Ermelinda Damko, Sr. Scientist at Regeneron, shared a post on LinkedIn:
“Data‑Rich, Insight‑Poor — CXXXIII
LKB1‑mutant lung adenocarcinoma is a textbook example of a data‑rich, insight‑poor system. We have deep sequencing, immune profiling, and survival curves that all say the same thing: these tumors are aggressive, highly inflammatory on paper, and disproportionately resistant to immunotherapy. We have catalogued gene signatures, neutrophil and macrophage infiltrates, EMT programs, and poor T‑cell function. What has been missing is a simple, mechanistically coherent answer to a basic translational question: what, exactly, holds this immune‑evasive state together in a way that we can drug.
The recent Cancer Discovery study by Pillai et al. offers a candidate. It does not just add another layer of descriptive complexity; it nominates leukemia inhibitory factor (LIF) as the cytokine that ties LKB1 loss to a specific tumor cell state and a specific myeloid niche, and then shows that turning LIF off collapses that architecture. It is one of the few places in the LKB1 field where dense mechanistic data truly narrow the space of plausible explanations and yield a clean therapeutic axis.
From ‘inflamed but immune‑cold’ to a defined circuit
The standard description of STK11/LKB1‑mutant lung cancer is ‘inflamed but immune‑cold’: high myeloid signatures, poor T‑cell infiltration, bad outcomes, and weak responses to PD‑1‑based regimens. That framing is descriptively accurate but not operationally helpful. Many genotypes are inflamed; many are immune‑cold; the combination does not, by itself, tell you where to intervene.
Pillai et al. start with autochthonous Kras‑driven GEMMs in which Lkb1 is deleted in the lung epithelium and tumors are allowed to evolve in situ under native immune surveillance. This preserves intratumoral heterogeneity and the slow emergence of cell states that simple transplant models miss. In that environment, Lkb1‑mutant tumors show the expected enrichment of inflammatory, hypoxic, and EMT Hallmark gene sets, with increased IL‑6, IL‑33, CSF3 and chemokines (CCL2, CCL7, CXCL1, CXCL5, CXCL7) that recruit myeloid cells.
The pivot is that one cytokine, LIF, consistently sits at the intersection of genotype and phenotype. LIF mRNA is selectively upregulated in Lkb1‑mutant tumor cells; RNA in situ detects Lif transcripts only in mutant lesions; multiplex cytokine profiling of bronchoalveolar lavage fluid shows elevated LIF protein in mice bearing Lkb1‑mutant tumors; and human LUAD analyses confirm higher LIF expression in LKB1‑mutant tumors, where high LIF associates with inferior survival. Mechanistically, Lif transcription is controlled by the LKB1–salt‑inducible kinase–CRTC2 axis: loss of LKB1 or SIK activity derepresses CRTC2 and increases Lif, and this regulation is conserved in lung and pancreatic cancer models. That is the first key insight: LIF is not just a marker of ‘inflamed LKB1‑mutant tumors’; it is one way that LKB1 loss is translated into immune‑evasive biology.
A LIF‑dependent tumor cell state, not just ‘more inflammation’
The second key move in the paper is that it refuses to treat the tumor as a single LIF‑high mass. Using ExCITE‑seq and related single‑cell platforms, the authors decompose the tumor into multiple cell states and identify a specific Sox17‑positive EMT‑like, endoderm‑like state that depends on autocrine LIF signaling.
This state is defined by loss of Nkx2‑1, gain of Sox17, expression of stemness markers, and a strong inflammatory transcriptional program. Pseudotime analysis suggests a trajectory from more differentiated AT2‑like cells toward this Sox17‑positive, EMT‑like state, consistent with LIF‑driven dedifferentiation and endoderm‑like plasticity. Immunohistochemistry in human LUAD shows that SOX17 is expressed in a subset of LKB1‑mutant tumors and declines when Lif or Lifr is deleted in the GEMMs, whereas Keap1‑mutant tumors lack this SOX17 pattern, underscoring genotype specificity.
