Abstract Background and Objectives Paraneoplastic neurologic syndromes (PNS) are cancer-related neurologic disorders caused by an autoimmune response targeting both the tumor and the nervous system. Identifying new biomarkers for early cancer detection could improve treatment outcomes. Tumor-derived cells release extracellular vesicles (EVs) carrying tumor-specific molecular signatures, which can help distinguish patients with cancer even in early stages. The aim of this study was to assess the potential of EVs as biomarkers to enhance cancer detection in patients with PNS. Methods This observational and multicenter study included 27 patients with tumor-associated PNS, 26 with suspected PNS without a tumor, 35 with cancer, and 32 healthy controls. Subsequently, among the patients with PNS, individuals were subclassified according to the PNS-Care Score as definite, probable, possible, and non-PNS. Total EVs were isolated from blood by precipitation and from B cells, T cells, and neurons by immunoisolation. To identify a biomarker for diagnostic refinement and clinical stratification, EV levels, size, and protein content were compared across study groups. To find a tumor biomarker in intermediate-risk cases, the possible association between EV protein content and cancer detection in intermediate-risk syndromes was analyzed. Results Patients with tumor-associated PNS showed significantly higher circulating EV levels compared with those with suspected PNS without evidence of a tumor (p = 0.028). Moreover, total EV levels, along with B cell–derived EVs, effectively differentiated patients with definite PNS from those with probable (p = 0.05) and possible (p = 0.006) PNS. A cutoff value of 2.10 × 10^10 particles/mL EVs was identified, above which diagnosis of PNS was definite, with 86% sensitivity and 81% specificity. Proteomic analysis identified specific proteins, including ACADM, HPT, ACTBL, and CCAR2, as markers of definite PNS, differentiating such patients from those with probable and possible PNS, contributing to diagnostic refinement for clinical stratification. It is important to note that increased levels of EVs and CD44 distinguished intermediate-risk cases with tumors from those without. Discussion EVs may act as tumor biomarkers in patients with PNS, even in intermediate-risk cases. Classification of Evidence This study provides Class IV evidence that higher circulating blood levels of EVs can distinguish between tumor-associated PNS from suspected PNS without tumor. Introduction Paraneoplastic neurologic syndromes (PNS) are neurologic disorders that occur in patients with cancer when the immune system activates an autoantibody response against the tumor; however, this response also affects the nervous system, causing collateral damage to the brain, spinal cord, and peripheral nerves.^[58]1 Currently, 3 levels of diagnostic certainty (possible, probable, and definite PNS) have been proposed based on a scoring system (PNS-Care Score) that considers the type of clinical phenotype, presence or absence of neuronal antibodies, and presence or absence of cancer.^[59]2 These classifications reflect the varying levels of risk that the neurologic syndrome is truly paraneoplastic, but a definitive diagnosis can only be confirmed when both the underlying cancer and associated antibodies are detected.^[60]3 However, the neurologic syndrome often precedes tumor detection by up to 5 years, and some patients lack detectable neuronal antibodies. In these patients, early detection of the underlying neoplasm allows for timely oncologic treatment, which is necessary to prevent the tumor from spreading and metastasizing. Moreover, despite treatment, the neurologic syndrome will not improve unless the tumor is addressed. Therefore, the identification of new biomarkers for early cancer detection, and consequently, true PNS diagnosis, would offer room for successful treatment and markedly improve patient prognosis. Recently, scientific attention has turned to extracellular vesicles (EVs) as potential biomarkers in many diseases, including cancer.^[61]4-8 EVs are membrane-bound nanostructures secreted by most cell types that circulate in all biofluids, including blood.^[62]6 EVs are important cargo carriers of cell-derived bioactive molecules such as proteins, nucleic acids, and lipids, which are selectively incorporated. This EV content reflects the characteristics of their originating cells.^[63]7 Specifically in cancer, tumor-derived cells secrete EVs that carry tumor-specific molecular signatures. This specific EV cargo differs from that found in healthy cell–derived EVs, allowing us to distinguish healthy patients from patients with cancer,^[64]8 even in early phases of the disease. Based on this background, EVs could be a promising tool in the field of cancer diagnosis, with significant potential as diagnostic biomarkers. The aim of this study was to evaluate blood EV molecular signatures as a noninvasive and accessible biomarker for early cancer detection in patients with PNS, including those with intermediate-risk conditions, such as possible or probable PNS. This study addresses the following key research questions: * Can EV profiles (size, levels, cellular origin, and proteomic content) help distinguish patients with tumor-associated PNS from those with suspected PNS without evidence of cancer? * Do EV profiles contribute to improved diagnostic stratification of PNS cases classified as definite, probable, possible, or non-PNS? * Can EV levels and specific proteins within EVs serve as biomarkers to detect cancer in patients with intermediate-risk PNS, thereby supporting early tumor diagnosis? Methods Study Design This clinical, observational, multicentric study included patients assigned to 4 groups, one case group and 3 control groups: (1) Patients with tumor-associated PNS: patients aged 18 years or older diagnosed with PNS and tumor detection according to updated diagnostic criteria for PNS.