cells of endometrial lesions and ovarian cancer tumors. Its
   local synthesis and deposition have been correlated with the
   progression of various cancer types ([169]52).

   Our data revealed an increased mRNA expression of C1q, C2, and C6, but
   especially C3 and C7, in the lesions ([170] Figures 4A–C ;
   [171]Supplementary Dataset ). C7, a complement cascade member
   responsible for initiation of the membrane attack complex, was the most
   overexpressed gene with the highest fold change in our comparison
   between diseased and control tissue, suggesting its significant role in
   lesion formation ([172] Figures 2A–C ). C7 was found to contribute to
   inflammation and tissue damage in endometriosis ([173]53); it has
   previously been shown to be overexpressed in ovarian cancer ([174]2)
   and stromal cells of endometriomas ([175]3).

   C3, a major effector at which all complement pathways converge, was one
   of the most differentially expressed genes in endometriosis ([176]
   Figure 4B ; [177]Supplementary Figure S1 ). C3 dysregulation is
   involved in most, if not all, inflammatory diseases; it has been found
   upregulated in cancer, cardiac and neurological diseases, asthma, and
   obesity ([178]54–[179]58). Patients with inflammatory bowel disease had
   a higher expression of C3 in their intestinal tissue, and this is
   thought to contribute to chronic inflammation and tissue injury
   ([180]59). Local C3 deposition has been suggested as a prognostic
   factor for gastric cancer ([181]60). A similar situation could occur in
   endometriosis; increased C3 expression could contribute to
   inflammation-driven peritoneal tissue injury, which in turn would
   facilitate lesion implantation. Glandular epithelial cells found in
   endometrial lesions were shown to produce C3 locally ([182]61). The
   activity of both complement cascade members C3 and C4 was higher in the
   serum of women with endometriosis than in those without ([183]62).
   Increased amounts of C1, C3, and C5 were detected in serum ([184]49)
   and peritoneal fluid of women with endometriosis ([185]50, [186]51).
   Similarly, in lesion-bearing mice, C3 was increased in their peritoneal
   fluid. Animals with C3 knockdown formed smaller endometrial cysts and,
   on average, fewer of them ([187]61).

   C3 seems pivotal to endometriosis pathology, and given its strong
   upregulation and presence both in tissue ([188] Figure 4B ) and well in
   peritoneal fluid of endometriosis sufferers, it poses an interesting
   target for early diagnosis and therapy. It has already been proposed as
   an endometriosis serum biomarker ([189]63). However, due to the
   observed discrepancies between tissue and plasma levels of C3, further
   investigation is necessary. In gastric cancer, C3 tissue deposition
   showed a negative correlation with plasma levels, highlighting the need
   for additional research to determine whether C3 is a suitable biomarker
   for endometriosis. As far as treatment is concerned, C3 inhibitors have
   entered clinical trials in anti-ovarian cancer therapy ([190]64) and
   treatment against inflammatory bowel disease ([191]59). Our results
   indicate that the biomarker and therapeutic potential of C3 should be
   studied in endometriosis in more depth.

   Our results revealed strong enrichment in the coagulation cascade and
   showed a dysregulation of SERPIN superfamily genes in endometrial
   lesions ([192] Figures 3D , [193]4D–F ; [194]Supplementary Figure S1 ),
   suggesting an imbalance in the coagulation-fibrinolysis processes
   ([195]19, [196]20).

   SERPINE1 and SERPINE2 were increased in endometrial lesions ([197]
   Figures 4D, E ). SERPINE1-encoded PAI-1 inhibits fibrinolysis and
   contributes to thrombosis/fibrosis. PAI-1 was shown to stimulate
   angiogenesis and facilitate tumor growth and metastasis in primary
   neuroblastoma tumors ([198]65) and in ovarian cancer, its increased
   expression was correlated with tumor cell proliferation and overall
   poor prognosis. The inhibition of PAI-1 suppressed ovarian cancer cell
   growth ([199]66). Its role in endometriosis has also been suggested; by
   inhibiting peritoneal fibrinolysis, PAI-1 is thought to contribute to
   the formation of endometriotic adhesions. PAI-1 was found to be
   increased in deep infiltrating endometriosis as compared to other
   subtypes and eutopic endometrium ([200]67). Moreover, endometriosis
   patients treated with dopamine receptor 2 agonist—quinagolide, showed
   decreased expression of tissue PAI-1 accompanied by a decrease in
   lesion size or its complete disappearance ([201]68). Our results
   further support the importance of PAI-1 in lesion maintenance.

   SERPINE2 was implicated in modulating DNA damage response ([202]69) and
   favoring cancer cell invasion ([203]70). Its pro-metastatic activity
   has been linked to extracellular matrix remodeling and an increase in
   matrix metalloproteinase 9 (MMP-9) expression ([204]71, [205]72). High
   SERPINE2 levels in the endometrium during the secretory phase suggest
   its involvement in tissue remodeling during implantation ([206]73). In
   a mouse model of endometriosis, SERPINE2 showed upregulated expression
   ([207]74); its role in human endometriosis remains unstudied.

