Identification of a circulating immunological signature as a liquid biopsy approach for the diagnosis of endometriosis

After signing informed consent, patients suffering from endometriosis, scheduled for surgical removal of the endometrioma (n = 78), or for salpingectomy for contraception purposes as controls (n = 48), were sequentially recruited at the Gynecology service at La Paz University Hospital. Inclusion criteria required patients to be of reproductive age and to have a diagnosis of endometriosis confirmed by histopathological examination. Minors and women with any infectious disease were excluded. Age-based comparative analyses employed a 35-year cut-off, as this threshold aligns with fertility-related clinical criteria, reflects relevant biological transitions, and facilitates comparison with prior studies 26 . Fertility status was assessed based on the obstetric history obtained during medical evaluation. Similarly, in terms of endometrioma size, we used 5 cm as the cut-off, as numerous studies and clinical protocols adopt this threshold when analyzing outcomes related to immune profiles, clinical symptoms, or treatment response 27 , 28 . Considering the immunological relevance of the affected organs and anatomical compartments 29 , patients were classified as ovarian or deep infiltrating endometriosis according to the Enzian classification 30 . No cases of peritoneal endometriosis were included in the study because symptoms were effectively controlled with hormonal therapy; therefore, surgical intervention was not required in these patients. Finally, BMI, measured and recorded by physicians during clinical assessment, was categorized as low or high based on the median value observed in our dataset. This study was conducted in accordance with the ethical guidelines of the 1975 Declaration of Helsinki and was approved by the local ethics committee of La Paz University Hospital, where the study was conducted (PI-4623). Demographic and clinical data of controls and patients are reported in Table 1 . Table 1 Demographic and clinical data of participants. Variable Control Group (n = 48) Endometriosis Group (n = 78) Median age, years (OR CI 95%) 36 (33–37) 37 (35–39) Body Mass Index (BMI), kg/m 2 (± SD) 24.66 (± 5.067) 22.99 (± 6.729) Infertility, n (%) NA 30 (38.46) Endometrioma location, n (%) 78 (100)  No endometrioma NA 8 (10.26)  Unilateral NA 51 (65.38)  Bilateral NA 19 (24.36) Endometrioma size, n (%) 78 (100)  No endometrioma NA 8 (10.26)    50 mm NA 25 (32.05) Endometriosis Subtypes, n (%) 78 (100)  Ovarian NA 33 (42.31)  Deep infiltrating NA 45 (57.69) Comorbidities, n (%) 5 (10.42) 30 (38.46)  Autoimmune diseases (Vitiligo, Psoriasis, Crohn Disease, Crest Syndrome, Multiple Sclerosis) 0 6 (7.69)  Digestive diseases (Celiac Disease, Gastritis, Hypercholesterolemia) 1 (2.08) 4 (5.13)  Endocrine diseases (Diabetes, Hypothyroidism, Thyroid cancer) 2 (4.17) 12 (15.38)  Respiratory diseases (Asthma) 2 (4.17) 8 (10.26) Treatment, n (%) 12 (25.00) 76 (97.44)  Hormonal Treatment (Estroprogestin/Progestins) 12 (100) 42 (53.85)  NSAIDs 0 34 (43.59) NA not applicable. Demographic and clinical data of participants. NA not applicable. Blood samples were collected before surgical intervention from the participants in K 3 EDTA coated tubes (REF:13060 Vacutest® Kima) and gel & clot activator tubes (REF:10313 Vacutest ® Kima). Soluble blood fractions were obtained after centrifugation for 10 min at 10.000 rpm, aliquoted and conserved at − 80 °C until their analysis. All patients included in the study were in the proliferative phase of the menstrual cycle, thereby minimizing the known impact of this factor on the assessment of immune markers 29 , 31 . Moreover, all blood samples were collected after induction of anesthesia and before the scheduled surgical