Liquid biopsy based on small extracellular vesicles predicts chemotherapy response of canine multicentric lymphomas

Canine multicentric lymphoma (cMCL) represents approximately 75% of canine lymphomas and shares biological features with human lymphomas, making dogs a valuable comparative model. Despite heterogeneity in subtypes (e.g., DLBCL as most common), most canine cases are treated with CHOP chemotherapy. While initial remission rates are generally high, relapse is frequent and cure is rare, underscoring the need to predict which dogs will not respond to CHOP before initiating therapy. Liquid biopsy analytes such as circulating tumor DNA, extracellular vesicles (exosomes), and miRNAs are promising for monitoring hematologic malignancies. Exosomes (30–150 nm) carry bioactive molecules, including miRNAs, and have been implicated in cancer progression and therapy resistance. Prior canine studies have reported circulating miRNA alterations and exosomal miRNAs associated with drug sensitivity in cell lines. This study tests the hypothesis that serum small extracellular vesicles (SEVs) and their miRNA cargo at diagnosis can predict chemotherapy response and outcome in dogs with multicentric lymphoma.

Background literature highlights that canine lymphoma, especially DLBCL, mirrors many aspects of human disease, including shared pathway alterations (e.g., PI3K/PTEN/NF-κB, MYC/BCL2 double expressors) and that PTCL-NOS has poor prognosis in both species. Standard CHOP therapy achieves high initial remission, but relapse within a year is common. Liquid biopsy biomarkers (ctDNA, exosomes, miRNAs) are being explored in human hematologic cancers. Exosomes are detectable in bodily fluids and can mediate therapy resistance. Prior canine studies: plasma miRNA profiling showed altered miR-127, miR-34a, miR-125b in relapsed dogs versus healthy; exosomal miRNAs (miR-151, miR-8908a-3p, miR-486) and CD82 differed between vincristine-sensitive and -resistant canine lymphoid cell lines. However, there is a lack of validated predictive tests for CHOP response in canine lymphoma, motivating evaluation of SEVs and their miRNAs as predictive markers.

Design and cohorts: Prospective/retrospective observational study with 19 dogs cytologically diagnosed with multicentric, large-cell lymphoma and 30 healthy control dogs. Samples collected January 2017–January 2019; clinical follow-up recorded through February 2020. Inclusion for lymphoma group: no prior cancer diagnosis, treated with a 19-week CHOP protocol (cyclophosphamide, doxorubicin, vincristine, prednisone), no concomitant diseases, and no drugs other than CHOP. Staging per WHO criteria using clinical exam, CBC/chemistry, imaging; PARR/immunohistochemistry when available. Seven of 19 had additional immunophenotyping: 5 DLBCL by IHC and 2 B-cell by PARR. Observation period post-CHOP end: 223–837 days. Ethical approval obtained (CEUA-FZEA-USP/9827200717) with owner consent. Controls: 30 healthy dogs (clinical exam and normal lab work), free of diagnosed diseases. Therapeutic response evaluation: Response per VCOG criteria at week 19 or at last assessment for dogs dying during treatment. Groups defined as complete response (CR) or progressive disease (PD). Survival analyzed via Kaplan–Meier. SEV isolation and characterization: Serum collected at diagnosis (D0) before CHOP. Processing: sequential centrifugations (300×g 10 min, 2,000×g 10 min, 16,500×g 30 min, 4 °C); supernatant filtered (0.20 µm PES) and ultracentrifuged twice at 120,000×g for 70 min to isolate SEVs; pellets resuspended in Ca2+/Mg2+-free PBS. Characterization by TEM (uranyl acetate staining), Western blot for exosomal marker CD9 and absence of mitochondrial marker cytochrome C (spleen tissue as control), and nanoparticle tracking analysis (NTA, NanoSight NS300) for particle size and concentration (5×30 s videos, camera level 14, 37 °C; threshold 5; valid tracks ≤2,500). Size range confirmation targeted 30–150 nm. miRNA (oncomir) profiling: From SEVs of 5 CR and 5 PD cases. RNA extraction from SEV pellets using TRIzol with PolyAcryl carrier, DNase treatment; quality assessed by NanoDrop (A260/A280 1.7–1.9). Reverse transcription with miScript II RT Kit; qPCR with miScript SYBR Green PCR Kit on QuantStudio 6 Flex. Custom primers for 95 canine mature miRNAs (miRBase). Ct positivity defined as Ct<37 in at least three biological replicates with appropriate melt curves. Normalization via geometric mean of miR-99b, Hm/Ms/Rt T1 sRNA, and RNU43 snoRNA. Analyses: frequency of detection per group, identification of group-exclusive miRNAs (Fisher’s exact), and differential abundance (P<0.10). Pathway enrichment for miRNAs associated with CR performed using miRNet (Reactome; hypergeometric test; significance P<0.05). Statistics: Normality by D’Agostino–Pearson; group comparisons by unpaired t-test or Mann–Whitney; contingency by Fisher’s exact; correlations by Pearson or Spearman; survival by Kaplan–Meier and log-rank; predictive value by ROC AUC; multivariate analysis by multiple logistic regression (also reported multiple linear regression R2 in results). Significance at P<0.05; suggestive at P<0.10.