   Reduction in SERPINA5 expression was linked with an aggressive tumor
   phenotype and poor prognosis in endometrial and ovarian serous
   carcinomas ([208]75, [209]76). The delivery of SERPINA5 through
   exogenous exosomes decreased the migratory potential of endometrial
   cancer cells ([210]76). Similarly, overexpression of SERPINA5 resulted
   in decreased invasion and angiogenesis in breast cancer ([211]77).
   Decreased SERPINA5 expression was correlated with downstream activation
   of MMP9 in ovarian serous carcinomas ([212]78). Our meta-analysis
   revealed lower SERPINA5 and higher MMP9 expression in endometrial
   lesions ([213] Figure 4F ). Moreover, ECM interactions were indicated
   in enrichment analysis ([214] Figures 3A, B, D ). Taken together, our
   results suggest that the imbalance in the coagulation pathway may be
   affecting extracellular matrix remodeling and contributing to the
   metastatic-like potential of endometriotic cells, thereby promoting
   lesion formation.

4.3. JAK/STAT3 pathway inhibition

   Our search for associations between the endometriosis gene signature
   and CMap reference perturbagens highlighted the role of inhibitors of
   JAK, CDK, and topoisomerase as possible therapy candidates ([215]
   Figure 3E ). JAK/STAT3 pathway dysregulation correlated with an
   increased proliferation and angiogenesis in cancer ([216]79) and with
   various immunodeficiency syndromes ([217]80, [218]81).

   Our meta-analysis revealed an increased expression of both STAT5A and
   STAT5B in lesions compared with control tissue ([219] Figures 5G, H ).
   Others have shown that phosphorylation of STAT3 was upregulated in
   endometriosis lesions ([220]82) and activated STAT3 increased
   proliferation of endometrial stromal cells ([221]83).

   Interestingly, increased C3 expression was shown to trigger the
   JAK/STAT3 pathway in gastric cancer, which led to a subsequent increase
   in cell proliferation. C3 inhibition with CR1 decreased that activation
   ([222]60). Our results present a similar picture; complement C3 as well
   as the JAK/STAT3 pathway seems to play a role in the development of
   endometriosis. This association needs further investigation.

4.4. Proposed pathways crosstalk in endometriosis

   It has been proposed that both the complement system and coagulation
   pathways are tightly linked; coagulation factors have been reported to
   cleave and activate complement members C3 and C5 ([223]84). On the
   other hand, C3 was shown to protect clots from fibrinolysis ([224]85).
   Increased amounts of C3 protein were shown to provoke mast cell
   activation, and various mast cell mediators were implicated in the
   regulation of coagulation and fibrinolysis in anaphylaxis ([225]86).
   Our meta-analysis revealed that endometrial lesions had a higher
   expression of mast cell markers, including KIT, CPA3, MS4A6A, FCGR2B,
   and S100A10 ([226] Figures 5A–E ). An increased mast cell burden was
   detected previously in animal and human endometrial tissue ([227]27).
   Moreover, our results showed that endometrial lesions had a higher
   level of STAT5A and STAT5B ([228] Figures 5G, H ), members of the
   JAK/STAT3 pathway, which regulate mast cells ([229]87). Targeting mast
   cells with JAK inhibitors for alleviation of symptoms of endometriosis
   was proposed almost two decades ago ([230]88) but not much research has
   been carried out on the topic since. Our current results fill this gap
   and suggest the use of JAK inhibitors as immunomodulators in
   endometriosis. Interestingly, a cooperation between mast cells,
   complement, and coagulation pathways has been reported in an
   inflammatory disease—chronic spontaneous urticaria ([231]89). Our
   analysis indicates that there exists an interplay between the
   complement and coagulation pathways, mast cell activation, ECM
   remodeling, and the JAK/STAT3 pathway (summarized in [232]Figure 6 ).
   To the best of the author’s knowledge, this relationship has not yet
   been studied in endometriosis, and our results warrant a further
   in-depth look into those processes.

Figure 6.