procedures (salpingectomy for controls and endometrioma excision for patients). Levels of circulating soluble immune checkpoints and cytokines were analyzed using a cytometry bead-based assay from BioLegend (immune checkpoint panel 1 Ref: 740867, and essential immune response panel 13-plex Ref: 740930, respectively). These kits include a panel of a pre-established set of soluble immune checkpoints (sCD25, s4-1BB, sCD27, sCD86 (B7.2), TGF-β1, sCTLA-4, sPD-L1, sPD-L2, sPD-1, sTim-3, sLAG3, Galectin-9) and cytokines (IL-4, IL-2, CXCL10 (IP-10), IL-1β, TNF, CCL2 (MCP-1), IL-17A, IL-6, IL-10, IFN-γ, IL-12p70, CXCL8 (IL-8), TGF-β1). The broad range of cytokines and immune checkpoints included in these panels has proven valuable for other clinical analyses across various pathological contexts, including neuroendocrine tumors 32 and COVID-19 33 . The quantifications were performed following the manufacturer instructions. Briefly, samples were twofold diluted in assay buffer and incubated for 2 h at room temperature with the coated beads. After that time, the beads were centrifuged, washed and incubated with the detection antibodies (conjugated to biotin) for one hour at room temperature. Finally, the beads were incubated with streptavidin–phycoerythrin for 20 min at room temperature, centrifuged and washed before flow cytometry. Human SIGLEC5 ELISA (Enzyme-linked immunoassay) kit (Sigma REF: RAB0433-1KT) was used to perform the determination of soluble Siglec-5. Following the instructions provided by the manufacturer, samples were incubated at room temperature for 2.5 h, followed by 4 washes with 300 µl of wash solution. Detection antibody was incubated for 1 h at room temperature and washed 4 times with 300 µl of wash solution. Finally, 100 µl of TMB One-Step substrate reagent were added to each well and, after 30 min the reaction was stopped by adding 50 µl of stop solution to each well. The absorbance was measured at 450 nm. Each sample was 250-fold diluted in this assay. Data of single soluble parameters are depicted as box and whiskers plots; lines inside the boxes delimitate the average and 10 – 90 quartiles, as well as standard deviation. D’Agostino and Pearson Normality test was performed to all the variables included in the study. Considering that data, one-way ANOVA followed by Fisher’s test or Kruskal–Wallis’ test followed by uncorrected Dunn’s test were performed for more than two groups comparison, and T-test or Mann–Whitney test for two groups comparison. The Backwards Wald method was used to generate binary logistic regression models using SPSS version 23 (IBM) software. The Wald automatic stepwise selection method starts from the complete set of independent variables to be studied and, gradually and randomly, removes and reintroduces them at each step until only the explanatory ones remain 34 . Based on this logistic regression model, classifying scores were generated as previously described 32 , 35 , 36 . Briefly, considering the concentrations of the soluble parameters included in every model, and the B factor provided by the algorithm associated to each of them, scores were generated following the formula: Score = ([parameter A] x B factor A) + ([parameter B] x B factor B) + … + ([parameter X] x B factor X). These scores were used to generate the Receiver Operating Characteristic (ROC) curves, determining the Area Under the Curve (AUC) and optimal cut-off values as the Youden index. The AUC, sensitivity, and specificity are shown, as well as their 95% confidence intervals. All along the figures, p -values ( p ) are denoted as * p  < 0.05, ** p  < 0.01, *** p  < 0.001, **** p  < 0.0001. Only variables showing statistically significant differences are depicted for clarity reasons. Statistical analyses were conducted using Prism 8.0 (GraphPad).

