The research team, led by investigators at Nationwide Children’s Hospital and Cincinnati Children’s Hospital Medical Center, began with a resource that few groups possess: autopsy-derived tumor specimens from twenty-six DIPG patients, each paired with matched normal frontal lobe tissue obtained through the Pediatric Brain Tumor Repository under institutional review board-approved protocols with informed consent from families. Bulk RNA sequencing of these paired samples revealed a transcriptome in upheaval. More than two thousand genes were significantly upregulated in tumor tissue, and among them were signatures of immune modulatory factors, including checkpoint genes such as PDCD1 and LAG3 and myeloid-associated genes including CSF1, CD163, CD86, and CCL2. Gene Ontology analysis showed that five of the top enriched pathways involved MHC-mediated antigen presentation, indicating that despite their reputation as immunologically cold tumors, DIPG tissues retain active immunological interfaces.

The apparent paradox, in which antigen presentation machinery is broadly induced while cytotoxic T cells remain scarce, likely reflects a dysfunctional immune interface rather than a functional anti-tumor response. Indeed, when the researchers correlated the expression of macrophage lineage markers such as CD11b, CD14, and STAT6 with patient survival, higher expression of each was significantly associated with shorter survival duration. This finding reinforced a growing consensus that tumor-associated myeloid cells, rather than functioning as anti-tumor sentinels, actively promote disease progression in DIPG. Consistent with earlier work showing sparse lymphocyte infiltration in these tumors, the data painted a portrait of a microenvironment dominated by microglia and monocyte-derived macrophages that have been recruited and shaped by the tumor itself.

To probe how that recruitment happens, the team turned to laboratory experiments with four patient-derived DIPG cell lines and THP-1 human monocytes in transwell migration assays. The results were striking in their asymmetry. SU-DIPG-IV and SU-DIPG-XXXVI, both carrying the H3.1K27M mutation, markedly enhanced monocyte migration and drove the cells toward an immunosuppressive CD11b-positive CD163-positive phenotype, while CCHMC-DIPG-1 and CCHMC-DIPG-2 showed no such capacity. Multiplex cytokine assays revealed that the two potent lines secreted far higher levels of CCL2, IL-10, and TGF-beta1. Yet when the investigators analyzed RNA sequencing data from forty-six pediatric high-grade glioma cell lines in the Childhood Cancer Model Atlas, they found that chemokine expression did not track with histone mutation status at all.

Instead, the decisive variable was the tumor cell’s lineage state. Pediatric glioma cells oscillate among four transcriptional identities, resembling neural progenitor cells, astrocytes, oligodendrocyte precursors, or mesenchymal cells. Using single-sample gene set enrichment analysis, the researchers computed chemokine scores and lineage scores for every cell line and found a strong positive correlation between chemokine expression and the mesenchymal-like program, with a Pearson coefficient of 0.73 and a p-value of 6.5 times ten to the minus nine. Chemokine scores correlated negatively with both the neural progenitor-like and oligodendrocyte precursor-like scores and showed no significant relationship with the astrocyte-like state. Mesenchymal-like cell lines expressed significantly higher levels of CCL2, CCL5, and CCL7 than their counterparts. This suggests that the mesenchymal program, potentially driven by NF-kappaB and STAT3 signaling, is the engine of myeloid recruitment, and it hints at a self-reinforcing loop in which mesenchymal tumor cells summon macrophages that in turn stabilize the mesenchymal phenotype.

Single-cell RNA sequencing of eight DIPG patients, integrated with public datasets spanning H3K27M-mutant, G34R/V-mutant, and wildtype pediatric high-grade gliomas, then allowed the team to reconstruct cellular communication networks across more than twenty-two thousand cells. CellChat analysis revealed that tumor-associated macrophages were present in every histone subtype examined and were the sole recipients of several incoming signals, including TGF-beta, semaphorin-3, and, most intriguingly, the amyloid precursor protein signaling pathway transmitted through the CD74 receptor. Immunofluorescence staining of tumor sections confirmed that APP and CD74-expressing cells occupy spatial proximity within the tumor, consistent with a real ligand-receptor interaction occurring in the microenvironment. APP was broadly expressed across most cell populations, whereas CD74 was restricted to tumor-associated macrophages, defining a one-way channel of communication from tumor to immune cell.

The most consequential observation, however, was what was missing. APP expression was significantly reduced in DIPG tumor tissue compared with matched normal brain at both the RNA and protein levels, a result confirmed by western blotting and mirrored by decreased expression of the related family members APLP1 and APLP2. To understand what this loss means for macrophage behavior, the researchers treated THP-1-derived macrophages with recombinant human APP protein. The stimulation triggered a robust proinflammatory transformation: interferon-stimulated genes such as IFIT1 and IFIT2 were induced, inflammatory cytokines including IL1B, CCL2, and CXCL10 rose, and immunosuppressive markers such as CD163, CD206, and ARG2 were suppressed. Gene set enrichment analysis confirmed strong activation of interferon alpha, interferon gamma, and TNF alpha signaling via NF-kappaB, and multiplex cytokine assays showed time-dependent increases in secreted IFN-gamma, TNF-alpha, IL-1beta, and IL-6.

