The clinical stakes are far from abstract. Organ shortages have made older donor livers a routine part of transplantation, and outcomes still track donor age in ways that stubbornly resist prediction. A registry-based analysis of more than 111,000 liver transplants cited in the review found that when older grafts went to younger recipients, five-year graft survival fell to 69.9 percent, compared with 77.8 percent for young-to-old combinations. Grafts from donors over 60 to 70 years fared particularly poorly in high-risk recipients with MELD scores of 35 or above, and in living donation, use of donors aged 50 and older carried a 61 percent increased hazard of graft failure. The paradox that frustrates clinicians is that age alone incompletely predicts post-transplant behavior. Some older livers retain substantial reserve while others collapse under metabolic overload, inflammatory injury or impaired recovery, which is precisely why the authors argue for a biologically grounded, module-based framework rather than another hunt for a single decisive marker.

That framework rests on five interrelated biological modules that emerged from converging multi-omics evidence: lipid stress and metabolic remodeling, diminished oxidative reserve, proteostatic strain, inflammatory sensitization, and reduced recovery capacity. Each module is supported by different layers of molecular data, and each connects back to mitochondria, the organelles that power fatty-acid oxidation, oxidative phosphorylation, redox homeostasis and detoxification in hepatocytes. The logic is that impaired mitochondrial metabolic flexibility, weakened redox buffering, disrupted organelle crosstalk and failing quality-control systems provide the common substrate on which all five disturbances build. Critically, the authors stress that these modules remain partly compensated at baseline. Molecular age-related changes accumulate quietly while histology and standard blood tests look normal, and only become conspicuous when the liver is challenged by metabolic overload, ischemia-reperfusion injury, cold preservation or ex vivo machine perfusion.

The lipid stress module illustrates how the pieces interlock. Multi-omics studies across mouse strains, diets and ages show that age-associated hepatic phenotypes are shaped not just by time but by genotype and diet, with alterations in fatty-acid handling, lipid-droplet distribution, peroxisomal programs and FGF21 and FABP4 signaling narrowing metabolic flexibility even when baseline function appears preserved. One mechanistic study linked senescence-associated 13-hydroxyoctadecadienoic acid, an oxidized lipid mediator, to age-related steatosis by showing that it directly inhibits catalase, a key antioxidant enzyme. That finding stitches lipid stress to the oxidative reserve module: aged rodent and primate livers repeatedly show altered glutathione balance, elevated reactive oxygen and nitrogen species, increased malondialdehyde, mitochondrial DNA damage and changes in catalase, superoxide dismutase and respiratory-chain proteins. In other words, lipid overload and antioxidant decline are not separate problems but two faces of the same mitochondrial vulnerability.

Proteostatic strain adds a third thread. Rodent phosphoproteomic data and primate single-nucleus studies converge on altered chaperone, protein-degradation and endoplasmic reticulum stress pathways, suggesting that the energetic cost of maintaining protein quality rises with age even in hepatocytes that look intact. The review links this to lipid and redox stress through a mechanistic chain: excess mitochondrial calcium and reactive oxygen species increase ER protein-folding demand, while ATP limitation weakens chaperone and degradation capacity, so proteostatic strain may act as a downstream amplifier of metabolic overload rather than an isolated defect. Inflammatory sensitization operates at the multicellular level. Single-cell studies identify altered endothelial, macrophage, neutrophil and hepatocyte programs in aged liver, and challenge models show that IL-1α signaling between liver sinusoidal endothelial cells and hepatocytes, along with Cxcl2-positive macrophage recruitment of neutrophils, can amplify ischemia-reperfusion injury in old livers. Finally, recovery capacity, the ability to restore metabolic, synthetic and structural homeostasis after injury, is compromised by dysregulated Hedgehog signaling, reduced hepatocyte plasticity and extracellular matrix remodeling, meaning two grafts can absorb similar early stress yet diverge sharply in lactate clearance, bile production and survival.

Why does mitochondria sit at the center rather than the periphery of this network? The review describes a feedback architecture in which lipid overload increases mitochondrial calcium loading and ROS generation, which promotes ER stress and inflammatory signaling; the resulting proteostatic and inflammatory burden then consumes energy and limits repair, further impairing mitochondrial quality control and metabolic flexibility in a self-reinforcing loop. This structure generates testable predictions. A liver carrying combined lipid stress and low redox reserve should show a larger stress-induced rise in mitochondrial, ER-stress and inflammatory readouts and a slower recovery trajectory than a liver with only one abnormal module. The authors are careful to note that mitochondrial vulnerability is the interface connecting the modules, not a separate module itself and not a validated measure of biological age, but it offers a unifying hypothesis for why aging livers with identical birthdays and similar biopsies can behave so differently under duress.

