When a tumor runs short of nutrients and oxygen, it does not simply endure the stress. It weaponizes it. A study published in Nature from researchers at NYU Langone Health and collaborating institutions reveals that the cellular machinery cancer cells use to survive metabolic hardship also secretes a molecular signal that corrupts nearby immune cells, shielding the tumor from destruction. The protein at the center of this scheme, lipocalin 2 (LCN2), now has a target painted on it.
A Stress Response with Two Jobs
The integrated stress response (ISR) is an ancient cellular defense system. When a cell faces misfolded proteins, amino acid starvation, or mitochondrial dysfunction, it activates a cascade that converges on a single transcription factor: ATF4. ATF4 then switches on genes that help the cell adapt, survive, and grow. In cancer, this pathway is chronically active, and it has long been studied as a driver of therapy resistance.
What had not been fully appreciated was whether the ISR also shapes the immune landscape surrounding a tumor. Most prior work on ATF4 in cancer was conducted either in cell culture or in immunodeficient animals, conditions that make it impossible to see any effect on host immunity. The new study was designed specifically to close that gap.
The Immune System Does the Work
The core observation is stark. ATF4-deficient lung cancer cells grow significantly more slowly than their wild-type counterparts when transplanted into mice with intact immune systems. Transplant those same cells into NSG mice, which lack functional T, B, and NK cells, and the growth difference vanishes entirely. The tumor cells themselves show no difference in proliferation rates or apoptosis markers in either setting, ruling out any cell-intrinsic growth defect.
The same pattern held across three independent tumor models: the KP lung adenocarcinoma line, Lewis lung carcinoma, and B16F10 melanoma. In each case, ATF4 loss slowed tumors in immunocompetent hosts and had no effect in immunodeficient ones. Pharmacological inhibition of the ISR with ISRIB, a small molecule that blocks the pathway upstream of ATF4, reproduced the genetic result. ISRIB-treated mice bearing orthotopic lung tumors survived significantly longer than vehicle-treated controls (log-rank p = 0.0011), while the same drug had no effect in NSG mice.
The conclusion is unambiguous: ATF4 is not helping tumors grow by making cancer cells more fit. It is helping them grow by suppressing the immune system.
Core Finding
A Screen Points to a Single Culprit
To find which ATF4-regulated gene was responsible for the immune suppression, the team ran an in vivo pooled CRISPR screen. They built a custom library of 3,240 guide RNAs targeting 470 putative ATF4 target genes, infected KP cancer cells with the library, and transplanted those cells into both immunocompetent C57BL/6J and immunodeficient NSG mice. After twelve days, they recovered the tumors and asked: which guides were selectively lost in the immunocompetent setting?
The logic is elegant. A guide that depletes a gene needed for immune evasion will be purged from tumors growing in mice with working immune systems, because those tumor cells will be killed. The same guide will survive in NSG mice, where no immune pressure exists. Plotting the differential dropout score between the two hosts produces a ranked list of immune evasion genes.
At the top of that list, alongside Atf4 itself as a positive control, sat Lcn2, the gene encoding lipocalin 2. LCN2 is a small secreted glycoprotein previously known for its role in inflammatory responses and bacterial defense. Its appearance here as a principal mediator of cancer immune evasion was unexpected.
Screen Result
Secretion, Not Iron, Is the Key
Validation experiments confirmed the screen result across multiple models. Lcn2-knockout KP tumors grew more slowly than wild-type tumors in immunocompetent mice, and reintroducing LCN2 by lentiviral overexpression restored growth to wild-type levels. Overexpressing LCN2 in ATF4-deficient cells rescued their growth defect, placing LCN2 directly downstream of ATF4 in the immune evasion pathway. The same pattern appeared in orthotopic pancreatic ductal adenocarcinoma (PDAC) models and in B16F10 melanoma.
LCN2 is classically known as a siderophore-binding protein that sequesters iron from bacteria. A reasonable hypothesis was that its immunosuppressive function might work through the same iron-chelation mechanism. The authors tested this directly by engineering LCN2 mutants that cannot bind iron-catechol complexes. Those mutants suppressed immunity just as effectively as wild-type LCN2. A secretion-deficient mutant, by contrast, had no immunosuppressive activity at all.