These functional data support a compelling hypothesis: when APP levels fall in the tumor, macrophages lose a signal that would otherwise push them toward an inflammatory, potentially anti-tumor state, and instead settle into the non-inflammatory, immunosuppressive phenotype that characterizes DIPG. The authors caution that the causal relationship remains to be established, and they note an apparent discrepancy with a recent glioblastoma study in which APP appeared to suppress macrophage phagocytosis via CD74, suggesting that context and the specific cellular readout matter. To lay groundwork for therapeutic intervention, the team went further, building three-dimensional structural models of membrane-bound APP and the CD74 trimer using I-TASSER-MTD, refining them in GROMACS, and performing protein-protein docking with HADDOCK. Membrane-restrained Gaussian network modeling then identified a mechanically rigid binding interface in which APP residues 19 through 24 engage a hydrophobic pocket in CD74 centered on residues 118 through 122.

That structurally defined interface offers an actionable target. Peptides engineered to engage the CD74 hydrophobic pocket could, in principle, mimic APP signaling and reprogram tumor-associated macrophages toward proinflammatory phenotypes with anti-tumor capacity, complementing emerging CAR-T cell therapies directed against GD2, B7-H3, and IL13-Ralpha2. The authors acknowledge important limitations: the sequencing work relied on post-mortem tissue from patients who had received varied treatments, the THP-1 model does not fully recapitulate the complexity of primary human macrophages, and the computational docking will require validation through site-directed mutagenesis and co-immunoprecipitation. Still, the study represents a substantial advance in understanding how lineage identity governs immune recruitment in pediatric high-grade glioma and how a protein famous in neurodegeneration may hold the key to thawing one of childhood cancer’s coldest tumors. Future studies using biopsy specimens, primary human systems, and syngeneic preclinical models will determine whether restoring APP-CD74 signaling can genuinely convert the immunosuppressive microenvironment of DIPG into one that fights back.

Subject of Research: Tumor-associated macrophage recruitment and the APP-CD74 signaling axis in pediatric high-grade glioma

Article Title: Integrated multi-omics identifies lineage-dependent myeloid cells recruitment and the APP-CD74 axis as an immunoregulatory target in pediatric high-grade glioma

Article References: Wang, Z., Kumar, A., Umaru, B., Iyer, A. M., Khan, K., Pang, H.-H., Fouladi, M., & Drissi, R. (2026). Integrated multi-omics identifies lineage-dependent myeloid cells recruitment and the APP-CD74 axis as an immunoregulatory target in pediatric high-grade glioma. Acta Neuropathologica, 152(1), Article 39. https://doi.org/10.1007/s00401-026-03089-0

Image Credits: AI Generated

DOI: 10.1007/s00401-026-03089-0

Keywords: diffuse intrinsic pontine glioma, pediatric high-grade glioma, tumor-associated macrophages, APP-CD74 axis, amyloid precursor protein, CD74, tumor microenvironment, single-cell RNA sequencing, mesenchymal-like lineage state, chemokines, immunotherapy, multi-omics

The study reviewed the medical records of patients diagnosed with ROS1-rearranged non-small cell lung cancer at Beijing Chest Hospital between February 2017 and January 2025. From 99 patients identified as harboring the rearrangement, 55 met the criteria for inclusion in the final analysis. The retrospective design allowed the team to capture the full texture of routine care, including the variety of diagnostic methods used, the sequencing of treatments, and the long-term outcomes that patients actually experienced outside the constraints of clinical trial eligibility criteria. Fusion partners were characterized using molecular techniques including next-generation sequencing and real-time polymerase chain reaction, and clinical responses were evaluated according to standard RECIST criteria for solid tumors.

The most striking structural insight from the study is that CD74 emerged as the most common fusion partner gene in this cohort, confirming patterns seen in other populations. ROS1 fusions arise when the ROS1 gene breaks and fuses with a variety of partner genes, producing a constitutively activated kinase that drives uncontrolled cell growth. Different fusion partners preserve different portions of the ROS1 kinase domain and can vary in breakpoint position, and there has long been speculation in the field about whether these structural differences translate into different clinical behaviors. The Beijing data now provide concrete real-world evidence that they do, at least for one fusion partner.