Not all evidence carries equal weight, and the review is unusually candid about this. Rodent studies provide most of the mechanistic depth, with large mouse multi-omics datasets and naturally aged models showing distributed remodeling across lipid metabolism, immune signaling, stress-kinase networks and protein-state regulation. Non-human primate data, though limited in scale and often drawn from female animals, strengthen confidence that ER stress, degradation pathways and hepatocyte pro-senescence programs are conserved rather than rodent-specific artifacts. Human baseline evidence remains the weakest layer: transcriptomic signals of aging are entangled with steatosis, fibrosis, metabolic disease and medication history, and matched human tri-omics datasets are scarce. Spatial and single-cell studies add a crucial caveat about anatomy, showing that hepatocytes do not age uniformly across the lobule. Pericentral cells preferentially accumulate large lipid droplets while periportal cells show more prominent mitochondrial and epigenetic disturbance, which means small biopsies or bulk assays can miss biologically important local variation.

This is where transplantation and machine perfusion enter as natural experiments. Procurement, cold preservation, reperfusion and metabolic reactivation impose rapid, defined swings in oxygen availability and energetic demand, converting modest baseline differences into clinically meaningful ones. Hypothermic oxygenated perfusion and normothermic machine perfusion have already proven their clinical value in randomized trials, reducing early allograft dysfunction, biliary strictures and organ discard, but their scientific value here lies in the sampling window they create. During HOPE, early perfusate flavin mononucleotide, a marker of mitochondrial injury, was higher in grafts that developed early allograft dysfunction, and in a 50-graft validation cohort the five-minute value yielded an area under the curve of 0.744 for early allograft dysfunction and 0.890 for one-year mortality. During NMP, serial perfusate transaminases, lactate dehydrogenase, von Willebrand factor and platelet changes associated with early dysfunction, while perfusate IL-6, TNF-α and IL-10 correlated with postoperative bilirubin, intensive-care stay or graft loss, and bile pH, bicarbonate and glucose tracked biliary viability.

The authors insist these signals are promising readouts, not decision rules. No published study has validated a combined panel across perfusion protocols, and none of the fluid-based markers is aging-specific; they reflect donor comorbidity, ischemia, procurement and preservation strategy as well as age-related vulnerability. The module framework is explicitly hypothesis-generating rather than outcome-validated, and the review warns against merging FMN trajectories, cytokine measurements and biliary chemistry from different cohorts into a clinical scoring system without prospective, matrix-matched validation. What the framework does offer is translational logic: tissue biopsies capture pathway state, serial perfusate captures stress and recovery kinetics, bile captures cholangiocyte function, and post-transplant outcomes provide the ultimate clinical anchor. Future priorities include paired tissue-fluid biobanks, attention to circadian timing and pre-analytical handling of metabolites, spatially informed biopsy interpretation, and explicit separation of immediate procurement injury from genuine recovery behavior in sampling schedules.

The bigger picture is a conceptual shift in how aging science approaches one of medicine’s hardest assessment problems. Rather than seeking a molecular birth certificate for the liver, the review proposes mapping which vulnerability modules are active in a given organ and how they will respond to defined stress. Candidate intervention targets, including FGF21 and FABP4 in lipid metabolism, the 13-HODE-catalase axis in antioxidant buffering and Hedgehog signaling in regenerative reserve, remain experimental, and the authors caution that any therapeutic claim requires direct evidence of benefit against clinically meaningful endpoints. But the reframing itself is consequential: if the most stable features of liver aging are process-level changes that become functionally visible under stress, then machine perfusion may be not just a preservation technology but a stress-testing platform, a controlled environment in which the hidden mitochondrial biography of an aging liver can finally be read before it matters.

