Classic Hodgkin lymphoma is a malignancy of B-cell origin characterized by a distinctive tumor microenvironment, in which malignant Hodgkin and Reed-Sternberg cells make up only a small fraction of the tumor mass while being surrounded by a dense infiltrate of immune cells. This unique biology has made the disease a fertile hunting ground for biomarker research. Identifying patients at high risk of relapse is crucial for tailoring therapy, allowing clinicians to intensify treatment for those most likely to progress while sparing others the long-term toxicities of aggressive regimens. Clinical risk factors, such as those used in the widely adopted EORTC and GHSG frameworks, remain the backbone of risk stratification, but their accuracy leaves room for improvement.

Over the years, a large number of diagnostic tissue-based biomarkers have been described in the literature as correlating with prognosis in classic Hodgkin lymphoma. Yet none has achieved the accuracy and reproducibility required to be incorporated into daily clinical practice. The three markers examined in the new study were among the strongest candidates. c-MET, a receptor tyrosine kinase implicated in cell proliferation and survival signaling, had been reported to be expressed on Hodgkin Reed-Sternberg cells and linked to adverse outcomes. HLA class II molecules, which present antigenic peptides to immune cells, had been shown to be frequently downregulated on tumor cells, and loss of their expression was thought to reflect immune escape and poorer prognosis. CD68, a marker of macrophages, was scored as the percentage of tissue surface area stained, reflecting the density of tumor-associated macrophages in the microenvironment, which previous reports had associated with treatment failure.

To test whether these markers could genuinely predict outcomes, the researchers turned to the EORTC-GELA H9 trial, a large intergroup randomised study in early-stage classic Hodgkin lymphoma. The study employed a case-cohort design, an efficient epidemiological strategy in which all patients who experienced the endpoint of interest are included alongside a randomly selected subcohort of the full trial population. The primary endpoint was progression-free survival, and the analysis encompassed 257 patients with 105 progression-free survival events. All biomarkers were assessed by immunohistochemistry on diagnostic tissue samples, the standard staining technique used in pathology laboratories worldwide.

The statistical approach was deliberately rigorous. Hazard ratios for progression-free survival were estimated using Cox regression models with Barlow weighting, a method specifically designed to account for the case-cohort structure and to produce unbiased estimates despite the sampling design. This careful methodology strengthens the reliability of the negative findings, since it minimizes the risk that the results were distorted by statistical artifacts or sampling bias.

The descriptive results confirmed that the biomarkers were indeed present in the expected patterns. c-MET expression was detected in the tumor cells of 67 percent of evaluable cases, with 108 of 162 samples showing positivity. Cell-surface expression of HLA class II was observed in 51 percent of cases, with 70 of 137 samples positive. Both figures are consistent with the frequencies reported in earlier studies, indicating that the staining and scoring procedures were technically sound and that the cohort was representative of the broader disease population. CD68 expression was scored as the percentage of stained tissue surface area, with scores above the median value of 9.4 percent designated as high.

When the researchers performed univariate Cox regression analyses, however, the results were unequivocally negative. The hazard ratio for progression-free survival was 0.752 for c-MET, with a 95 percent confidence interval of 0.456 to 1.241. For HLA class II, the hazard ratio was 0.854, with a confidence interval of 0.503 to 1.448. For CD68, the hazard ratio was 0.864, with a confidence interval of 0.518 to 1.440. In every case, the confidence interval crossed the value of 1.0, meaning that none of the biomarkers showed a statistically significant association with progression-free survival. Because none of the three markers reached significance on their own, the researchers did not proceed to analyses combining the biomarkers with each other or with established clinical risk factors.

The authors acknowledge that several factors may help explain the lack of validation. The H9 trial employed risk-adapted treatment, meaning that patients received therapy tailored to their clinical risk profile, which could obscure or dilute any prognostic signal from tissue biomarkers. The trial also excluded patients with advanced-stage disease, restricting the analysis to early-stage patients in whom prognostic discrimination is generally harder to achieve because overall outcomes are favorable and event rates are lower. In addition, scoring challenges inherent to immunohistochemistry, including variability in staining intensity and the subjective nature of threshold selection, may have limited the ability to detect true biological associations.

Despite these caveats, the conclusion of the study is firm: these three biomarkers provide insufficient prognostic value for implementation in clinical practice. The finding carries practical weight for the field, because c-MET, HLA class II, and CD68 have each been the subject of multiple publications proposing their use in risk stratification. The new results, derived from a large, well-characterized randomised trial population and analyzed with appropriate statistical methods, suggest that the earlier positive reports may have reflected small sample sizes, heterogeneous patient populations, or publication bias favoring positive findings.