In an attempt to address whether the presence of endometriosis has an impact on circulating immunological factors, we evaluated the levels of a panel of cytokines (IL-4, IL-2, IL-1β, IL-6, IL-10, IFN-γ, IL-12p70, IL-17A, TNF, TGF-β1, IP-10(CXCL10), MCP-1(CCL2) and IL-8) and soluble immune checkpoints (sSiglec-5, sB7.2(CD86), sCD25, sPD-L2, Galectin-9, sPD-L1, sPD-1, sCTLA-4, sTim-3, s4-1BB, sCD27, and sLAG3) in samples from patients suffering from endometriosis. Blood samples from women scheduled for salpingectomy for contraception were considered as controls. Among the cytokines, patients suffering from endometriosis showed increased TNF expression while the levels of TGF-β1, IP-10, MCP-1, and IL-8 were reduced (Fig.  1 A). Interestingly, the presence of endometriosis showed a consistent expression pattern of soluble immune checkpoints, with reduced levels of all the analyzed factors except sSiglec-5 and sB7.2 (Fig.  1 B). Fig. 1 Circulating levels of cytokines and soluble immune checkpoints in patients suffering from endometriosis. The concentration of cytokines ( A ) IL-4, IL-2, IL-1β, IL-6, IL-10, IFN-γ, IL-12p70, IL-17A, TNF, TGF-β1, IP-10(CXCL10), MCP-1(CCL2) and IL-8 and soluble immune checkpoints ( B ) sSiglec-5, sB7.2(CD86), sCD25, sPD-L2, Galectin-9, sPD-L1, sTim-3, sCTLA-4, sPD-1, s4-1BB, sCD27, and sLAG3 was analyzed in blood soluble fractions of controls (women scheduled for salpingectomy with contraception purposes) and patients diagnosed with endometriosis. Data are shown as box and whiskers showing quartiles. Comparisons between the two groups of participants were performed by Mann–Whitney test. P values are represented as * p  < 0.05, ** p  < 0.01, *** p  < 0.001, **** p  < 0.0001. Circulating levels of cytokines and soluble immune checkpoints in patients suffering from endometriosis. The concentration of cytokines ( A ) IL-4, IL-2, IL-1β, IL-6, IL-10, IFN-γ, IL-12p70, IL-17A, TNF, TGF-β1, IP-10(CXCL10), MCP-1(CCL2) and IL-8 and soluble immune checkpoints ( B ) sSiglec-5, sB7.2(CD86), sCD25, sPD-L2, Galectin-9, sPD-L1, sTim-3, sCTLA-4, sPD-1, s4-1BB, sCD27, and sLAG3 was analyzed in blood soluble fractions of controls (women scheduled for salpingectomy with contraception purposes) and patients diagnosed with endometriosis. Data are shown as box and whiskers showing quartiles. Comparisons between the two groups of participants were performed by Mann–Whitney test. P values are represented as * p  < 0.05, ** p  < 0.01, *** p  < 0.001, **** p  < 0.0001. This first analysis indicates that the presence of endometriosis generates specific circulating immunological patterns that could differentiate healthy women from patients suffering from this condition. Based on the observations above, we evaluated the predictive capacity of soluble immune checkpoints and cytokines in circulation to identify patients suffering from endometriosis. With this aim, we performed a binary logistic regression model including those immunological variables that showed statistical differences between the control and endometriosis groups in Fig.  1 . This regression model generated a score that was assessed by receiver operating characteristic (ROC) curves, determining the area under the curve (AUC), and the optimal cut-off score, estimated by the Youden index 32 , 35 , 36 . The regression model generated a score including eight variables (Fig.  2 A). The ROC analysis showed that this immunological signature identified endometriosis with a remarkable yield, displaying an AUC = 0.8881 (95% CI 0.8297–0.9464) (Fig.  2 B). Furthermore, the optimal cut-off exhibited 72.92% specificity (95% CI 59.00–83.43%) and a great 92.31% sensitivity versus controls (95% CI 84.22–96.42%) (Fig.  2 B, C). These results indicate that the analysis of circulating immune checkpoints and cytokines represents a highly efficient tool to identify patients suffering from endometriosis, providing a liquid biopsy approach. Fig. 2 Regression model considering circulating immunological factors identify patients suffering from endometriosis. Wald backward stepwise regression, including as variables those statistically different in Fig.  1 (TNF, TGF-β1, IP-10, MCP-1, IL-8, sCD25, sPD-L2, Galectin-9, sPD-L1, sTim-3, sCTLA-4, sPD-1, s4-1BB, sCD27 