^[65]2 Antibody-positive cases were identified using positivity of both tissue-based immunofluorescence and immunoblot. We excluded patients with clinical neurologic symptoms as a consequence of brain injury; direct invasion by the tumor; tumor metastasis; metabolic or nutritional disorders; infections; toxicity as a consequence of radiotherapy, chemotherapy, or immunotherapy for cancer treatment; current drug or alcohol dependence; coexisting serious illness; pregnancy; or lactation. (2) Patients with suspected PNS without evidence of a tumor (suspected PNS): patients aged 18 years or older with a neurologic syndrome suggestive of paraneoplastic origin who had not developed a tumor at least 2 years after blood extraction. (3) Patients diagnosed with cancer (tumor): patients 18 years of age or older diagnosed with lung or breast cancer without symptoms of neurologic disease. (4) Healthy controls: individuals who voluntarily agreed to participate in the study ([66]Figure 1). Figure 1. Flowchart Showing the Study Groups. [67]Figure 1 [68]Open in a new tab The study included patients with tumor-associated PNS, those with suspected PNS without evidence of tumor, those with tumors without neurologic syndromes, and healthy controls. Patients with PNS were classified based on their risk of being truly paraneoplastic as definite, probable, or possible PNS or as non-PNS. PNS = paraneoplastic neurologic syndrome. In a subsequent study, we subclassified patients from groups 1 and 2 based on their risk of being truly paraneoplastic, considering clinical phenotype, presence or absence of neuronal antibodies, and presence or absence of cancer. We categorized them into the following subgroups based on the PNS-Care Score:^[69]2 definite PNS (≥8 points), probable PNS (6–7 points), possible PNS (4–5 points), or non-PNS (≤3 points) ([70]Figure 1). Last, patients with intermediate risk comprising the probable and possible PNS groups were further divided into 2 subgroups: those with tumors and those without. Standard Approval Protocols, Registrations, and Patient Consents The study was approved by the Research Ethics Committee of La Paz University Hospital (PI-5280), and all patients signed the informed consent. All data management was governed by the principles of Spanish Law 14/2007 of July 3 on Biomedical Research, ensuring the confidentiality of all personal data. We adhered to the relevant Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist. Demographic Data The following demographic and clinical data were collected for the study: sex, age, clinical phenotype, neuronal antibodies, and tumor characteristics ([71]Table1). Table 1. Demographic and Clinical Data Tumor-PNS (n = 27) Suspected PNS (n = 26) Tumor (n = 35) Healthy (n = 32) p Value Data  Women, n (%) 15 (55.55%) 13 (44.4%)* 34 (97.14)* 14 (43.8)* 0.001  Age, y 71.78 (7.45) 59.1(18.26)* 63.8 (10.34) 53.73(12.14)* 0.001 Clinical phenotype  Encephalitis, n (%) 5 (18.5) 16 (61.53) - - 0.051  Polyradiculoneuropathies, n (%) 3 (11.11) 5 (19.23) - - 0.051  Cerebellar syndrome, n (%) 8 (29.6) 1 (3.84) - - —  Lambert-Eaton syndrome, n (%) 2 (7.40) 0 (0) - - —  Retinopathy, n (%) 1 (3.7) 0 (0) - - —  Sensory neuronopathy, n (%) 5 (18.51) 1 (3.84) - - —  Myelitis, n (%) 1 (3.7) 2 (7.69) - - 0.051  Stiff-person syndrome, n (%) 0 (0) 1(3.84)  Brainstem encephalitis, n (%) 2 (7.40) 0 (0) - - — Autoantibodies  Anti-Hu, n (%) 6 (25) 0 (0) — — —  Anti-Yo, n (%) 7 (29.16) 1 (7.14) — — 0.020  Anti-titin, n (%) 1 (4.16) 0 (0) — — —  Anti-CV2, n (%) 3 (12.5) 0 (0) — — —  Anti-ZIC4, n (%) 1 (4.16) 0 (0) — — —  Anti-recoverin, n (%) 1 (4.16) 0 (0) — — —  Anti-SOX1 2 (8.32) 0 (0) — — —  Anti-NMDA-R 0 (0) 9 (64.28) — — —  Anti-LGI1 0 (0) 3 (21.42) — — —  Anti-Ri 1 (4.16) 0 (0) — — —  Anti-amphiphysin 1 (4.16) 0 (0)  Anti-GAD65 0 (0) 1 (7.14) Tumor type  NSCLC, n (%) 2 (7.40) — 0 (0) — —  SCLC, n (%) 13 (48.1) — 1 (2.85) — 0.001  Breast adenocarcinoma, n (%) 5 (18.51) — 34 (97.14) — 0.001  Colon adenocarcinoma, n (%) 3 (11.1) 0 (0)  Ovary adenocarcinoma, n (%) 3 (11.1) — 0 (0) — —  Melanoma, n (%) 1 (3.7) 0 (0) [72]Open in a new tab Abbreviations: PNS = paraneoplastic neurologic syndrome; LGL1 = leucine-rich glioma inactivated 1; n; number; RNMDA = anti-N-methyl-d-aspartate receptor; Sox1 = sex-determining region Y-box 1; ZIC4 = zinc finger protein of the cerebellum 4. *Indicates groups with significant differences. Sample Collection and Processing A total of 7 mL peripheral blood was collected from each participant in 9-mL EDTA tubes at diagnosis and immediately centrifuged at 3,000 × g for 15 minutes at 4°C. Plasma was collected in tubes and stored at a temperature of −80°C until further analysis. Extracellular Vesicle Isolation This research focused on the total blood-circulating EVs and further on those derived from key cells involved in the pathogenesis of the disease: B cells, T cells, and neurons. Neuron-derived EVs reflect neuronal injury due to immune attacks occurring in PNS. B-cell EVs offer insights into humoral immune responses, indicating pathologic B-cell activation. T-cell EVs reveal cellular immune dysregulation and T cell–mediated cytotoxicity. Collectively, these EV populations provide complementary perspectives on the immune-neuronal interaction. For total blood-circulating EV isolation, a commercially available high-throughput particle precipitation method known as the ExoQuick EV Isolation Kit (System Biosciences, USA) was used as previously described.