   [233]Schematic representation of proposed key molecular processes
   contributing to lesion development. The diagram illustrating
   interactions in immune response and inflammation. The complement and
   coagulation cascade is central, mast cells activation, JAK/STAT3
   pathway, and ECM remodeling. Arrows show the directional change in gene
   expression in endometriosis lesions compared with control endometrium.
   [234]Open in a new tab

   Schematic representation of proposed key molecular processes
   contributing to lesion development. This diagram illustrates the
   complex interplay between the complement system, coagulation cascade,
   extracellular matrix remodeling, and immune signaling pathways in the
   context of lesion formation. Coagulation factors cleave and activate
   complement components C3 and C5, initiating pro-inflammatory responses.
   Complement factor C3 also functions to stabilize clots by protecting
   them from fibrinolysis. The coagulation cascade is tightly regulated by
   members of the serpin superfamily, which modulate extracellular matrix
   remodeling through the induction of matrix metalloproteinase-9 (MMP-9)
   expression. Complement activation further engages the JAK/STAT
   signaling pathway, particularly STAT5, promoting mast cell activation.
   Activated mast cells, in turn, influence coagulation dynamics.
   Pharmacologic inhibition of the JAK/STAT pathway using JAK inhibitors
   can suppress STAT3 signaling and reduce mast cell degranulation,
   thereby modulating both inflammatory and thrombotic processes. These
   mechanisms highlight the therapeutic potential of JAK inhibitors in the
   context of endometriosis. Genes names in green signify differentially
   expressed genes, arrows show the directional change in gene expression
   in endometriosis lesions compared with control endometrium. Created in
   [235]BioRender.com.

4.5. Strengths and limitations

   Our network meta-analysis enabled us to arrive at a consensus
   endometriosis signature. The use of publicly deposited endometriosis
   transcriptomic data collected on three different continents, spanning
   various age groups, and ethnicities, as well as various types and
   stages of endometriosis, enabled a comprehensive, unbiased, and
   multi-demographic comparison of endometriotic and control tissue.

   The following limitations should be considered when interpreting our
   results. Our meta-analysis included only nine datasets because most of
   the studies lacked a control group, included therapeutic intervention,
   or performed RNA isolation on processed tissue. Secondly, only
   published studies where the absence of endometriosis was excluded by
   laparoscopy were included in this meta-analysis. Therefore, publication
   bias may have occurred, although none was indicated by the funnel plot.

4.6. Conclusions and clinical implications

   We highlight the role of complement and coagulation cascade in
   endometriosis and propose an interplay between both those processes and
   mast cells, ECM interaction, and the JAK/STAT3 pathway that needs
   further investigation. We underscore the significance of C3 and call
   for further research into its diagnostic and therapeutic potential.
   Furthermore, we propose JAK inhibitors discovered in drug repurposing
   analysis and validated in vitro as potential therapy candidates.

   Our results show differences in expression in eutopic endometrium from
   patients with and without endometriosis. Those should be further
   explored to understand if they contribute to endometrial seeding.
   Detected gene differences may be potential biomarkers that could be
   used in the less invasive endometriosis biopsy and should be further
   studied.

Funding Statement

   The author(s) declare that financial support was received for the
   research and/or publication of this article. This research is part of
   the project No. 2022/47/P/NZ5/02484 co-funded by the National Science
   Centre and the European Union Framework Programme for Research and
   Innovation Horizon 2020 under the Marie Skłodowska-Curie grant
   agreement No. 945339. For the purpose of Open Access, the author has
   applied a CC-BY public copyright licence to any Author Accepted
   Manuscript (AAM) version arising from this submission;”. The authors
   would like to acknowledge Medical University of Lodz – Digital Medicine
   Center (MULDiMediC) project (2023/ABM/02/00009) financed by Medical
   Research Agency in Poland for access to its computational resources.

   Inline graphic

Abbreviations

   EM, endometriosis; EH, endometrium from healthy controls; EEM,
   endometrium from women with endometriosis; EL, endometrial lesions;
   DEGs, differentially expressed genes; GEO, Gene Expression Omnibus.

Data availability statement

   The original contributions presented in the study are included in the
   article/[236] Supplementary Material . Further inquiries can be
   directed to the corresponding author.

Author contributions

   MG: Conceptualization, Writing – review & editing, Supervision, Funding
   acquisition, Writing – original draft, Visualization, Data curation,
   Methodology, Project administration, Formal Analysis. AR: Formal
   Analysis, Writing – original draft, Visualization, Data curation. MS:
   Data curation, Writing – original draft. JC: Methodology, Writing –
   original draft. KS: Methodology, Writing – original draft. WF:
   Supervision, Writing – review & editing, Methodology, Writing –
   original draft.

Conflict of interest

   The authors declare that the research was conducted in the absence of
   any commercial or financial relationships that could be construed as a
   potential conflict of interest.

Generative AI statement

   The author(s) declare that no Generative AI was used in the creation of
   this manuscript.

Publisher’s note

   All claims expressed in this article are solely those of the authors
   and do not necessarily represent those of their affiliated
   organizations, or those of the publisher, the editors and the
   reviewers. Any product that may be evaluated in this article, or claim
   that may be made by its manufacturer, is not guaranteed or endorsed by
   the publisher.

Supplementary material

   The Supplementary Material for this article can be found online at:
   [237]https://www.frontiersin.org/articles/10.3389/fimmu.2025.1619434/fu
   ll#supplementary-material
   [238]DataSheet1.xlsx^ (1.7MB, xlsx)
   [239]Table1.docx^ (493.5KB, docx)

References