and sLAG-3) was performed. ( A ) Optimal model differentiating patients suffering from endometriosis from controls (women scheduled for salpingectomy with contraception purposes). ( B ) ROC curve analysis for identification of endometriosis. C Distribution of controls and patients diagnosed with endometriosis according to the score generated as the optimal Youden index from the ROC curve in B. Area under the curve (AUC) is shown, as well as sensitivity and specificity of 95% confidence intervals are shown in brackets. **** p  < 0.0001, Mann–Whitney test. Regression model considering circulating immunological factors identify patients suffering from endometriosis. Wald backward stepwise regression, including as variables those statistically different in Fig.  1 (TNF, TGF-β1, IP-10, MCP-1, IL-8, sCD25, sPD-L2, Galectin-9, sPD-L1, sTim-3, sCTLA-4, sPD-1, s4-1BB, sCD27 and sLAG-3) was performed. ( A ) Optimal model differentiating patients suffering from endometriosis from controls (women scheduled for salpingectomy with contraception purposes). ( B ) ROC curve analysis for identification of endometriosis. C Distribution of controls and patients diagnosed with endometriosis according to the score generated as the optimal Youden index from the ROC curve in B. Area under the curve (AUC) is shown, as well as sensitivity and specificity of 95% confidence intervals are shown in brackets. **** p  < 0.0001, Mann–Whitney test. Next, we wanted to evaluate whether demographic or clinical features of patients could influence the good performance of the score as diagnostic tool. Of note, the score’s diagnostic capacity was neither affected when women were split into two groups based on their age, as over or below 35 years old (Fig.  3 A), nor based on their BMI (Fig.  3 B). Furthermore, the score maintained its efficient diagnostic capacity independently of endometriosis subtypes, patient’s infertility status, endometrioma size, clinical comorbidities, endometrioma location, and treatment (Fig.  3 C–H). Therefore, the evaluation of an immunological signature performs an efficient diagnosis of patients suffering from endometriosis across diverse clinical conditions. Fig. 3 Performance of the diagnostic score for endometriosis considering different demographic and clinical features of the patients. Distribution of the diagnostic score generated in Fig.  2 between controls and patients diagnosed with endometriosis according to different clinical factors. ( A ) Age ( 35 years). ( B ) BMI was categorized as low or high based on the median value of each group. Distribution of controls and patients based on endometriosis subtypes ( C ), infertility status of patients ( D ), endometrioma size ( E ), different comorbidities ( F ), endometrioma location ( G ), and treatment ( H ). Data are shown as box and whiskers showing quartiles. Comparisons between the groups of individuals were performed by Kruskal–Wallis’ test. p values are represented as * p  < 0.05, ** p  < 0.01, *** p  < 0.001, **** p  < 0.0001. Performance of the diagnostic score for endometriosis considering different demographic and clinical features of the patients. Distribution of the diagnostic score generated in Fig.  2 between controls and patients diagnosed with endometriosis according to different clinical factors. ( A ) Age ( 35 years). ( B ) BMI was categorized as low or high based on the median value of each group. Distribution of controls and patients based on endometriosis subtypes ( C ), infertility status of patients ( D ), endometrioma size ( E ), different comorbidities ( F ), endometrioma location ( G ), and treatment ( H ). Data are shown as box and whiskers showing quartiles. Comparisons between the groups of individuals were performed by Kruskal–Wallis’ test. p values are represented as * p  < 0.05, ** p  < 0.01, *** p  < 0.001, **** p  < 0.0001. Demographic data such as age and BMI are easy-to-collect variables from patients, which can be informative and enrich the regression model. When age and BMI were considered, the model generated a signature based on just four immunological variables (sCD25, sPD-L1, sLAG-3, and IP-10), along with both age and BMI (Fig.  4 A). This is