^[73]9 For neuronal, T cell–associated, and B cell–associated EV subpopulation isolation, we used a 2-step process including precipitation followed by immunoprecipitation using biotinylated antibodies against specific EV surface markers such as the anti-L1CAM antibody (Thermo Fisher Scientific, USA) for those of neuronal origin, the anti-CD20 antibody (Merck Millipore, Germany) to capture B cell–derived EVs, and the anti-CD3 antibody (Merck Millipore, Germany) to capture T cell–derived EVs.^[74]9 Characterization of Purified Extracellular Vesicles EVs were characterized by the following: transmission electron microscopy (TEM) (JEOL JEM1010), which allows visualization of EV morphology and precise size measurement, in accordance with the methodology previously described^[75]9; nanoparticle tracking analysis (NTA) to assess EV size distribution and quantification using NanoSight NS500 (Malvern Instruments, United Kingdom); and Western blot to analyze vesicle-specific surface tetraspanin proteins with anti-Alix (1:250, Cell Signal, USA), anti-CD81 (1:250, Abcam, United Kingdom), and anti-CD63 (1:250, Abcam) antibodies ([76]Figure 2, A-D). Figure 2. Characterization of Extracellular Vesicles. [77]Figure 2 [78]Open in a new tab (A and B) Characterization of EVs by size dispersion and shape of EV samples analyzed by NTA. C) Electron microscope image of an EV smaller than 200 nm. D) Western blot image demonstrating the positivity of EV-specific markers CD9, CD81, and Alix in the EV membrane. Negative control samples are from EV-depleted plasma. The gel image was cropped. EVs = extracellular vesicles; PNS = paraneoplastic neurologic syndrome; kDa = kiloDalton; ml = mililiters; nm = nanometers. Extracellular Vesicle Quantification and Size Assessment We analyzed the levels and size of total circulating, neuronal, B cell–derived, and T cell–derived EVs and compared them across groups using the NanoSight NS500 nanoparticle analyzer (Malvern Instruments, United Kingdom). Small (<200 nm) and large (>200 nm) EVs were selected for the NTA. Three 60-second videos were recorded at a detection threshold of 3. This process was run in triplicate. Proteomic Analysis The protein content of the vesicles was analyzed and compared across groups using the “label-free” sequential window acquisition of all theoretical mass spectra (SWATH-MS) method. Three sample pools were prepared from each condition to obtain biological replicates, ensuring a relevant biological representation. Protein identification of each sample was performed in a 2-step analysis with a combination of liquid chromatography and mass spectrometry, followed by a data-dependent acquisition method in a TripleTOF 6600 system. Data were processed using ProteinPilot 5.0.1 and MarkerView 1.3.1 software from Sciex (Sciex, Canada), as previously described.^[79]10 To minimize batch effects in SWATH-MS, we first applied retention time correction in PeakView, aligning peptides using endogenous signals for consistent peptide area comparisons. PeakView extracted intensities from up to 10 peptides per protein and 7 fragment ions per peptide. MarkerView was used to assess missing values—minimal because of pooled libraries—and apply most likely ratio (MLR) normalization. MLR, which assumes that most proteins remain constant across samples, aligns ratio histograms to a centered reference, reducing technical variance and improving quantification. It performs best with ≥500 transitions and normally distributed data. MLR is applied first within replicates (yielding normalized peak areas and reproducibility scores) and then across groups to standardize scales. SCIEX recognizes MLR as a robust method for reducing intersample variability and improving downstream reliability. Biomarker Analysis We analyzed the levels, sizes, and protein content of total circulating, neuronal, B cell–derived, and T cell–derived EVs in blood to identify tumor biomarkers by comparing patients with tumor-associated PNS, patients with suspected PNS without evidence of tumor, patients with tumors without neurologic syndromes, and healthy controls. To refine diagnostic biomarkers for clinical stratification, we conducted a follow-up study on patients with neurologic syndromes, comparing EV levels, size, and protein content across definite, probable, possible, and non-PNS cases. To identify a tumor biomarker for intermediate-risk cases, we performed a substudy among patients classified as probable or possible PNS, comparing those with tumors with those without. The individual who performed the analysis of the EV levels, size, cellular origin, and protein signatures was blinded to the final clinical diagnosis of the patients, ensuring objectivity and minimizing bias measurement. The interpretation of the EV profile and the categorization of patients as definite, probable, possible, or non-PNS were conducted by different evaluators working independently, thus reducing the risk of interpretation bias. Biological Functions and Pathway Study To elucidate possible mechanisms governing the manifestation of PNS as a consequence of tumor, we explored the potential biological functions in which the proteins identified in the various study groups are involved. To this end, we used the Reactome pathway database.