The study emerged from a confluence of two research currents. Cancer remains the leading cause of death in dogs, yet veterinary oncology has lagged behind its human counterpart, particularly in immunotherapy. Immune checkpoint inhibitors, which revolutionized human cancer care, have only just begun entering canine clinical trials. In parallel, however, other immunotherapeutic approaches have advanced in dogs over the past decade. Among them is a vaccine designed to stimulate immunity against EGFR and HER2, two cell-surface proteins frequently exploited by tumors. Earlier work showed that this vaccine could provoke anti-tumor activity in both T cells and B cells and reduce metastatic burden in canine patients. Meanwhile, in human oncology, a growing body of evidence has tied gut microbiome composition to immunotherapy outcomes, spurring clinical trials of fecal microbiota transplants as adjuncts to checkpoint blockade. What remained entirely unexplored was whether the same microbiome-survival relationship exists in dogs receiving anti-cancer immunization.

To find out, a collaboration spanning the National Cancer Institute, Oregon State University, Yale University, and the Bridge Animal Referral Center in Edmonds, Washington, enrolled 51 canine cancer patients in a clinical trial of the EGFR/HER2 peptide vaccine. The cohort included 28 females and 34 purebred dogs, with a mean age at diagnosis of 9.5 years. The most common diagnosis was osteosarcoma, a malignant bone tumor affecting 22 dogs, followed by hemangiosarcoma, an aggressive cancer of blood-forming vessels, in 8 dogs; the remaining 21 animals carried a mix of other tumor types. On the day each dog received its first vaccine dose, the team collected rectal swabs to capture the baseline gut microbiome—the microbial community as it existed before any vaccine-driven immune changes could take hold. All animals also continued to receive standard veterinary care, and the patients were scheduled to receive two vaccine injections three weeks apart.

The laboratory work relied on 16S rRNA gene sequencing, a standard technique that reads a conserved bacterial gene to identify which microbes are present and in what proportions. DNA extracted from the swabs was amplified and barcoded at the National Cancer Institute’s Microbiome and Genetics Core Facility and sequenced on an Illumina MiSeq platform, yielding an average of roughly 32,000 reads per sample and a total of 899 amplicon sequence variants—the fine-grained units used to distinguish bacterial populations. Bioinformatic pipelines built on DADA2 and QIIME2 processed the reads, assigned taxonomy against the SILVA reference database, and predicted functional potential using PICRUSt2. The team then applied survival analysis, including Kaplan-Meier curves and Cox proportional hazards regression, to ask a single pointed question: did the abundance of any bacterial taxon at baseline correlate with how long each dog lived?

The first results were, in a sense, reassuringly unremarkable. Neither alpha diversity—the richness and evenness of microbes within a sample—nor beta diversity, which measures compositional differences between samples, varied significantly by cancer type, sex, or breed. No specific taxa or predicted functional pathways distinguished one diagnostic group from another. Within the limits of the sample size, the canine gut microbiome did not appear to encode a signature of tumor type. But that null result carried an important implication: it suggested the study population carried no inherent microbiome bias that could confound the search for survival-related microbes. The microbiome differences the team eventually found would therefore be less likely to reflect pre-existing differences between cancer groups.

Clinical covariates told a familiar story. Dogs with hemangiosarcoma and dogs older than 11.9 years at diagnosis had significantly poorer survival, consistent with prior veterinary reports, while breed made no difference. Neither recent antibiotic use, chemotherapy, nor surgery predicted outcome. Armed with these confounders, the researchers ran Cox regression separately within the osteosarcoma and other-cancer subgroups, deliberately excluding hemangiosarcoma because its grim prognosis and small sample size could distort the analysis. Under a relaxed false discovery rate threshold, they identified 57 survival-associated taxa among osteosarcoma patients and 46 among the other-cancer group. Crucially, 15 taxa appeared in both groups, and 11 of those—73 percent—pointed in the same direction in each, with matching hazard ratios. That convergence across biologically distinct tumors hinted at shared microbial signatures tied to survival rather than to any single malignancy.