relevant because a predictive signature based on a reduced number of parameters would facilitate the technical implementation of this liquid biopsy approach. This model rendered also a high performance, with an AUC of 0.8243 (95% CI 0.7486–0.8999) (Fig.  4 B), 60.42% specificity (95% CI 46.31–72.98%) and a remarkable 92.31% sensitivity versus controls (95% CI 84.22–96.43%) (Figs.  4 B, C). Therefore, the analysis of only four circulating immunological parameters, plus two commonly recorded demographic variables, allows for an efficient diagnosis of endometriosis. Fig. 4 Regression model including age and BMI generates a minimal signature of clinical application for the diagnosis of endometriosis. Wald backward stepwise regression, including as variables those immunological factors statistically different in Fig.  1 (TGF-β1, TNF, IP-10, MCP-1, IL-8, sCD25, sPD-L2, Galectin-9, sPD-L1, sPD-1, sCTLA-4, sTim-3, s4-1BB, sCD27 and sLAG-3) plus age and body mass index (BMI) of the participants was performed. ( A ) Optimal model differentiating patients suffering from endometriosis from controls (women scheduled for salpingectomy with contraception purposes). ( B ) ROC curve analysis for identification of endometriosis. ( C ) Distribution of controls and patients diagnosed with endometriosis according to the score generated as the optimal Youden index from the ROC curve in B. Area under the curve (AUC) is shown, as well as sensitivity and specificity of 95% confidence intervals are shown in brackets. **** p  < 0.0001, Mann–Whitney test. Regression model including age and BMI generates a minimal signature of clinical application for the diagnosis of endometriosis. Wald backward stepwise regression, including as variables those immunological factors statistically different in Fig.  1 (TGF-β1, TNF, IP-10, MCP-1, IL-8, sCD25, sPD-L2, Galectin-9, sPD-L1, sPD-1, sCTLA-4, sTim-3, s4-1BB, sCD27 and sLAG-3) plus age and body mass index (BMI) of the participants was performed. ( A ) Optimal model differentiating patients suffering from endometriosis from controls (women scheduled for salpingectomy with contraception purposes). ( B ) ROC curve analysis for identification of endometriosis. ( C ) Distribution of controls and patients diagnosed with endometriosis according to the score generated as the optimal Youden index from the ROC curve in B. Area under the curve (AUC) is shown, as well as sensitivity and specificity of 95% confidence intervals are shown in brackets. **** p  < 0.0001, Mann–Whitney test. As done before, we wanted to test the strength of this diagnostic score considering different clinical variables of the patients, evaluating its universality. Consistent with our prior findings using only soluble immunological variables (Fig.  3 ), this minimal signature also identified patients suffering from endometriosis with diverse clinical conditions and treatments, supporting the score’s universality (Fig.  5 A–F). Fig. 5 Performance of the minimal diagnostic score for endometriosis considering different demographic and clinical features of the patients. Distribution of controls and endometriosis patients according to the minimal diagnostic score based on endometriosis subtypes ( A ), infertility status of patients ( B ), endometrioma size ( C ), different comorbidities ( D ), endometrioma location ( E ), and treatment ( F ). Data are shown as box and whiskers showing quartiles. Comparisons between the groups of individuals were performed by Kruskal–Wallis’ test. p values are represented as * p  < 0.05, ** p  < 0.01, *** p  < 0.001, **** p  < 0.0001. Performance of the minimal diagnostic score for endometriosis considering different demographic and clinical features of the patients. Distribution of controls and endometriosis patients according to the minimal diagnostic score based on endometriosis subtypes ( A ), infertility status of patients ( B ), endometrioma size ( C ), different comorbidities ( D ), endometrioma location ( E ), and treatment ( F ). Data are shown as box and whiskers showing quartiles. Comparisons between the groups of individuals were performed by Kruskal–Wallis’ test. p values are represented as * p  < 0.05, ** p  < 0.01, *** p  < 0.001, **** p  < 0.0001.