Subject of Research: Mitochondrial vulnerability and multi-omics biological modules of the aging liver and their translational relevance to transplantation

Article Title: Multi-omics perspectives on mitochondrial vulnerability in the aging liver: biological modules and translational implications

Article References: Chen, H., Lu, W., Wang, K., Fang, R., Lee, S., Chen, S., & Liu, Z. (2026). Multi-omics perspectives on mitochondrial vulnerability in the aging liver: biological modules and translational implications. Frontiers in Cell and Developmental Biology, 14, Article 1954948. https://doi.org/10.3389/fcell.2026.1954948

Image Credits: AI Generated

DOI: 10.3389/fcell.2026.1954948

Keywords: liver aging, mitochondria, multi-omics, machine perfusion, liver transplantation, oxidative stress, proteostasis, inflammation, biomarkers, FGF21, ischemia-reperfusion injury, hepatocytes

The study, led by researchers at Fujita Health University, set out to probe a long-standing puzzle in metabolic physiology. Insulin and glucagon, the two pancreatic hormones that jointly govern nutrient handling, respond very differently to protein than to carbohydrate. Glucose ingestion robustly stimulates insulin and suppresses glucagon in healthy individuals, but in type 2 diabetes the early insulin response is blunted and glucagon paradoxically rises. Protein, by contrast, stimulates both hormones in everyone, making protein a particularly potent glucagon secretagogue regardless of glycemic status. Because insulin drives amino acid uptake into peripheral tissues for protein synthesis while glucagon promotes hepatic gluconeogenesis, converting amino acids into glucose, the researchers hypothesized that the balance between these hormones, known as the insulin-to-glucagon ratio, would determine how the liver handles a sudden flood of dietary amino acids.

To test this, the team compared nondiabetic mice with mice made diabetic by injections of streptozotocin, a compound that destroys insulin-producing beta cells and produces a model of severe insulin deficiency. After an overnight fast, both groups received either a liquid normal control diet, containing roughly 67 percent carbohydrate and 15 percent protein, or a liquid high-protein diet with 58 percent protein and only 23 percent carbohydrate, matched for calories. Blood and liver samples were collected two hours after feeding, a window chosen to capture the immediate postprandial response rather than the slower transcriptional adaptation that emerges over days of high-protein feeding.

The hormonal results were revealing. In nondiabetic mice, the high-protein meal produced lower insulin and GIP secretion at 60 minutes than the control diet, an effect the authors attribute largely to the reduced carbohydrate content of the meal. Glucagon, however, rose significantly higher with the protein load. In diabetic mice, insulin secretion in response to the control diet was markedly attenuated, as expected, but the response to the high-protein diet was comparable to that of nondiabetic animals. Strikingly, despite lower insulin secretion, postprandial blood glucose was actually lower in diabetic mice given the high-protein meal than in those given the carbohydrate-rich control diet, suggesting that cutting carbohydrate from a meal can blunt glycemic excursions even when insulin secretion is impaired.

Plasma amino acid profiling then exposed a sharp divergence between the two metabolic states. The high-protein diet elevated nearly all essential amino acids, including the branched-chain amino acids leucine, isoleucine, and valine, in both diabetic and nondiabetic mice. But on the normal control diet, diabetic mice generally showed lower plasma amino acid levels than their nondiabetic counterparts, a pattern consistent with accelerated gluconeogenesis draining amino acids from the circulation. When the protein load arrived, several amino acids, including arginine, tyrosine, phenylalanine, tryptophan, and isoleucine, rose significantly higher in diabetic mice than in nondiabetic ones, hinting that the diabetic body handles a protein surge differently from the very first hours.

The liver told an even more dramatic story. In nondiabetic mice, two hours of high-protein feeding left hepatic amino acid content essentially unchanged, with glutamine the only significant increase. In diabetic mice, the same meal produced significant hepatic accumulation of alanine, aspartic acid, proline, threonine, and all three branched-chain amino acids, and the increase extended to nearly every essential amino acid except histidine. Glutamine was the lone exception, falling significantly below control levels. The combination of lower plasma levels for some amino acids and higher hepatic content in diabetic mice led the authors to conclude that amino acid uptake into the liver is accelerated in diabetic animals given a protein load, effectively overwhelming the organ’s metabolic capacity and leaving substrate pooled inside hepatic tissue.

Crucially, this accumulation occurred without any corresponding transcriptional program. Quantitative PCR of sixteen genes involved in glucose and amino acid metabolism, including gluconeogenic enzymes such as Pepck and G6pc, aminotransferases, urea cycle genes, and glutaminase, revealed almost no differences between diet groups within the two-hour window. The only significant diet-related change was Cps1 in diabetic mice. This stands in sharp contrast to the team’s earlier work showing that seven days of high-protein feeding induces robust glucagon-mediated upregulation of hepatic amino acid-catabolizing enzymes, which keeps plasma amino acids, apart from branched-chain amino acids, near normal. The new data indicate that the immediate postprandial amino acid excursion is governed by non-transcriptional mechanisms, such as altered tissue uptake and substrate diversion toward gluconeogenesis, rather than by rapid changes in enzyme gene expression.