^[80]11 Statistical Analysis The statistical analysis was supervised by the Biostatistics Unit of La Paz University Hospital using SPSS 23.0 (IBM, USA). Categorical variables were described as percentages, and proportions between groups were compared using the χ^2 test. The Fisher exact test was used for dichotomous variables. Continuous variables were expressed as mean (SD). A t test and analysis of variance with Bonferroni post hoc correction were used for multiple comparisons of normally distributed data. Kruskal-Wallis or Mann-Whitney U tests were applied for the comparison of non-normally distributed data sets. Using a receiver operating characteristic analysis, a Youden index corresponding to the maximum sensitivity and specificity values was calculated to identify a score for tumor presence in patients with PNS. Data Availability The mass spectrometry proteomics data have been deposited with the ProteomeXchange Consortium via the PRIDE partner repository with data set identifier PXD038003. Results The study included 27 patients with tumor-associated PNS, 26 patients with suspected PNS but without tumor, 35 patients with cancer without neurologic syndrome, and 32 healthy controls. Patient characteristics are summarized in [81]Table 1. In a further study, patients were subclassified according to the PNS-Care Score: definite PNS (n = 21), probable PNS (n = 6), possible PNS (n = 11), and non-PNS (n = 15) ([82]Figure 1). Extracellular Vesicle Characterization EVs were characterized by size, morphology, and presence of specific EV marker profiles in their membrane. The EV samples showed the typical morphology of lipid bilayer spheres, with a size less than 300 nm by NTA ([83]Figure 2, A and 2B) and 200 nm by TEM ([84]Figure 2C). Triple positivity for the EV-specific tetraspanins CD9, CD81, and Alix allowed for a robust characterization of the EV samples by Western blot ([85]Figure 2D). Levels and Size of Circulating Extracellular Vesicle Reflect Tumors in Patients With PNS Patients with tumor-associated PNS showed significantly higher levels of total blood-circulating EVs (3.918 × 10^10 ± 1.94 × 10^10 particles/mL) compared with those with suspected PNS (1.98 × 10^10 ± 1.92 × 10^10 particles/mL) ([86]Figure 3A), identifying a cutoff of 2.10 × 10^10 EVs, with 80% sensitivity and 87% specificity (Figure 4A). Figure 3. Comparison of EV Levels and Size Across Cohorts of Tumor-Associated PNS, Suspected PNS Without Tumor, Tumor Without Neurologic Syndrome, and Healthy Controls. [87]Figure 3 [88]Open in a new tab (A) Levels of total circulating EVs and those derived from B cells, neurons, and T cells. (B) Size of total small EVs and small EVs across those derived from B cells, neurons, and T cells. (C) Size of total large EVs and large EVs across those derived from B cells, neurons, and T cells. EVs = extracellular vesicles; PNS = paraneoplastic neurologic syndrome. A significant increase in the size of small EVs was found in patients with tumors (128 ± 23 nm) compared with those with tumor-associated PNS (97 ± 48 nm; p = 0.002) ([89]Figure 3B). Patients with tumor-associated PNS showed a smaller size of small B cell–derived EVs (117 ± 40 nm) than patients with suspected PNS (151 ± 42 nm; p = 0.006) and a larger size of small and large B cell–derived EVs (117 ± 40 and 366 ± 87 nm, respectively) than patients with tumors (102 ± 23 and 262 ± 61 nm, respectively; p = 0.05) ([90]Figure 3, B and C). Patients with tumor-associated PNS exhibited higher levels of neuron-derived EVs (6.01 × 10^9 ± 6.4 × 10^9 particles/mL) than healthy controls (1.17 × 10^9 ± 4.38 × 10^9 particles/mL; p = 0.001). Significant differences in the size of large neuron-derived EVs were found between patients with tumor-associated PNS (332.93 ± 94.86 nm) and healthy controls (253 ± 32 nm; p = 0.003) ([91]Figure 3C). In addition, we did not observe associations between age, sex, and clinical phenotype of the patients with EV levels (eTable 1), suggesting that the specificity of EV signatures is related to tumor association rather than patient characteristics. Levels and size of normal EV in healthy controls are presented in eTable 2. Tumor-Specific Molecular Signatures in EVs Distinguished Patients With Tumor-Associated PNS On proteomic analysis of the vesicle content, we identified a total of 755 proteins in circulating EVs common to all study groups. The differentially expressed proteins found in patients with tumor-associated PNS are presented in the volcano plots of [92]Figure 4B. Figure 4. Proteomic Analysis of EVs. [93]Figure 4 [94]Open in a new tab (A) Receiver operating characteristic curve showing a cutoff point of 2.10 × 10^10, above which PNS is tumor-associated PNS, with 80% sensitivity and 87% specificity. Volcano plots representing protein abundance in patients with tumor-associated PNS compared with patients with suspected PNS, those with tumor, and healthy controls. (C ) Tumor-specific proteins and (D) PNS-specific proteins. ACADM = Acyl-CoA dehydrogenase medium chain; CXCL7 = chemokine (C-X-C motif) ligand 7; EVs = extracellular vesicles; HEMO = hemopexin; IMB1 = karyopherin subunit beta 1; PGRC2 = progesterone receptor membrane component 2; PNS = paraneoplastic neurologic syndrome; ZYX = zyxin. ZYX (zyxin) and ACADM (acyl-CoA dehydrogenase, medium-chain) proteins were defined as tumor-specific markers, given that they were differentially expressed in patients with tumor-associated PNS (1,982.494 ± 793.4662 NSAF and 58,567.98 ± 10,239.14 NSAF, respectively) and in patients with tumors (2,863.071 ± 1499 NSAF and 50,431.86 ± 17,665 NSAF) compared with the rest of the study groups ([95]Figure 4C). EV Content Differentiated Patients With Neurologic Syndromes IMB1 (importin subunit beta-1), HEMO (hemopexin), CXCL7 (C-X-C motif chemokine ligand 7), and PGRC2 (membrane-associated progesterone receptor component 2) proteins were identified as neurologic syndrome–specific biomarkers, regardless of tumor presence: tumor-associated PNS—IMB1 (106,125.8 ± 21,153.37 NSAF), HEMO (4,155,214 ± 1,163,027 NSAF), CXCL7 (139,467.2 ± 41,109.37 NSAF), and PGRC2 (150,375.3 ± 9,626.952 NSAF); suspected PNS—IMB1 (103,343.9 ± 15,404.32 NSAF), HEMO (2,828,013 ± 951,698 NSAF), CXCL7 (73,368.02 ± 14,490.79 NSAF), and PGRC2 (115,420.8 ± 34,712.26 NSAF) (eFigure 1). Circulating EVs and B Cell–Derived EVs Reflect the Levels of Evidence for PNS: Definite, Probable, and Possible Cases Patients with definite PNS exhibited significantly higher levels of total circulating EVs in the blood (16.34 × 10^10 ± 14.83 × 10^10 particles/mL) compared with those with possible PNS (8.43 × 10^9 ± 6.95 × 10^9 particles/mL; p = 0.006) (eFigure 1A). Both groups of definite and possible PNS were found to have significantly higher EV levels than non-PNS patients (1.68 × 10^10 ± 2.63 × 10^10 particles/mL; p = 0.017). Levels of B cell–derived EVs were significantly lower in patients with definite PNS compared with those with probable PNS (eFigure 1C). No significant differences were observed in EVs derived from neurons or T cells across the study groups. A cutoff value of 2.10 × 10^10 particles/mL EVs was identified above which PNS is definite, with 86% sensitivity and 81% specificity (eFigure 1E). Increased Levels of ACADM, HPT, ACTBL, and CCAR2 Distinguish Definite PNS From Probable, Possible, and Non-PNS Cases On proteomic analysis of the vesicle content in patients with definite, probable, and possible PNS and non-PNS patients, we identified a total of 914 proteins in circulating EVs common to all study groups. The differentially expressed proteins found in definite PNS are presented in the volcano plots of eFigure 1F. Significantly higher levels of ACADM, HPT (haptoglobin), ACTBL (beta-actin-like protein), and CCAR2 (cell cycle and apoptosis regulator 2) were observed in patients with definite PNS compared with probable, possible, or non-PNS groups. For ACADM, the levels in the definite PNS group were (47,080.38 ± 26,877.14 NSAF) compared with (12,943.94 ± 22,419.56 NSAF), (10,845 ± 987 NSAF), and (3,173.74 ± 1993.51 NSAF) in the probable, possible, and non-PNS groups, respectively (p = 0.016; p = 0.03; p = 0.049). Similarly, HPT levels were significantly higher in the definite PNS group (19,367,568.14 ± 5,912,115.15 NSAF) relative to the probable PNS (4,736,143.39 ± 2,387,927.93 NSAF), possible PNS (1,506,768.56 ± 1,043,859.478 NSAF), and non-PNS (158,629.5 ± 33,340.25 NSAF) groups (p = 0.016; p = 0.006; p = 0.0002). ACTBL levels in patients with definite PNS (137,163.66 ± 73,060.60) were also higher than in probable PNS (11,709.99 ± 20,282.29 NSAF), possible PNS (10,949.64 ± 18,965.34 NSAF), and non-PNS (46,976.99 ± 29,311.40 NSAF) groups (p = 0.017; p = 0.017; p = 0.009; respectively). The same trend was observed for CCAR2, with significantly higher levels in the definite PNS group (88,859.77 ± 29,227.34 NSAF) compared with probable PNS (3,568.18 ± 6180.28 NSAF), possible PNS (3,009.37 ± 5,212.38 NSAF), and non-PNS (18,816.47 ± 24,717.52 NSAF) groups (p = 0.007; p = 0.007; p = 0.002; respectively) ([96]Figure 5). Figure 5. Proteins Distinguishing Definite, Probable, Possible, and Non-PNS Cases. [97]Figure 5 [98]Open in a new tab Proteins such as ACADM, HPT, ACTBL, and CCAR2 show differential expression across these groups, highlighting their potential as biomarkers for stratifying PNS diagnosis. ACADM = Acyl-CoA dehydrogenase medium chain; ACTBP = beta-actin-like protein = CCAR2 = cell cycle and apoptosis regulator 2; EVs = extracellular vesicles; HPT = haptoglobin; PNS = paraneoplastic neurologic syndrome. ACADM, ZYX, CXCL7, IBM, and HEMO are proteins involved in various biological processes and may lack specificity for paraneoplastic conditions. To evaluate their specificity in PNS, we analyzed their expression in 3 additional control groups: 41 patients with rheumatoid arthritis (a non-neurologic autoimmune disease), 21 patients with ischemic stroke (a nonimmune-mediated neurologic disorder), and 38 patients with MS (an immune-mediated neurologic disease not associated with paraneoplastic syndromes). We found that these proteins were overexpressed in the groups of the study compared with these additional controls, suggesting a specific association with paraneoplastic conditions rather than with general neuroinflammation or autoimmunity (eFigure 2). Increased Levels of EVs and CD44 Distinguish Intermediate-Risk