The findings from this study demonstrate that endometriosis has a systemic effect on the immune system by modulating circulating soluble immune checkpoints and cytokine levels. Patients with endometriosis displayed increased TNF expression, while TGF-β1, IP-10, MCP-1, and IL-8 levels were lower. Elevated TNF levels were consistent with findings in Malutan et al. in endometriosis patients compared to controls 37 , as well as the reduced serum concentration of MCP-1 (CCL2) and the same trend for IL-8 (CXCL8) reported by Sommar et al. 38 . Nevertheless, the available studies evaluating blood cytokine concentrations as potential biomarkers for endometriosis remain limited and yield inconsistent findings, as outlined below. Furthermore, most research to date has focused on evaluating the diagnostic potential of inflammatory proteins in PF and/or tissue samples (i.e., eutopic/ectopic endometrium) 39 – 41 . Regarding analyses in blood, Podgaec et al. detected no differences in serum concentrations of the five analyzed cytokines (TNF, IFN-γ, IL-2, IL-4, IL-10), but did find significantly higher IFN-γ and IL-10 levels in the PF of endometriosis patients 42 . Elevated PF concentrations of IL-8 and MCP-1 in endometriosis were also reported by Borrelli et al., along with increased MIP-3β 43 . More recently, Fan et al. studied seven cytokines, including IL-6, IL-4, and IL-2, in both serum and PF from patients with endometriosis and control subjects. Consistent with our findings, they observed significantly lower levels of IL-2 in serum, while IL-2 levels were significantly higher in PF 44 . Though PF appears to be the most representative diagnostic sample, closely reflecting the inflammatory changes linked to the pathogenesis of endometriosis 45 , its collection requires a semi-invasive procedure, making it less suitable for clinical practice. Interestingly, our data revealed that endometriosis was associated with a consistent pattern of soluble immune checkpoint expression, with 10 out of the 12 factors analyzed showing lower plasma levels in patients with endometriosis. Studies analyzing soluble immune checkpoint levels are even rarer than those examining cytokine levels, and the reported values also fluctuate among studies. In this sense, and in contrast to our study, Santoso et al. did not find significant differences in the serum levels of sCTLA-4, sHLA-G, and sPD-1 between the endometriosis and control groups, but noted significantly higher levels of serum sPD-L1 and all four molecules in PF 25 . Similarly to our findings, Suszczyk et al. found lower plasma sPD-L1 levels in endometriosis patients compared to healthy controls 46 . However, the immunological factors in these studies were measured using different ELISA kits for individual factors, which may limit the establishment of a comprehensive immunological profile of the disease. Furthermore, several studies have highlighted that plasma is a more reliable sample than serum for systemic cytokine profiling, as coagulation processes in serum can cause artificial fluctuations in immune factors levels 47 , 48 . Given the highly dynamic nature of these soluble proteins, plasma-based measurements better reflect their physiological levels, minimizing artifacts from clot formation. Therefore, we believe that our findings in terms of the global study of several immunological factors have at least two key implications. First, the biological data obtained contribute to understanding the disease’s pathophysiology. In this sense, the increased systemic TNF levels, along with the robust downmodulation of soluble immune checkpoints, illustrate the proinflammatory nature of the endometriosis. This is likely associated with a strong peritoneal recruitment of inflammatory cells mediated by a gradient of chemokines indicated by the reduced levels of IP-10 (CXCL10), MCP-1 (CCL2), and IL-8 (CXCL8) in plasma coupled to elevated levels of some of these chemokines in PF 43 . Indeed, conditioned media from primary cultures of surgically resected specimens from endometriosis patients produce MCP-1 and IL-8, among other chemoattractant factors 49 , supporting this model. Second, the specific circulating immunological patterns identified might be useful for distinguishing healthy women from those with endometriosis. Regarding the first point, given that single biomarkers are often insufficient for disease diagnosis or prediction, the study of multiple proteins is increasingly recognized as crucial 50 – 53 . Recent biomarker studies make use of multiplex platforms, capable of screening thousands of markers, which have been used in diseases such as Kawasaki disease, aortic aneurysms and rheumatoid arthritis to identify individual markers and explore pathways and multiple involved proteins 50 – 54 . Additionally, multivariate statistical methods, in contrast to univariate approaches, allow for simultaneous analysis of multiple variables, improving data interpretation and prediction. Logistic regression is a statistical approach used when modeling the probability of disease presence or absence, making it valuable for endometriosis research 55 . Considering the diagnostic value of multiparametric analysis, we used logistic binary regression models to enhance the predictive ability of circulating immunological parameters in endometriosis patients. The regression model generated a score based on eight immunological factors (s4-1BB, sCTLA-4, sPD-L1, sPD-L2, sLAG-3, IP-10, TNF, MCP-1), achieving a specificity of 72.92% and a high sensitivity of 92.31%. Recently, Knific et al. assessed the diagnostic potential of various cytokines in plasma from 210 patients with different types of endometriosis and controls 56 . Despite analyzing the soluble cytokines and clinical data with appropriate statistical methods, their model achieved only a sensitivity of 