Treatment patterns in the cohort reflect how thoroughly targeted therapy has displaced chemotherapy in this molecularly defined subgroup. Targeted therapy with tyrosine kinase inhibitors, drugs that directly block the aberrant ROS1 signaling machinery, accounted for 87.3 percent of first-line treatment decisions, and ultimately every patient in the study received targeted therapy at some point during their treatment course. The results were impressive by any measure: the objective response rate for first-line treatment reached 72.7 percent, meaning nearly three out of four patients saw their tumors shrink measurably, and the median progression-free survival across the cohort was 18.9 months, a substantial span of disease control in advanced lung cancer.

But when the investigators stratified outcomes by fusion subtype, a clear divide appeared. Patients whose tumors carried an SDC4 fusion partner had an objective response rate of only 44.4 percent in first-line treatment, compared with 86.1 percent for those with other fusion subtypes, a statistically significant difference. Progression-free survival told the same story: the SDC4 subgroup experienced a median progression-free survival of 7.5 months, dramatically shorter than the 22.4 months seen in patients with non-SDC4 fusions. Univariate and multivariate Cox regression analyses converged on the same conclusion, identifying SDC4 subtype as an independent prognostic factor for shorter progression-free survival on first-line therapy. The biological explanation remains an open question, but the finding suggests that the architecture of the fusion itself may influence how completely kinase inhibitors can suppress the oncogenic signal, or how quickly resistance mechanisms emerge.

Beyond fusion subtype, the study illuminated other biological markers with prognostic weight. Programmed death-ligand 1 expression, a measure of the immune-evading machinery of tumor cells, also mattered: patients with PD-L1 expression below one percent had remarkably longer median progression-free survival that had not yet been reached at the time of analysis, compared with 17.3 months for those whose tumors expressed PD-L1 at one percent or higher. Baseline metastatic site added further prognostic nuance. Patients with bone metastases at diagnosis had a median progression-free survival of just 7.5 months versus 24.2 months for those without, while for overall survival, smoking history and baseline liver metastasis each predicted markedly shorter survival, cutting median overall survival to 22.6 and 23.8 months respectively, compared with 63.2 months for patients without those risk factors.

Treatment choices themselves proved consequential in ways that reinforce current clinical thinking. Patients who received tyrosine kinase inhibitors as first-line therapy had a median progression-free survival of 22.4 months, compared with only 6.4 months for those who began with other regimens, and multivariate analysis confirmed first-line TKI use as an independent prognostic factor. Perhaps the most powerful predictor of all was the earliest signal of treatment effect: patients who responded to their initial targeted therapy lived dramatically longer without progression, 23.6 months versus 4.4 months, and their overall survival reflected this advantage, with responders living significantly longer than non-responders. The message for clinicians is unambiguous, that securing an early response with the right targeted agent sets the trajectory for the entire course of the disease.

Taken together, the survival figures offer a genuinely encouraging bottom line for this rare patient population. The median overall survival of the entire cohort reached 59.8 months, a figure that would have been unthinkable in the era before molecularly targeted lung cancer therapy, when advanced ROS1-rearranged disease was treated as an undifferentiated malignancy with cytotoxic chemotherapy. The study also demonstrated that patients benefit consistently from targeted therapy across the fusion subtype spectrum, even if the degree of benefit varies, and that the results achieved in this real-world Chinese cohort align well with the response rates reported in registration trials for ROS1 inhibitors, providing reassurance that trial outcomes translate to routine practice.

The implications for precision oncology extend well beyond the statistics. As comprehensive molecular profiling becomes standard for advanced lung cancer, this study argues that reporting should go beyond simply flagging ROS1 rearrangement as positive or negative and should instead specify the fusion partner and co-variant markers such as PD-L1 expression and metastatic pattern, because these variables together define distinct prognostic strata within what has been treated as a single molecular class. Patients with SDC4 fusions, bone metastases, or higher PD-L1 expression may warrant closer surveillance, earlier sequencing to second-generation inhibitors, or combination strategies, while those with favorable profiles can be counseled with well-founded optimism. For a biomarker that occurs in a small minority of lung cancers, ROS1 continues to punch far above its weight in shaping the future of personalized cancer medicine, and studies like this one, grounded in the messy reality of clinical care, are essential for ensuring that every patient within a molecular subgroup receives the nuance their tumor biology demands.

Subject of Research: Real-world survival outcomes of patients with advanced ROS1 fusion-positive non-small cell lung cancer

Article Title: Survival of patients with advanced non-small cell lung cancer harboring different ROS1 fusion subtypes: a real-world study

Article References: Survival of patients with advanced non-small cell lung cancer harboring different ROS1 fusion subtypes: a real-world study. (n.d.). https://doi.org/10.1186/s12885-026-16863-w

Image Credits: AI Generated

DOI: 10.1186/s12885-026-16863-w

Keywords: non-small cell lung cancer, ROS1 fusion, SDC4 subtype, CD74, targeted therapy, tyrosine kinase inhibitor, progression-free survival, overall survival, PD-L1, bone metastasis, precision oncology, real-world study