Signaling analyses added further nuance. Western blotting of liver proteins showed that Akt phosphorylation, a readout of insulin signaling, was significantly reduced by the high-protein meal in diabetic mice, and reduced glycogen content followed the same pattern in both groups, likely reflecting the meal’s lower carbohydrate content. Yet phosphorylation of GSK3β, Akt’s downstream effector in glycogen synthesis, and of CREB, the glucagon-responsive transcription factor, were unchanged across all four groups, suggesting that the two-hour time point captures signaling dynamics whose temporal relationship to hormone levels is more complex than simple feed-forward regulation. The authors note that time-course studies will be needed to untangle these kinetics.

Metabolomics also uncovered a paradox in hepatic glycogen. Despite impaired insulin secretion and a glucagon-dominant hormonal profile, diabetic mice had higher liver glycogen than nondiabetic mice under both diets, echoing recent evidence that glucagon may contribute to postprandial glycogen repletion and that hyperglycemia itself, through sheer substrate availability, can drive glycogen accumulation even without insulin action. Meanwhile, diabetic livers showed elevated early glycolytic and gluconeogenic intermediates, including glucose 6-phosphate and fructose 6-phosphate, which the authors interpret not as inefficient metabolic flux, as reported in obese diabetic models, but as a distinct state in which overwhelming substrate availability from accumulated amino acids feeds into glucose production pathways.

The clinical implications are intriguing and carefully hedged. In type 1 diabetes, where insulin secretion is severely compromised, protein-rich meals can promote hyperglycemia by stimulating glucagon. But the present findings suggest that when residual insulin secretion is preserved, as in many cases of type 2 diabetes, high-protein, low-carbohydrate feeding enhances hepatic amino acid uptake and glucose production without raising postprandial blood glucose, because the limited carbohydrate content offsets the glucose-raising effect of glucagon-stimulated gluconeogenesis. The study also raises new questions about FGF21, a liver-derived hormone acutely lowered by the protein meal in nondiabetic mice, hinting that its secretion is rapidly tuned by dietary protein or insulin action. The authors acknowledge limitations, including the absence of fasting-baseline metabolomics, no direct measurement of glycogen turnover, and no assessment of intestinal nutrient absorption. Even so, the work delivers a memorable message: in diabetes, the liver’s first response to a protein feast is a silent traffic jam of amino acids, one that precedes and perhaps shapes the slower genetic adaptation that follows.

Subject of Research: Acute high-protein feeding and hepatic amino acid metabolism in diabetic versus nondiabetic mice

Article Title: Acute high‐protein feeding induces hepatic amino acid accumulation in diabetic male mice prior to transcriptional adaptation

Article References: Sumioki, H., Tomatsu, E., Seino, Y., Hidaka, S., Ueno, S., Kanie, S., Nishida, K., Haraguchi, T., Murao, N., Fujisawa, H., Takayanagi, T., Sugimura, Y., Hayashi, Y., & Suzuki, A. (2026). Acute high‐protein feeding induces hepatic amino acid accumulation in diabetic male mice prior to transcriptional adaptation. Physiological Reports, 14(18), Article e71117. https://doi.org/10.14814/phy2.71117

Image Credits: AI Generated

DOI: 10.14814/phy2.71117

Keywords: diabetes, glucagon, insulin, amino acids, liver metabolism, high-protein diet, branched-chain amino acids, gluconeogenesis, FGF21, metabolomics, streptozotocin, postprandial metabolism

The research, conducted by a team led by Andrea De Giglio of the University of Bologna and published in Cancer Immunology, Immunotherapy, followed 40 patients with advanced non-small cell lung cancer who were beginning first-line treatment with immune checkpoint inhibitors, either alone or in combination with chemotherapy. Rather than focusing exclusively on conventional tumor-side markers such as PD-L1 expression or tumor mutational burden, the investigators tracked a panel of host-derived immunometabolic biomarkers in the blood: soluble receptor for advanced glycation end products, better known as sRAGE, fibroblast growth factor 21, or FGF21, growth differentiation factor 15, or GDF15, and cell-surface-associated perilipin 2, or cPLIN2. In parallel, they used computed tomography scans — which these patients undergo routinely for disease monitoring — to quantify body composition parameters, including measures of lean mass and fat distribution.