The study, registered as NCT00005584 on ClinicalTrials.gov, is a reminder of the importance of prospective validation in biomarker research. Promising associations observed in retrospective single-center cohorts frequently fail to withstand the scrutiny of large, rigorously designed trials. For patients with early-stage classic Hodgkin lymphoma, risk stratification will continue to rest on established clinical factors for now, while researchers continue the search for tissue-based or molecular markers robust enough to guide individualized treatment decisions in the clinic.

Subject of Research: Prognostic value of immunohistochemical biomarkers in early-stage classic Hodgkin lymphoma

Article Title: Immunohistochemistry for c-MET, HLA class II, and CD68 does not show prognostic value in early-stage classic Hodgkin lymphoma patients in the EORTC-GELA H9 intergroup randomised trial

Article References: Martinez Alcala, R., Visser, L., Fortpied, C., Fermé, C., Henry-Amar, M., Aleman, B., Arens, A., Juul, S., Hayat, A., Hude Dragičević, I., Hutchings, M., Meijnders, P., Mutsaers, P., Paszkiewicz-Kozik, E., Rossetti, S., Sureda, A., Syrykh, C., Tonino, S., Plattel, W., & Diepstra, A. (2026). Immunohistochemistry for c-MET, HLA class II, and CD68 does not show prognostic value in early-stage classic Hodgkin lymphoma patients in the EORTC-GELA H9 intergroup randomised trial. PLOS One, 21(10), e0360080. https://doi.org/10.1371/journal.pone.0360080

Image Credits: AI Generated

DOI: 10.1371/journal.pone.0360080

Keywords: classic Hodgkin lymphoma, c-MET, HLA class II, CD68, immunohistochemistry, biomarkers, prognosis, EORTC-GELA H9 trial, progression-free survival, case-cohort design, risk stratification, PLOS One

The rationale for targeting macrophages directly is rooted in their complicated role in cancer biology. Macrophages are versatile immune cells that maintain tissue homeostasis and defend against pathogens, but within the tumor microenvironment they are notoriously plastic and heterogeneous, adopting functions that range from pro-inflammatory to strongly immunosuppressive. Dense infiltration of tumors by CD68-positive macrophages has been linked in breast cancer to larger tumor size, higher histologic grade, hormone-receptor negativity, lymphovascular invasion, and reduced overall and progression-free survival. Existing imaging probes have struggled to capture this picture accurately. Some tracers, such as those targeting CD163 or CD206, preferentially label the immunosuppressive M2-like subset, while probes aimed at the translocator protein TSPO are not exclusive to macrophages and are affected by genetic polymorphisms that vary between individuals. A copper-64-labeled CD68-Fc fusion protein has been explored, but it targets scavenger receptors indirectly rather than macrophages themselves.

To fill this gap, the researchers turned to phage display, a technique that screens enormous libraries of antibody fragments for those that bind a chosen target. They used a synthetic Fab library, designated Library F, that was deliberately designed without lysine residues in its complementarity-determining regions, the loops that make contact with the antigen. Lysines were replaced with arginines, which preserve the basic character of the loops but lack the reactive nucleophilic amine that would otherwise be chemically modified during radiolabeling. Framework lysines elsewhere in the molecule were retained to allow controlled conjugation to a chelator without disturbing the antigen-binding surface. Screening this library against immobilized recombinant human CD68 yielded 14 unique sequences, of which five, named Fab3, Fab4, Fab7, Fab8, and Fab14, could be expressed successfully in bacterial cultures and purified on Protein A columns.

Characterization of the five candidates revealed important differences. Enzyme-linked immunosorbent assays showed that all five bound human CD68 with high significance compared with a bovine serum albumin control, but Fab3 also bound strongly to BSA itself and cross-reacted with mouse CD68, so it was eliminated. The remaining four Fabs bound human CD68 in a concentration-dependent manner with half-maximal effective concentrations ranging from 1.7 to 15 nanomolar, an affinity range well suited to immunoPET probes, and showed no detectable binding to BSA or mouse CD68. Production metrics then became the deciding factor. Fab4 and Fab7, despite their favorable affinities, expressed inconsistently and lost purity during storage, retaining only 71 and 75 percent monomer after 60 days at 4 degrees Celsius. Fab14 expressed at high yields but with purities of only 60 to 70 percent. Fab8, by contrast, delivered consistent yields of about 1.7 milligrams per liter, the highest purity of 80 to 90 percent straight from Protein A purification, and 96 percent monomer after two months of storage, making it the clear lead candidate.