Cases With Tumors From Those Without Patients with an intermediate risk of PNS with tumors were found to have higher levels of total circulating EVs (1.75 × 10^11 ± 2.30 × 10^11 particles/mL) compared with those with intermediate risk of PNS without evidence of tumor (1.073 × 10^10 ± 6.28 × 10^9 particles/mL; p = 0.049). Moreover, significantly higher levels of CD44 were observed in patients with an intermediate risk with tumors (75,776.75 ± 26,659.86 NSAF) compared with those without (22,371.01 ± 11,885; p = 0.03) ([99]Figure 6, A–C). Figure 6. Extracellular Vesicle Levels and Proteins Indicative of Tumors in Patients at Intermediate Risk of Being Truly Paraneoplastic and Biological Functions Associated With the Dysregulated Proteins in PNS. [100]Figure 6 [101]Open in a new tab (A) Levels of EVs in patients with intermediate-risk PNS with tumor vs without tumor. (B) Volcano plot representing CD44 abundance/expression in patients with intermediate-risk PNS associated with tumor vs those with intermediate-risk PNS without tumor. (C) CD44 proteomic analysis in patients with intermediate risk with tumor vs those with intermediate-risk PNS without tumor. (D) Pathway enrichment analysis using Reactome databases of 71 differentially upregulated proteins from patients with tumor-associated PNS (in red), 54 proteins of patients with cancer (in green), 47 proteins identified in the possible PNS group (in yellow), 48 proteins found in the probable PNS group (in purple), and 27 proteins from patients in the non-PNS group (in blue). EVs = extracellular vesicles; PNS = paraneoplastic neurologic syndrome; NF-kappaB = nuclear factor kappa B. To further explore the diagnostic utility of EV profiling in the intermediate-risk PNS group, we included an additional validation cohort of 10 additional patients with probable PNS. Among them, 5 had a confirmed tumor and 5 had no identifiable oncologic association. When applying the EV-based model to this external cohort, patients with tumors exhibited significantly higher levels of total circulating EVs (3.31 × 10^11 ± 3.12 × 10^11 particles/mL) compared with those without tumors (1.52 × 10^10 ± 1.26 × 10^10 particles/mL) (p = 0.021). All tumor-associated patients exceeded the predefined cutoff of 2.10 × 10^10 particles/mL. These findings validate the predictive performance of the model in an independent group of patients with probable PNS. In addition, neuron-derived EVs were significantly higher in the tumor group (353.2 ± 34.19 vs 278 ± 32.31 nm; p = 0.016) (eTable 3). Differentially Expressed Proteins in Patients With Tumor-Associated PNS Were Enriched for Specific Biological Functions [102]Figure 6D illustrates the proteins specifically identified in each study group and the biological functions in which they are involved. The 71 differentially expressed proteins identified in the group of patients with tumor-associated PNS compared with the rest of the groups are involved in the following functions: phagocytosis, interleukin (IL)-4 and IL-13 signaling, complement activation, activation of the MAPK oncogenic pathway, DNA synthesis, and axonal growth, among others. The 54 proteins found in patients with only tumors participate in the activation of nuclear factor kappa B in B lymphocytes, regulation of RUNX1, variation of the Hedgehog pathway, and apoptosis regulation, among other processes. The 47 proteins identified in the possible PNS group are involved in the detoxification of reactive oxygen species, among other functions, while the 48 proteins found in the probable PNS group are implicated in the regulation of small RNAs. Finally, the 27 differentially expressed proteins found in non-PNS patients participate in SEMA4D-induced cell migration, mitochondrial beta-oxidation, and muscle contraction ([103]Figure 6D). Classification of Evidence This study provides Class IV evidence that higher levels of blood-circulating EVs can distinguish tumor-associated PNS from suspected PNS without tumor. Discussion Since the term “extracellular vesicle” was first used in 1971, many studies have focused on identifying the functional roles of EVs in disease and enhancing their clinical benefits. In an attempt to identify blood-derived EVs as clinical biomarkers in PNS, this study sought to investigate the possible associations between EVs and cancer. To this end, we analyzed the levels of EVs as critical indicators of cellular activity that may reflect underlying pathologic processes, providing valuable diagnostic and prognostic insights and the sizes of EVs. In addition, we explored the proteomic content of EVs, which can provide a deeper understanding of the pathogenic mechanisms underlying PNS, potentially supporting clinicians in the diagnosis of patients with intermediate-risk neurologic syndromes or cases where PNS is suspected but cannot be confirmed by current diagnostic methods. In our study, patients with tumor-associated PNS exhibited significantly higher levels of total blood-circulating EVs compared with patients with suspected PNS without evidence of a tumor. This parameter had previously been investigated in patients diagnosed with breast cancer^[104]12,13 and small-cell lung cancer^[105]14 (without accompanying neurologic syndromes), revealing elevated levels of circulating EVs compared with healthy controls, probably as a result of increased tumor cellularity. Consistent with these findings, patients in our