The implication is clean: LCN2 must leave the cancer cell to do its damage, but once outside, it acts through a mechanism entirely independent of iron. That distinction matters for therapeutic design, because it means blocking LCN2's secreted form, rather than its iron-binding pocket, is the right target.
Corrupting the Macrophage
With LCN2 established as the key secreted factor, the question became: what does it do once it leaves the tumor cell? The answer required deep immune profiling. Using ExCITE-seq (a method that simultaneously captures the transcriptome and surface protein profile of single cells) on immune populations from tumor-bearing lungs, the team found that silencing LCN2 in tumors shifted the macrophage landscape substantially. The proportion of alveolar macrophages rose from 20% to 39%, while two populations of interstitial macrophages (IMs), the Spp1 and Cx3cr1 subtypes, fell from 35.9% and 28.7% to 16.4% and 18.2%, respectively.
Interstitial macrophages are the relevant players here. Gene set enrichment analysis showed that in LCN2-silenced tumors, IMs upregulated interferon-gamma response pathways and allograft rejection programs, while downregulating oxidative phosphorylation, a metabolic signature associated with immunosuppressive macrophage states. The functional readout was equally clear: LCN2 suppresses CXCL9 (a T cell-attracting chemokine) and promotes IL-6 (a cytokine associated with weakened anti-cancer immunity) in macrophages. Treating bone-marrow-derived macrophages with recombinant LCN2 protein in the presence of low-dose LPS reproduced both effects in vitro.
The receptor mediating this effect is SLC22A17. Using acoustic force spectroscopy to measure binding directly, the team showed that LCN2-coated beads bind to macrophages, and that siRNA knockdown of Slc22a17 abolishes this binding. Critically, Slc22a17 expression in the ExCITE-seq data was restricted almost entirely to macrophages, with negligible levels in other immune cell types. That specificity makes macrophages the primary cellular target of tumor-derived LCN2 in the tumor microenvironment.
Mechanism
The Human Picture
Mouse models are persuasive, but the clinical data in this paper carry independent weight. Analyzing tumor microarrays from 105 patients with lung adenocarcinoma, the team found that LCN2 staining intensity rose with tumor grade, a measure of clinical aggressiveness. High LCN2 staining was associated with the absence of tumor-infiltrating lymphocytes (TILs), and the two markers showed a striking spatial anti-correlation: regions of high LCN2 staining corresponded to regions of low CD3 staining.
The same pattern appeared in 33 PDAC biopsies, where LCN2 score and CD3 staining showed a negative correlation (R² = 0.28, p = 0.002). Nearest-neighbor spatial analysis of PDAC tissue confirmed that LCN2-positive cells were physically farther from T cells than LCN2-negative cells, supporting a model in which LCN2 creates local micro-domains of immune exclusion rather than acting systemically.
Across a database of 2,166 non-small-cell lung cancer samples, high LCN2 expression was associated with shorter overall survival, with a median of 52 months in the high-LCN2 group versus 79 months in the low-LCN2 group (HR = 1.28, 95% CI 1.13-1.45, p = 8.1 x 10⁻⁵). Among 976 patients treated with immune checkpoint inhibitors (anti-PD1, anti-PD-L1, or anti-CTLA4), the LCN2-high cohort fared worse, with a hazard ratio of 1.43 (95% CI 1.2-1.69, p = 5 x 10⁻⁵). A pan-cancer analysis of TCGA data showed that high LCN2 expression was associated with lower intra-tumoral to stromal T cell ratios and higher macrophage density across 23 epithelial solid tumor types.
Clinical Data
Turning Cold Tumors Hot
Because LCN2 is a secreted protein circulating in the tumor microenvironment, it is, in principle, an accessible antibody target. The team developed synthetic monoclonal antibodies against both mouse LCN2 (anti-mLCN2) and human LCN2 (anti-hLCN2), with each antibody binding its target with nanomolar affinity and showing no cross-reactivity to the other species' protein.
In female mice bearing orthotopic KP lung tumors, biweekly treatment with anti-mLCN2 at 10 mg/kg significantly reduced tumor progression. In an orthotopic PDAC model, the same antibody reduced final tumor weight (p = 0.0169). Neither treatment produced detectable toxicity: mouse weight was stable, blood cell counts across lymphocyte, neutrophil, monocyte, eosinophil, and basophil populations were unchanged, and post-mortem examination of liver, heart, spleen, and kidney showed no gross abnormalities.