Converting Fab8 into a PET tracer required attaching a chelator that could hold a positron-emitting radionuclide. The team selected p-SCN-Bn-deferoxamine, or DFO, and zirconium-89, a pairing with a well-established chemistry and a track record of clinical translation. After optimizing the conjugation stoichiometry to a 4:1 ratio of DFO to Fab, which minimized free chelator without inducing protein aggregation, the researchers radiolabeled the conjugate at a specific activity of 7.4 kilobecquerels per nanomole, achieving a radiochemical yield of 99 percent. Stability testing showed that the resulting tracer, [89Zr]Zr-DFO-Fab8, remained more than 95 percent intact monomer in phosphate-buffered saline over 48 hours, and about 80 percent intact in human serum at 37 degrees Celsius over the same period, with the remainder of the radioactivity associated with serum proteins.

In vitro experiments confirmed that the tracer behaved as intended on real cells. The human monocytic leukemia line THP-1 was differentiated into macrophages, which expressed human CD68 on roughly 25 percent of cells compared with about 4 percent of undifferentiated THP-1 cells, as measured by flow cytometry. When fixed macrophages were incubated with the tracer, binding reached about 10 percent of the added activity, five times higher than an irrelevant radiolabeled Fab control and roughly three times higher than in undifferentiated THP-1 cells. Importantly, neither the unlabeled conjugate nor the radiolabeled tracer compromised macrophage viability at the concentrations tested, with only a modest 19 percent decrease at the highest radiotracer concentration of 125 nanomolar, far above the estimated nanomolar circulating levels in the animal studies.

The first in vivo test used a cleverly simple model. Because Fab8 does not cross-react with mouse CD68, the team implanted subcutaneous plugs of Matrigel mixed with either recombinant human CD68 protein or BSA into opposite shoulders of immunocompetent BALB/c mice. PET/CT imaging at 1 and 4 hours after tracer injection showed significantly higher uptake in the human CD68 plugs than in the BSA control plugs, with plug-to-muscle ratios of 11 and 13 versus roughly 5 and 6, respectively. Voxel-level histogram analysis confirmed that the spatial distribution of tracer within the antigen plugs differed significantly from the controls over time. Biodistribution measurements correlated strongly with the imaging data, with a Pearson correlation coefficient of 0.97, and showed the expected renal dominance typical of antibody fragments of this size.

The more biologically demanding test came next. The researchers grew bilateral MDA-MB-231 triple-negative breast cancer xenografts, a tumor line that forms highly vascularized tumors and does not express CD68, in immunodeficient NSG mice. Once the tumors had developed vasculature, they injected differentiated THP-1 macrophages directly into one tumor and undifferentiated THP-1 cells into the other, then administered the tracer systemically. PET imaging at 1 and 6 hours revealed significantly greater uptake in the macrophage-containing tumors, with tumor-to-muscle ratios of about 5.2 and 5.9 compared with 2.7 and 3.6 in the control tumors. Immunofluorescence staining of the excised tumors provided the decisive validation: macrophage-injected tumors contained roughly 660 CD68-positive cells per area versus only 16 in the controls, and the CD68 density correlated with the PET standardized uptake values with a Pearson coefficient of 0.9663. This is the study’s central achievement, a direct, quantitative bridge between a noninvasive PET signal and the histopathological gold standard for pan-macrophage burden.

The authors are candid about the limitations that must be addressed before clinical translation. Kidney uptake was high, as expected for a 50-kilodalton Fab, and about 10 percent aggregation in serum may have contributed to elevated liver and spleen signal through sequestration by the reticuloendothelial system. Because the target uptake in the xenograft model was inherently low, background signal was comparatively prominent, and tumor-to-heart ratios below one suggested incomplete blood clearance at the early imaging timepoints used. Future work will explore polyethylene glycol modification to reduce renal retention, an approach already validated clinically by the approved drug certolizumab pegol, along with better animal models of human macrophage infiltration. Even so, the study establishes a complete pipeline, from phage display discovery through radiopharmaceutical engineering to correlated PET and pathology, for a first-generation human CD68 immunoPET tracer. Such a probe could eventually help select patients for macrophage-targeting therapies, monitor treatment response, and predict resistance to immunotherapy, all from a single noninvasive scan.

Subject of Research: Development of a zirconium-89-labeled anti-CD68 Fab immunoPET tracer for imaging human macrophages

Article Title: A FABulous approach to image human macrophages

Article References: Roohani, B., Vaughn Embs, A., Kumar, R., Lee, S., Katz, S. R., La Prairie, C., Nash, D., Dangarwala, M., Kluger, H. M., Klein, D. E., Zhang, P., Larimer, B. M., Nelson, B., & Marquez-Nostra, B. (2026). A FABulous approach to image human macrophages. European Journal of Nuclear Medicine and Molecular Imaging. https://doi.org/10.1007/s00259-026-08180-y

Image Credits: AI Generated

DOI: 10.1007/s00259-026-08180-y

Keywords: CD68, macrophages, immunoPET, phage display, Fab, zirconium-89, tumor-associated macrophages, PET imaging, triple-negative breast cancer, molecular imaging, radiopharmaceutical, tumor microenvironment