study with definite PNS were found to have significantly higher levels of total circulating EVs compared with those with possible PNS. Subsequently, patients with higher levels of B cell–derived EVs in comparison with those with probable PNS were further studied, aiding in the differentiation of truly paraneoplastic cases. A cutoff value of 2.10 × 10^10 particles/mL EVs was identified, above which diagnosis of PNS is definite, with 86% sensitivity and 81% specificity. Moreover, among patients with probable or possible PNS, where antibody profiles or clinical syndromes suggest an intermediate risk of being truly paraneoplastic, EV analysis revealed higher levels of EVs in those with underlying malignancy compared with those without a tumor, helping address an important gap in current clinical practice. The identification of a specific cutoff value also aligns with previous studies in patients with cancer (without associated neurologic syndromes), determining a level of circulating EVs that reflects the presence of neoplasm.^[106]14 Establishing an EV level cutoff with reasonable sensitivity and specificity for patients with tumors could be valuable in differentiating PNS from idiopathic cases or those arising from other causes. This information may assist clinicians in diagnostic and treatment planning. For example, elevated EV levels may indicate a greater likelihood of a hidden tumor, prompting earlier and more targeted cancer screening, such as whole-body PET scans, because such individuals may have a higher probability of an undetected malignancy. In addition, these patients might benefit from more frequent and longer cancer screening after the acute phase, addressing a current area of debate among neuroimmunology experts.^[107]15 The strong association between elevated EV levels and tumor presence enhances diagnostic certainty, even in cases where antibody findings are inconclusive. This supports the timely initiation of treatment strategies tailored to paraneoplastic etiologies. From a therapeutic perspective, individuals with levels above the cutoff may respond better to treatments targeting T cell–mediated autoimmunity, such as corticosteroids and cyclophosphamide, commonly used in managing paraneoplastic syndromes in conjunction with tumor removal. Last, in non-PNS cases, EV levels below the cutoff point could provide additional reassurance against a paraneoplastic cause, reducing unnecessary diagnostic procedures and allowing clinicians to focus on alternative diagnoses. Another goal in this study involved analyzing the content of EVs in PNS, given that this could provide insights into the molecular composition and communication networks of PNS-associated cancer cells. Initially, our research demonstrated that patients with cancer exhibit different levels of the proteins ZYX and ACADM compared with those with suspected PNS and with healthy controls. In a subsequent study, we successfully validated ACADM as a biomarker to differentiate definite PNS from probable, possible, and non-PNS cases. This enzyme plays an important role in the mitochondrial fatty acid beta-oxidation pathway, essential for the metabolism of medium-chain fatty acids. Although ACADM is primarily known for this role in metabolic disorders, there is emerging evidence of a potential relationship between ACADM and cancer.^[108]16 Previous studies have shown that ACADM expression levels can be altered in various types of cancer,^[109]17,18 influencing energy production and the metabolic state of cancer cells. Dysregulation of ACADM may lead to altered energy homeostasis, which can induce metabolic reprogramming to support rapid growth and proliferation, promoting the survival and growth of cancer cells, highlighting its potential role as a biomarker for cancer progression.^[110]18 Although further research is needed to fully elucidate the mechanisms connecting ACADM to cancer, it shows significant promise as a biomarker for identifying tumor presence in patients with definite PNS and distinguishing them from those at intermediate risk. Other proteins included in the biomarker panel that can distinguish definite PNS from other PNS categories in this study are HPT, ACTBL, and CCAR2. These proteins have been previously implicated in cancer-related processes. Elevated serum HPT has been identified as a prognostic marker in various solid tumors and is correlated with poor prognosis.^[111]19 ACTBL is typically upregulated in most tumor cells and tissues, with abnormal expression and polymerization affecting the cytoskeleton, contributing to cancer invasiveness and metastasis.^[112]20 CCAR2, a regulator of apoptosis and the cell cycle, has been suggested as a tumor promoter, with its upregulation in patients with cancer often associated with poor prognosis. Its depletion has been shown to reduce cancer cell growth in vitro.