To confirm that the therapeutic effect was specifically due to neutralizing tumor-derived LCN2 rather than host-derived LCN2, the team engineered a chimeric system: Lcn2-knockout KP cells were made to express human LCN2, then transplanted into mice and treated with anti-hLCN2. This antibody targets only the tumor-derived human protein, leaving host mouse LCN2 untouched. The result was the same: tumor growth was suppressed, and CD4+ and CD8+ T cell infiltration increased (p = 0.0009 for CD8+ cells). Cytokine analysis of bronchoalveolar lavage fluid showed lower IL-6 and higher IFN-gamma concentrations in treated mice, consistent with a shift toward a pro-inflammatory macrophage state.
The combination experiment is where the therapeutic logic becomes most compelling. KP lung tumors are immunologically cold and refractory to anti-PD1 therapy alone. Anti-PD1 treatment did not significantly extend survival in this model. Anti-LCN2 treatment alone extended median survival to 24.5 days from a control of 19 days. The combination of anti-LCN2 and anti-PD1 extended median survival to 26 days. The interpretation is that LCN2 neutralization recruits T cells into the tumor, and anti-PD1 then prevents those T cells from being shut down by checkpoint signals, a sequential logic that mirrors how combination immunotherapy is being developed more broadly.
Therapeutic Data
What the Data Support and Where Gaps Remain
The mechanistic chain here is well-constructed. The genetic screen is a strong starting point because it is unbiased and conducted in vivo, where immune pressure is real. The validation across multiple tumor types (lung, pancreatic, melanoma) and multiple genetic and pharmacological perturbations (ATF4 knockout, LCN2 knockout, ISRIB, inducible shRNA) builds a robust case. The ExCITE-seq data add granularity that flow cytometry alone cannot provide, and the acoustic force spectroscopy binding assay is a direct, physical confirmation of the LCN2-SLC22A17 interaction on macrophages.
The clinical correlations are consistent and statistically strong, but they are correlational. High LCN2 expression may co-occur with other features of aggressive tumors that independently exclude T cells. The authors acknowledge this, and the spatial nearest-neighbor analysis in PDAC tissue is a reasonable attempt to show that LCN2's effect is locally confined rather than a proxy for overall tumor aggressiveness.
The combination therapy result (26 days median survival versus 19 days for control) is a modest absolute extension in a mouse model. That number should not be over-interpreted as a prediction of clinical effect size. What it does show is proof-of-concept that LCN2 neutralization can sensitize a cold tumor to checkpoint blockade, which is the mechanistic claim the authors are making.
Two open questions stand out. First, the molecular details of how LCN2 binding to SLC22A17 reprograms macrophage transcription are not resolved. The authors flag this explicitly. Second, the therapeutic window of systemic LCN2 neutralization in humans is unknown. LCN2 has established roles in antibacterial defense and inflammatory regulation; blocking it systemically could have consequences that mouse experiments, conducted over weeks, would not detect. The authors' suggestion of tumor-selective antibody engineering is worth taking seriously.
A New Target in the Immunotherapy Landscape
The broader significance of this work is that it adds a new layer to how we think about the tumor microenvironment. The ISR has been studied as a cell-intrinsic survival mechanism for years. This paper reframes it as a communication system: stressed cancer cells broadcast a signal, via LCN2, that corrupts the immune cells in their neighborhood. That reframing opens a therapeutic angle that was not previously visible.
For the field of immunotherapy, the most pressing problem is the large fraction of solid tumors that do not respond to checkpoint inhibitors. Lung and pancreatic cancers, the two tumor types most extensively studied here, are among the most refractory. A strategy that converts immune-excluded tumors into inflamed ones, by blocking the signal that drives exclusion, addresses that problem at its root rather than trying to amplify an immune response that was never recruited in the first place.
The path from these mouse experiments to a clinical antibody is long, and the toxicity profile of LCN2 neutralization in humans will need careful evaluation. But the target is secreted, the antibodies are specific, and the mechanism is now defined with enough resolution to guide that development. The ISR-ATF4-LCN2 axis is a real and targetable pathway of immune evasion, and the case for pursuing it clinically is now substantially stronger.