When the team pooled all 51 dogs, including the hemangiosarcoma cases, the signal strengthened rather than dissolving. All 11 taxa remained significantly associated with survival under the stricter criterion of a false discovery rate below 0.1. Seven taxa, when abundant, were linked to an elevated risk of death, while four were associated with prolonged survival. Kaplan-Meier curves built on uniform low-versus-high abundance thresholds made the separation visible across the entire cohort. The associations also withstood the study’s most stringent test: after statistically adjusting for hemangiosarcoma diagnosis and advanced age—the two strongest predictors of poor outcome—every one of the 11 microbes retained its significant relationship with survival, confirming the robustness of the microbial signal in the face of powerful clinical confounders.

Which microbes made the list? While the full taxonomic details reside in the study’s supplementary tables, the published discussion highlights several recognizable genera. Some, such as Enterococcus and Ruminiclostridium, showed associations that diverge from certain human studies, a discrepancy the authors attribute partly to differences in treatment modality and partly to the resolution limits of 16S sequencing, which often cannot identify bacteria at the species or strain level—precisely the classifications most likely to share functional traits. Others, including Flavonifractor, members of the Lachnospiraceae family, and Shigella, have been reported in human cancer immunotherapy research with directions of effect that echo the canine findings. The partial overlap, and partial divergence, underscores both the promise and the early stage of cross-species microbiome oncology: canine and human gut communities are related but not identical, and therapies differ fundamentally between the species.

The authors are careful about what the study can and cannot claim. A baseline sample taken before vaccination cannot, by itself, prove whether the 11 taxa are general prognostic markers of cancer survival or biomarkers specific to the anti-EGFR/HER2 immunotherapy response—though their independence from cancer type, age, and poor-prognosis factors tips the balance toward the latter interpretation. The cohort of 51 dogs was too small to resolve cancer-type-specific biomarkers, and diet data, a major microbiome determinant, were unavailable. Generalizing the results will require larger cohorts and independent validation, which the collaborative group says it is actively pursuing. Still, the mechanistic stakes are high. Gut microbes have been shown in mouse models and human patients to either enhance or blunt immunotherapy effects, and the systemic immune influence of the gut may reach even anatomically distant tumors like osteosarcomas of the bone.

The road ahead, the researchers argue, runs from correlation toward causation. Computational approaches designed to infer causal relationships from observational data—Mendelian randomization, transkingdom network analysis, and mediation analysis—could help identify microbes that do not merely flag better survival but actively drive it. The ultimate test would be interventional: randomized clinical trials of fecal microbiota transplants or anti-cancer probiotics in dogs undergoing immunotherapy. If those experiments succeed, the implications could flow in both directions across the species line. Dogs develop spontaneously occurring cancers in a way few laboratory models replicate, making veterinary oncology a genuine translational bridge. Eleven bacterial taxa, measured from a simple rectal swab on the day of a first vaccine dose, may prove to be the first microbiome-based prognostic tools in canine cancer—and a comparative oncology resource that could ultimately sharpen immunotherapy for human patients as well.

Subject of Research: Gut microbiome biomarkers of survival in dogs with cancer receiving anti-EGFR/HER2 immunotherapy

Article Title: Gut microbiota of dogs with cancer receiving anti-EGFR/HER2 immunization reveals potential biomarkers of patient survival

Article References: Rodrigues, R. R., Karumuru, V., Nuss, S., Elliott, M., Shriver, I., Chao, C.-M., Berriatua, R. C., Doyle, H. A., Tripp, C., Mamula, M. J., Dzutsev, A., Morgun, A., & Shulzhenko, N. (2025). Gut microbiota of dogs with cancer receiving anti-EGFR/HER2 immunization reveals potential biomarkers of patient survival. Veterinary Oncology, 2(1), Article 36. https://doi.org/10.1186/s44356-025-00048-x

Image Credits: AI Generated

DOI: 10.1186/s44356-025-00048-x

Keywords: canine cancer, gut microbiome, immunotherapy, EGFR/HER2 vaccine, 16S rRNA sequencing, survival biomarkers, veterinary oncology, Cox regression, osteosarcoma, hemangiosarcoma, comparative oncology, fecal microbiota transplant