40%, a specificity of 65%, and an AUC of 0.61, which is inadequate for diagnostic purposes. Similarly, Jørgensen et al. conducted a cross-sectional study measuring 48 cytokines in PF using multiplex immunoassays 57 . Among them, five cytokines showed significant differences, and a logistic regression model using three parameters predicted endometriosis with 86% sensitivity and 67% specificity. However, this model was based on PF, which is not easily accessible for diagnosing endometriosis. Considering our findings compared with the studies previously mentioned, our regression model indicates that analyzing circulating immune checkpoints and cytokines is a robust and effective method for detecting endometriosis, providing a liquid biopsy approach. This diagnostic score maintained its performance across demographic variables such as age and BMI, as well as clinical factors including fertility status, endometrioma location, size, treatment, and endometriosis subtype (ovarian or deep infiltrating). Along this line, it also effectively distinguished patients with and without comorbidities. Although no significant differences were found for respiratory, autoimmune, and digestive comorbidities, most probably due to the small number of patients, the score showed a similar trend. Therefore, the signature obtained from analyzing soluble circulating immunological factors functions as an efficient diagnostic tool for endometriosis, proving useful across different demographic and clinical conditions of the patients. Demographic factors, such as age and BMI, are easily collected and can enhance regression models for diagnosing endometriosis. In this sense, Kocbek et al. employed multiple regression models incorporating cytokine and secretory protein panels to diagnose ovarian endometriosis, achieving a high AUC of 0.90 58 . Their models relied on biomarker ratios (biglycan/leptin, RANTES/IL-6, and ficolin-2/glycodelin-A) and IL-8, adjusted for age, showed strong diagnostic power but posed challenges for clinical implementation due to its complexity. In contrast, our study demonstrates that a minimal immunological signature, combined with demographic data, is sufficient for high-efficiency diagnosis. Our model, incorporating four immunological markers (sCD25, sPD-L1, sLAG-3, and IP-10) with age and BMI, simplifies the diagnostic approach while maintaining strong performance (AUC = 0.8243). This minimal signature enables differentiation between groups using a single regression model with an easily implementable multiplex assay panel similar to our immunological parameter-based signature, this approach successfully identified endometriosis patients regardless of fertility status or comorbidities. It was even more robust, as it distinguished nearly all endometriosis patients with comorbidities from healthy women. In a recent study, Fauconnier et al. developed a predictive model based on a self-reported pain questionnaire, achieving high diagnostic accuracy, with an AUC of 0.92 59 . However, reliance on self-reported symptoms introduces biases and lacks biological confirmation. In contrast, our logistic regression model, incorporating a minimal immunological signature with age and BMI to diagnose endometriosis, enhances symptom-based screening using circulating biomarkers, which offer objective and quantifiable data. Importantly, the model achieves a high sensitivity (92.31%), reducing false-positive diagnoses, a limitation in symptom-based approaches. Although our specificity (60.42%) is lower than that reported in the model mentioned before, it still provides a valuable minimally diagnostic tool. Future studies integrating both strategies could further refine non-invasive diagnostic workflows. Additionally, factors such as endometrioma location, size, treatment, or endometriosis subtypes had no impact on the diagnostic performance of the obtained signatures, suggesting a minimal impact of these clinical features on the circulating immunological factors analyzed. In a previous research conducted by Stegmann et al., they used the histological diagnoses in their regression model, achieving a sensitivity of 88.4% and a specificity of 24.6%, suggesting the need for additional diagnostic methods like imaging or biomarker-based approaches 60 . They excluded age and BMI due to their lack of predictive value, whereas our results suggest that combining four soluble immunological markers with these demographic variables provides a reliable, non-invasive diagnostic tool. Overall, this represents a promising liquid biopsy approach with significant clinical potential. Nevertheless, some limitations of our study need to be standout. Despite the relevance of the findings, the limited number of participants (78 patients and 48 controls) may constrain the extent to which these results can be generalized to broader populations. Additionally, the predictive models were developed and evaluated within the same cohort, lacking external validation in independent populations to confirm their diagnostic performance. Finally, as an observational study, it does not explore the mechanistic roles of the altered immune checkpoints and cytokines in the pathogenesis of endometriosis, nor the status of the pathology at the local level. Future studies in this sense, addressing the potential role of ICs within endometrial tissue, might shed light on the pathophysiology of endometriosis. Despite these limitations, our results provide a new diagnostic tool of potential clinical application for the early diagnosis of endometriosis, a complex disease of increasing awareness. This tool is based on a blood test, providing a liquid biopsy approach that could provide an efficient and easy diagnosis. This approach emerges as a promising instrument in endometriosis research as well as in other fields, including cancer research.