Cohort and clinical outcomes: Of 19 lymphoma dogs, 8 (42%) achieved CR at the end of CHOP; 11 (58%) had PD (non-responders or relapsed before completion). Overall survival differed markedly between CR and PD (log-rank P<0.0001; HR≈8.104 [2.615–25.12]); median survival 573 days (CR) vs 124 days (PD). CR vs PD differed by substage distribution (P=0.0408; substage B more frequent in PD) and age (PD older; P=0.0006). SEV characterization: Serum SEVs from controls and lymphoma patients showed typical morphology by TEM and size 30–150 nm by NTA. Western blot: SEVs positive for CD9, negative for cytochrome C; spleen tissue showed the converse. No significant difference in SEV concentration or size between lymphoma and control groups. SEV concentration predicts response and survival: At diagnosis (D0), PD dogs had higher serum SEV concentration than CR dogs (P=0.034). ROC for predicting CHOP response using SEV concentration yielded AUC=0.8011 (P=0.0287). Dogs that died due to lymphoma had higher SEV concentration at D0 than survivors (P=0.0448); ROC for survival AUC=0.8286 (P=0.0332). Kaplan–Meier using a cutoff of 2.48×10^10 particles/mL showed worse survival for higher SEV concentration: >2.48×10^10 vs <2.48×10^10 particles/mL had median survival 143 vs 461 days, respectively (log-rank P=0.0111; HR≈3.424 [1.067–10.99]). SEV size was not predictive/prognostic (P=0.1233). Correlates and multivariate modeling: Therapeutic response (CR vs PD) inversely correlated with SEV concentration (Spearman r=−0.516, P=0.024), age (r=−0.735, P=3.34×10^−4), and substage (r=−0.567, P=0.011). SEV concentration was not affected by age (P=0.5853), stage (P=0.7532), or substage (P=0.4142). Multivariate model including SEV concentration, age, and substage significantly predicted therapeutic response (R^2=0.7789, P<0.0001); excluding SEV concentration reduced fit (R^2=0.6746, P=0.0001). Exosomal oncomirs: Of 95 screened, 85 miRNAs were detected in at least one sample; 76 in both groups; 7 only in CR (miR-151-5p, miR-190a, miR-200c, miR-204, miR-488, miR-183, miR-205); 2 only in PD (miR-196a, miR-10b). Frequency analysis showed higher occurrence in CR of miR-205 (3/5 vs 0/5, P=0.0384) and a trend for miR-222 (4/5 vs 1/5, P=0.0578). Expression analysis indicated trends: miR-20a more abundant in CR (P=0.085), miR-93 more abundant in PD (P=0.09). Pathway enrichment for CR-associated miRNAs (miR-20a, miR-205, miR-222) included activation of BH3-only proteins (P=0.009), PIP3 activates AKT signaling (P=0.018), and signaling by SCF-KIT (P=0.018).

This study demonstrates that serum SEV concentration at diagnosis is inversely associated with response to CHOP chemotherapy and overall survival in dogs with multicentric lymphoma, addressing the key clinical need to identify non-responders prior to or early in therapy. A single biomarker—SEV concentration—provided significant predictive value (AUC≈0.80), and its integration with clinical variables (age, substage) further improved prediction. Although SEV concentration did not distinguish lymphoma from healthy dogs, the primary goal is predicting therapeutic response and outcome rather than diagnosis, where current clinical methods suffice but may overcall complete responses. The exosomal miRNA profiling suggests a candidate molecular signature associated with favorable response (miR-205, miR-222, miR-20a) and with poor response (miR-93). Enrichment of pathways related to apoptosis (BH3-only), PI3K/AKT signaling, and SCF-KIT signaling aligns with known mechanisms of lymphoma biology and chemoresistance in both canine and human contexts. These findings support the relevance of SEVs as a prognostic/predictive liquid biopsy and provide initial leads for mechanistic and translational exploration of exosomal miRNAs in canine lymphoma.

A liquid biopsy based on serum small extracellular vesicles can predict response to CHOP chemotherapy and overall survival in canine multicentric lymphoma. Higher SEV concentration at diagnosis associates with progressive disease and shorter survival, with an AUC of about 0.80 for predicting response. An exosomal oncomir signature—characterized by presence or higher levels of miR-205, miR-222, and miR-20a in responders and higher miR-93 in non-responders—was identified, implicating apoptosis and PI3K/AKT/SCF-KIT pathways. These results lay the groundwork for integrating SEV quantification and miRNA profiling into clinical decision-making to identify refractory patients early and to inform development of targeted therapies. Future work should validate these biomarkers in larger, well-characterized cohorts with full immunophenotyping/histopathology, assess longitudinal changes during therapy, and develop clinically scalable assays (e.g., ELISA-based exosome quantification) for widespread adoption.

Key limitations include the small sample size (19 lymphoma dogs total; miRNA profiling in 10 dogs), which limits statistical power and generalizability. Not all cases had comprehensive immunophenotyping/histopathology, precluding subgroup analyses by lymphoma subtype. The oncomir findings are preliminary (several associations at suggestive P<0.10) and require validation. Recruitment challenges (older dogs with comorbidities) and field conditions may introduce selection bias. SEV isolation relied on ultracentrifugation, which may not be standardized across laboratories; clinical translation will require simplified, robust assays. Cross-sectional sampling at diagnosis did not evaluate dynamic biomarker changes during treatment.