In an effort to improve on those dismal numbers, veterinary oncologists have tested a series of combination chemotherapy protocols incorporating agents such as vincristine, cyclophosphamide, and dacarbazine alongside doxorubicin. None has proven clearly superior to single-agent doxorubicin, and most carry greater toxicity. One combination, however, generated genuine excitement: doxorubicin paired with dacarbazine, administered as a true concurrent combination for dogs with advanced-stage hemangiosarcoma, produced an objective response rate of 47.4 percent in a prospective study, although its impact on overall survival remains undefined. The catch is logistical and toxic. Dacarbazine must be given as an eight-hour intravenous infusion every three weeks, and beyond the usual myelosuppressive risks of neutropenia and thrombocytopenia, it is notorious for causing acute gastrointestinal upset that requires prophylactic antiemetic therapy.

That is where temozolomide enters the story. Temozolomide is an oral chemotherapeutic closely related to dacarbazine, but far easier to administer and generally well tolerated in dogs, with favorable safety data already available when it is combined with doxorubicin in canine lymphoma patients. In human medicine, temozolomide is the first-line chemotherapy for glioblastoma multiforme, a deadly brain cancer, and its mechanism of action is well characterized. Once in the bloodstream, the drug spontaneously hydrolyzes into its active metabolite, 5-(3-methyl-1-triazeno) imidazole-4-carboxamide, or MTIC. MTIC methylates nucleotide bases in DNA, with guanine as the most important target. The cell’s mismatch repair pathway then detects these methylated bases and attempts to repair them in a futile cycle that ultimately triggers apoptosis, the programmed death of the cancer cell.

That mechanism also explains why some tumors resist the drug. In cells with an intact mismatch repair system, sensitivity to temozolomide is governed largely by a DNA repair protein called O-6-methylguanine-DNA methyltransferase, or MGMT. MGMT works by directly removing the methyl groups that temozolomide installs, effectively erasing the damage before it can become lethal. When MGMT is overexpressed, the drug’s fingerprints are wiped away almost as fast as they are written, and the tumor survives. In human glioblastoma, this relationship is so reliable that testing MGMT status, particularly epigenetic silencing of the MGMT promoter, is routinely used to decide whether temozolomide is a viable option for a patient. The question for veterinary researchers was whether the same logic could be imported into canine hemangiosarcoma: if MGMT expression varied among tumors, it might serve as a biomarker to identify which dogs would benefit from adding temozolomide to doxorubicin.

A new study led by Brianna Moore, Julie Nguyen-Edquilang, and Matthew R. Berry at the University of Illinois set out to test exactly that hypothesis in the laboratory. The team worked with five canine hemangiosarcoma cell lines, named DHSA-1426, Emma Brain, Emma Spleen, FITZ, and SBHSA, along with a canine aortic endothelial cell line as a comparison. The cell lines had been shared by collaborators at the University of Minnesota, Colorado State University, and the University of Wisconsin-Madison. To anchor their measurements, the researchers used a human cervical cancer cell line, HeLa, and a pair of mouse glioma cell lines, one wild type and one engineered to carry a human MGMT knock-in, as controls. All cells were cultured under standard conditions at 37 degrees Celsius in a humidified incubator with five percent carbon dioxide.

The researchers measured MGMT at two levels. Quantitative PCR, using canine-specific TaqMan primers with GAPDH as the reference gene, quantified MGMT transcript abundance across three biological replicates with five technical replicates each. Western blotting, performed with a rabbit monoclonal antibody and normalized to the loading control beta-actin, assessed protein expression. The two approaches told a generally concordant story. Four of the five hemangiosarcoma cell lines, DHSA-1426, Emma Brain, Emma Spleen, and SBHSA, showed low or absent MGMT transcript and protein. FITZ was the striking exception, displaying the highest MGMT expression of the panel, roughly twenty-fold greater protein expression than the canine endothelial cell line based on normalized densitometry. The protein bands appeared at the expected sizes, approximately 37 kilodaltons in human controls and approximately 40 kilodaltons in canine cells, a slight species difference confirmed against UniProt database entries.