^[113]21 The findings in our study underscore the potential of these biomarkers in distinguishing patients with definite PNS from those with intermediate-risk PNS. In addition, CD44, a cell matrix adhesion protein essential to cancer invasion and metastasis,^[114]22 can differentiate between patients with tumor and those without in intermediate-risk cases. While high-risk antibodies are strongly predictive of cancer association, EV analysis could be particularly helpful in patients with intermediate-risk antibodies or neurologic syndromes, or when a tumor is suspected but not yet confirmed. EV profiling bridges a critical diagnostic gap by distinguishing cases of definite PNS from those of other causes and by making earlier tumor detection possible in patients where imaging alone is inconclusive. The analysis of EV protein content can also shed light on the biological processes by which tumors influence distant organs via the immune system and the mechanisms governing the manifestation of neurologic syndrome. Autoimmune neurologic diseases begin when antigen-presenting cells present nervous system proteins to T cells in lymph nodes, triggering an autoreactive response.^[115]23 IMB1, a protein involved in nuclear-cytoplasmic transport,^[116]24 may affect EV composition and disrupt immune tolerance mechanisms, potentially contributing to the initiation of autoimmunity.^[117]25 Our study found IMB1 overexpressed in EVs from patients with immune-mediated diseases, suggesting its role in fundamental autoimmune mechanisms. After activation in lymph nodes, autoreactive T cells cross the blood-brain barrier, where they interact with local antigen-presenting cells and release inflammatory mediators, recruiting additional immune cells.^[118]23 CXCL7, a chemokine enriched in EVs of patients with neurologic syndromes, promotes immune cell recruitment and chemotaxis, linking its dysregulation to prolonged inflammation process.^[119]24 In agreement with our results, dysregulated levels of CXCL7 have been previously noted across various autoimmune diseases. This result is in line with the findings from the functional enrichment analysis, which indicated enrichment of the complement activation pathway, promoting chemotaxis at sites of inflammation^[120]26 and prolonging autoimmune response. Once within the nervous system, the aberrant T-cell response can become perpetuated. The T cells differentiate to a proinflammatory Th17 cell lineage,^[121]27 perpetuating T-cell responses in the nervous system and contributing to chronic inflammation. HEMO, a glycoprotein overexpressed in patients with neurologic syndrome, regulates Th17 activity and may reflect efforts to modulate excessive immune responses.^[122]28 Within the scenario of neuroinflammation-related neurologic injury, the nervous system initiates intrinsic natural repair mechanisms aiming to mitigate the damage caused by the immune system. Within this framework, PGRC2 supports neural progenitor proliferation and synapse remodeling and also inhibits inflammatory gene expression and promotes cell survival.^[123]29,30 This finding aligns with the results from the functional enrichment analysis of the protein content, which indicated that axonal growth is enhanced. These EV proteins likely play dual roles in mediating immune responses and facilitating nervous system repair in immune-mediated neurologic diseases. In this study, we preferred to analyze the proteome of EVs rather than serum or plasma because the lipid bilayer of EVs safeguards their protein cargo from enzymatic degradation, preserving biomarker integrity. This allows for detection of low-abundance or fragile proteins that are often undetectable in serum because of degradation, dilution, or high-abundant proteins. EVs originate from specific cell types, reflecting their cellular origin and state, thus providing more precise and biologically relevant information. By contrast, free serum proteins represent a complex mixture of systemic signals, complicating biomarker attribution. In fact, several tumor-associated proteins were identified, including ANXA, STAT1, PARK7, DDX5, NPM, ZYX, and ACADAM, which are known to be overexpressed in various cancers,^[124]31,32 although none of them are recognized as classical paraneoplastic antibody targets. Future studies will be needed to determine whether these EV-associated tumor proteins can themselves become immunogenic and trigger antibody responses in patients with PNS. Several potential confounding factors could influence EV levels, including age, sex, clinical phenotype, therapies, comorbidities, and lifestyle factors such as smoking. In our analysis, we found no significant effects of age, sex, or clinical phenotype on EVs. This finding supports the hypothesis that the EVs we characterized function as specific biomarkers of tumor pathology itself. It is important to note that patients were not undergoing active therapy at the time of sampling. However, we were unable to perform a comparable analysis for the protein content of the EVs. This limitation stems from the fact that proteomic analyses were conducted using technical triplicates, which presents a considerable challenge, given the large number of variables involved. Conducting proteomic assays in triplicate for each individual sample would be logistically impractical and prohibitively resource-intensive. In conclusion, our study highlights the potential of EVs as valuable biomarkers for tumor diagnosis in patients with suspected PNS, even in those at intermediate risk. The identification of specific EV cutoff values not only assists clinicians in distinguishing definite PNS from those arsing from other causes but also enhances the detection of tumors in patients at intermediate risk, ultimately facilitating earlier diagnosis, targeted cancer screening, more personalized immunomodulatory treatment strategies, and timely oncologic interventions to prevent tumor progression and metastasis. Acknowledgment