Endometriosis is a chronic inflammatory disease characterized by the growth of endometrial tissue in extrauterine locations. It predominantly affects women of reproductive age, typically between the ages of 25 and 45 1 . The World Health Organization estimates that around 10% of women of reproductive age, approximately 190 million globally, are diagnosed with this condition 2 , 3 . There are three subtypes of endometriosis: peritoneal, ovarian and deep infiltrating endometriosis. Regarding risk factors, some studies have addressed the potential role of autoimmune and digestive diseases in the development of endometriosis 4 – 6 . The symptoms of the disease consist of pelvic pain, dysmenorrhea, dyspareunia, gastrointestinal issues, and infertility, with 40–50% of infertile women being diagnosed with the condition 7 . Typically, the diagnosis of endometriosis is often delayed by 8 to 10 years due to nonspecific symptoms 8 . This can be attributed to the disease’s complex and multifactorial nature, with previously identified contributions from genetic, hormonal, environmental, and immunological factors 9 . Considering the clinical impact of endometriosis, a significant challenge in modern gynecology is understanding the pathophysiology of this pathology and working toward its prevention and early detection. Currently, ultrasound and magnetic resonance imaging represent the main diagnostic methods 10 , 11 . Nonetheless, we are faced in clinical practice with some symptomatic patients even when endometriosis lesions have not yet developed 12 . Therefore, current research efforts aim to diagnose the disease early, seeking non-invasive diagnostic approaches that incorporate clinical indicators with high specificity and sensitivity for identifying endometriosis. For several years, CA-125 was one of the most widely used biomarkers; however, it is no longer used for the diagnosis and follow-up of endometriosis due to its low sensitivity and specificity 13 , as well as to the fluctuation of its blood levels throughout the menstrual cycle 14 . MicroRNAs (miRNAs) have been recognized as potential biomarkers for the early diagnosis of women suffering from endometriosis through liquid biopsy methods. Indeed, a test based on next generation sequencing (NGS) of miRNAs in saliva is being clinically tested 15 . Furthermore, in a previous study, we identified miR-30c-5p as a promising candidate for a minimally invasive biomarker in blood 16 . Nevertheless, the robustness of this test can be impacted by the degree of serum hemolysis 16 . Alternatively, the need for NGS hampers a general clinical application. Therefore, the need for reliable, easy-to-access diagnostic markers for endometriosis is still unmet. Looking for biomarkers, despite the complete pathophysiology of endometriosis remaining unclear, it is known that the immune system plays a critical role in this disease. There is substantial evidence of abnormal immune cell function in women with endometriosis, including reduced T cell reactivity 17 , decreased NK cell cytotoxicity 18 , activation of polyclonal B cells leading to increased antibody production 19 , heightened numbers and activity of peritoneal macrophages 18 , and changes in inflammatory mediators 20 . In this context, elevated levels of soluble factors, including autoantibodies, growth factors, and oxidative stress markers, have been reported in the blood of patients suffering from endometriosis 21 – 23 . However, further studies investigating their associations are required to establish these factors as non-invasive biomarkers for the diagnosis of endometriosis. Additionally, other studies have evaluated the levels of several cytokines, including TGF-β, IL-1β, IL-10, and IL-17A, as well as immune checkpoints such as sCTLA-4, sPD-1, and sPD-L2, in both blood and peritoneal fluid (PF) from patients suffering from endometriosis 24 , 25 . Nevertheless, the analysis of individual immunological factors might limit the capacity to obtain an integral immunological profile of this disease. Furthermore, PF extraction, as an invasive method, is less favorable in clinical settings. Therefore, our work aims to study, based on a blood liquid biopsy, the potential application of determining a large panel of soluble immunological factors, including both cytokines and soluble immune checkpoints, for the efficient diagnosis of endometriosis.