With the expression map in hand, the team turned to cytotoxicity testing using the sulforhodamine B assay. Cells were seeded in 96-well plates, pretreated for two hours with O6-benzylguanine, an inhibitor that depletes the functional pool of MGMT, and then exposed to temozolomide at concentrations ranging from 0.033 micromolar to 1000 micromolar for seven days. Because temozolomide is unstable in solution, it was dissolved fresh immediately before each experiment. Cytotoxicity values were normalized against a DMSO live-cell control and a raptinal control that induces complete cell death, allowing relative responses to be compared across conditions. The control cell lines behaved exactly as biology predicts. GL261 wild-type cells, which lack MGMT, were sensitive to temozolomide even without the inhibitor, while GL261 MGMT knock-in cells were strongly resistant, with IC50 values comparable to the hemangiosarcoma lines, until O6-benzylguanine pretreatment sensitized them, producing the largest drop in IC50 of any line tested.

The hemangiosarcoma results defied expectations. Every canine cell line proved resistant to temozolomide across the entire tested concentration range, regardless of MGMT status. FITZ, the only line expected to be sensitized by MGMT inhibition based on its high expression, remained stubbornly resistant, showing an undulating dose-response curve with no concentration-dependent effect in any of three biological replicates, and no reliable IC50 could be determined. DHSA-1426 likewise showed minimal cytotoxic response across the range, precluding curve fitting. In other words, MGMT expression status could not predict temozolomide sensitivity in canine hemangiosarcoma cells, and blocking MGMT did not rescue drug sensitivity even where the protein was abundant. The data point instead to an MGMT-independent resistance mechanism, and the authors hypothesize that mismatch repair deficiency is the culprit.

That hypothesis has independent support. A comprehensive analysis of microsatellite instability in canine cancers, a hallmark of mismatch repair defects marked by mutations accumulating in repetitive DNA sequences, found a surprisingly high incidence of 63 percent across canine tumors, with hemangiosarcoma among the tumor types showing elevated instability. If canine hemangiosarcoma is indeed frequently mismatch repair deficient, the implications extend well beyond temozolomide. Mismatch repair deficiency also means deficient methylated DNA goes unnoticed by the cell, undermining the drug’s lethal futile-repair cycle. More intriguingly, tumors with high microsatellite instability carry increased mutational burden and therefore heightened immunogenicity, making them particularly susceptible to immune checkpoint blockade. The authors suggest this provides biological rationale for moving beyond chemotherapy combinations entirely and exploring chemo-immunotherapy, pairing doxorubicin with agents such as PD-1 or PD-L1 inhibitors.

The study’s conclusions are a sober but valuable course correction. MGMT remains a highly promising predictive biomarker for temozolomide selection in human glioblastoma, but this in vitro investigation demonstrates that canine hemangiosarcoma cell lines resist the drug through a mechanism that MGMT testing cannot capture. On that basis, the authors conclude the data do not support clinical exploration of temozolomide, alone or combined with doxorubicin, for treating dogs with hemangiosarcoma. Instead, the findings redirect attention toward characterizing mismatch repair deficiency in canine hemangiosarcoma and toward immunotherapy-based strategies that exploit the high mutational burden such deficiency creates. For a disease that claims most of its canine victims within a year of diagnosis, a result that rules out one dead-end combination while pointing toward a biologically grounded alternative is a meaningful step forward for comparative oncology.

Subject of Research: MGMT expression and temozolomide resistance mechanisms in canine hemangiosarcoma cell lines

Article Title: Investigating MGMT expression as a resistance mechanism to temozolomide in canine hemangiosarcoma cell lines

Article References: Moore, B., Nguyen-Edquilang, J., & Berry, M. R. (2026). Investigating MGMT expression as a resistance mechanism to temozolomide in canine hemangiosarcoma cell lines. Veterinary Oncology, 3(1), Article 24. https://doi.org/10.1186/s44356-026-00076-1

Image Credits: AI Generated

DOI: 10.1186/s44356-026-00076-1

Keywords: canine hemangiosarcoma, temozolomide, MGMT, doxorubicin, chemotherapy resistance, mismatch repair deficiency, microsatellite instability, veterinary oncology, DNA repair, immunotherapy, cell lines, dog cancer