The rationale behind this approach is rooted in a growing recognition that cancer immunotherapy is not simply a dialogue between a drug and a tumor. Checkpoint inhibitors work by unleashing T cells, and T cells are exquisitely sensitive to the metabolic environment in which they operate. Systemic inflammation, insulin resistance, sarcopenia and obesity-related metabolic signaling can all reshape that environment. FGF21 and GDF15, for example, are stress-responsive endocrine factors that rise during mitochondrial dysfunction, cellular stress and cachexia, and they have been implicated in reprogramming immune cell behavior. sRAGE, meanwhile, serves as a decoy receptor that mops up inflammatory damage-associated molecular patterns, and its circulating levels are increasingly viewed as a readout of chronic low-grade inflammation and lung epithelial health.

The findings were striking. In the exploratory cohort, patients with high baseline levels of sRAGE experienced significantly longer progression-free survival and overall survival than those with lower levels. Even more intriguingly, the trajectory of sRAGE during treatment carried information: patients whose sRAGE levels increased while on immunotherapy were more likely to respond to the treatment, suggesting that this biomarker might serve not only as a static prognostic factor but as a dynamic indicator of an unfolding, favorable immune response. This kind of longitudinal signal is particularly valuable in a clinical setting where early assessment of treatment efficacy can determine whether to continue, intensify or switch therapy long before radiographic changes become apparent.

FGF21 told a very different story. Elevated circulating levels of this hormone were associated with worse overall survival in the exploratory cohort, with a hazard ratio of 1.11 per unit increase and a p-value of 0.026. FGF21 levels were also inversely correlated with lean mass, linking the biomarker to the loss of skeletal muscle — a hallmark of cancer cachexia and a well-established adverse prognostic factor in lung cancer. This connection between a metabolic stress hormone and body composition underscores a central theme of the study: the metabolic state of the patient’s body as a whole may help determine whether the immune system, once released from checkpoint inhibition, can mount and sustain an effective anti-tumor campaign.

To guard against the risk that these associations were statistical accidents in a small cohort, the team validated the two most promising markers — FGF21 and sRAGE — in an independent group of 76 patients with advanced non-small cell lung cancer. The result was a partial replication that sharpened the biological picture. FGF21 remained an independent predictor of outcome, with higher levels associated with both shorter progression-free survival and shorter overall survival. sRAGE, however, lost its association with survival in the validation cohort, a sobering reminder of the reproducibility challenges that plague biomarker research, particularly for inflammatory markers that fluctuate with comorbidities, medications and subclinical disease activity. The authors are appropriately cautious, framing sRAGE as a hypothesis-generating signal that warrants further study rather than a validated clinical tool.

The study went beyond circulating proteins to interrogate the tumor microenvironment itself. Using in silico transcriptomic analyses of tumor gene expression data, the researchers explored how the genes related to these biomarkers associate with immune signatures within tumors. The results revealed two distinct patterns. Tumor expression of PLIN2 and sRAGE was linked to immune-inflammatory phenotypes — tumors characterized by an active, infiltrated immune landscape where checkpoint blockade is most likely to succeed. In contrast, FGF21 and GDF15 were associated with immune-desert phenotypes, tumors that lack meaningful immune cell infiltration and are notoriously resistant to immunotherapy regardless of how the drug is delivered.

This convergence of clinical and transcriptomic evidence gives the study its conceptual weight. It suggests that the same immunometabolic axes measurable in a patient’s blood may mirror, or even shape, the immune architecture of the tumor. A patient whose blood chemistry reflects chronic metabolic stress — elevated FGF21, dwindling lean mass — may also harbor a tumor microenvironment depleted of T cells, rendering checkpoint inhibitors pharmacologically futile. Conversely, a favorable inflammatory-metabolic profile, signaled by higher sRAGE and rising levels during therapy, may coincide with an inflamed tumor that is primed to respond. The blood biomarkers, in other words, could act as accessible windows into the tumor’s immunological personality.

The clinical implications are significant, though the investigators and outside experts alike emphasize that the work is preliminary. A blood test measuring four circulating proteins, combined with body composition data already embedded in routine CT scans, could one day help stratify patients before or early during immunotherapy — identifying those who need closer monitoring, metabolic interventions such as nutritional support and exercise to preserve muscle, or alternative treatment strategies altogether. Prospective cohorts of 40 and 76 patients cannot support immediate changes to clinical practice, and neither biomarker has yet demonstrated the predictive — as opposed to purely prognostic — performance needed to guide individual treatment decisions. Larger, multi-center validation studies will be essential, ideally with standardized assay protocols and serial sampling designed to test whether early changes in FGF21 or sRAGE genuinely anticipate radiographic response.

Still, the SARC-LUNG study adds to a compelling and fast-moving body of evidence that the success of cancer immunotherapy depends on the whole organism, not just the tumor. As immunometabolism moves from the laboratory bench toward the oncology clinic, biomarkers like FGF21, GDF15, sRAGE and PLIN2 may help clinicians see what conventional tumor profiling has missed: the metabolic soil in which the immune response must grow. For patients with advanced lung cancer — a population in which cachexia, inflammation and metabolic derangement are common — that broader view could eventually mean smarter treatment selection, earlier detection of failure, and new opportunities to intervene on the host side of the cancer-immunity equation.

One notable aspect of the SARC-LUNG design is its prospective structure. Unlike most biomarker studies in lung cancer, which are assembled retrospectively from banked samples of uncertain provenance, this cohort enrolled patients at the moment treatment began and collected blood specimens and imaging at defined time points. That prospective framework reduces the risk of selection artifacts and makes the longitudinal observations — such as the rise in sRAGE among responders — considerably more meaningful than a single cross-sectional measurement would allow.

The study also reflects a broader shift in how biomarker research is being organized in Italy and across Europe. The work received support through national research programs, including funding linked to the HEAL ITALIA initiative and projects addressing aging in an aging society, and it was carried out jointly by the University of Bologna and collaborating centers in Genoa and Rome. This multi-institutional arrangement matters because host-derived biomarkers are sensitive to pre-analytical conditions — how blood is drawn, processed and stored — and harmonized procedures across sites strengthen the credibility of the measurements.

It is worth emphasizing what distinguishes prognostic from predictive information in this setting. A prognostic marker tells us something about the likely course of disease regardless of the therapy given, whereas a predictive marker identifies patients who will specifically benefit from a particular treatment. The associations reported in SARC-LUNG, including the independent relationship between FGF21 and survival in the validation cohort, are best understood as prognostic signals. Demonstrating true predictive value would require comparing biomarker-defined subgroups across randomized treatment arms, something a single-arm observational study cannot do.

The biological plausibility of the FGF21 finding deserves attention as well. FGF21 is produced primarily by the liver in response to mitochondrial stress and nutrient deprivation, and chronically elevated levels are observed in conditions ranging from obesity-related metabolic dysfunction to cardiac cachexia. Its inverse correlation with lean mass in this cohort fits a model in which rising FGF21 reflects an emerging catabolic state — one that may deplete the energy reserves and muscle-derived substrates that a robust anti-tumor immune response demands.

Finally, the publication appears as an open-access article shared ahead of final copyediting, a format that accelerates access to accepted peer-reviewed findings while the Version of Record is being prepared. For a field where biomarker hypotheses often circulate slowly through conference presentations, this early availability allows other research groups to begin designing the larger validation studies that will determine whether these immunometabolic signatures can ultimately earn a place in clinical decision-making.

Subject of Research: Host-derived immunometabolic biomarkers associated with prognosis and immune microenvironment profiles in advanced non-small cell lung cancer treated with immune checkpoint inhibitors.

Article Title: Host-derived immunometabolic biomarkers identify prognostic profiles under immunotherapy in advanced NSCLC: the SARC-LUNG study

Article References: De Giglio, A., Conte, M., Coco, S., Galuppi, F., Santamaria, S., Rosa, A., Lo Bianco, F., Naddeo, M., Ricciotti, I., Di Federico, A., Favorito, V., Trofarello, L., Mantuano, F., Sperandi, F., Gelsomino, F., Brocchi, S., Mosconi, C., Genova, C., Salvioli, S., & Ardizzoni, A. (2026). Host-derived immunometabolic biomarkers identify prognostic profiles under immunotherapy in advanced NSCLC: the SARC-LUNG study. Cancer Immunology, Immunotherapy. https://doi.org/10.1007/s00262-026-04565-y

Image Credits: AI Generated

DOI: 10.1007/s00262-026-04565-y

Keywords: non-small cell lung cancer, immunotherapy, immune checkpoint inhibitors, biomarkers, FGF21, sRAGE, GDF15, immunometabolism, body composition, cachexia, tumor microenvironment, prognosis