This matters because it turns a vague bulk descriptor—’LKB1‑mutant tumors are inflamed and plastic’—into a concrete object: a LIF‑dependent Sox17‑high tumor cell state that you can count, track, and eliminate. It gives you a targetable unit of resistance instead of an amorphous cloud of ‘more inflammation.’
How that state builds an immune‑evasive niche
The Sox17‑high state is not just a passenger. It produces IL‑6, CSF3, CCL2 and other mediators that recruit long‑lived SiglecF‑high neutrophils and Arg1‑high interstitial macrophages. Neutrophils in Lkb1‑mutant tumors express angiogenic and immunosuppressive programs; interstitial macrophages express Arg1, deplete extracellular arginine, and blunt CD4 and CD8 T‑cell effector cytokines. These myeloid populations expand early, before major differences in tumor burden, arguing against tumor size as the main driver and pointing back to tumor genotype and cell state.
The paper closes the loop with both mouse and human data. An Arg1‑high interstitial macrophage signature built from mouse single‑cell data is enriched in KRAS/LKB1‑co‑mutant human LUAD and stratifies LKB1‑wild‑type LUAD by survival. Depleting interstitial macrophages with clodronate reduces tumor burden and improves T‑cell effector function. T cells in Lkb1‑mutant tumors show impaired IFNγ and TNF responses, and this impairment reverses when the Arg1‑high niche is dismantled. The resulting causal chain is straightforward: LKB1 loss induces LIF, LIF stabilizes a Sox17‑high tumor state, that state constructs an Arg1‑rich myeloid niche, and that niche silences T‑cell immunity.
Mechanistically, this is where a large amount of data crystallizes into something usefully simple. Instead of ‘many things are wrong in LKB1‑mutant lung cancer,’ you get a single axis—LKB1–SIK–CRTC2–LIF–Sox17–myeloid niche—that explains why this genotype looks and behaves the way it does.
What blocking LIF actually buys you
The obvious next question is what happens if you cut the wire. Here, the wire is autocrine LIF–LIF receptor signaling.
Genetic deletion of LIF or its receptor in Lkb1‑mutant tumor cells reduces tumor burden, lowers intratumoral phospho‑STAT3, contracts the Sox17‑high EMT‑like states, and dampens IL‑6, CSF3 and CCL2 levels in bronchoalveolar lavage fluid. Deleting LIF in Lkb1‑wild‑type tumors has negligible impact, indicating that the axis is a genotype‑linked dependency, not a general feature of NSCLC.
At the microenvironment level, ExCITE‑seq shows that disrupting tumor‑intrinsic LIF signaling reduces SiglecF‑high neutrophils and Arg1‑high interstitial macrophages, increases interferon‑stimulated macrophage subsets compatible with antitumor responses, and boosts T‑cell infiltration and clonal expansion. Recombinant LIF does not directly suppress cytokine production in activated T cells, so the dominant immunosuppressive effect runs through the tumor–myeloid circuit rather than direct T‑cell signaling.
T‑cell depletion experiments align mechanism with therapy. In Lkb1‑mutant tumors with intact LIF signaling, depleting CD4 and CD8 T cells barely changes tumor burden because T cells are non‑functional in that context. In Lkb1‑mutant tumors where the LIF receptor has been deleted and the Sox17‑dependent state is gone, T‑cell depletion removes a large fraction of the antitumor effect. That is the point at which you can say that LIF blockade re‑enables T‑cell‑mediated control, rather than simply slowing tumor growth in a cell‑intrinsic way.
Therapeutic modalities targeting the LIF axis
The mechanistic story would be less compelling if it floated above an empty therapeutic landscape. In LKB1‑mutant NSCLC, however, LIF sits in a crowded but increasingly structured space of vulnerabilities.
The most direct modality is monoclonal antibody blockade of LIF itself. MSC‑1/AZD0171 (falbikitug), a humanized IgG1 anti‑LIF antibody, has completed a phase I first‑in‑human trial in advanced solid tumors, showing acceptable safety, pharmacokinetics, and evidence of STAT3 pathway inhibition and macrophage reprogramming. Phase II studies are now testing AZD0171 in resectable NSCLC, and in combination with durvalumab and platinum/taxane chemotherapy in advanced pancreatic adenocarcinoma. Preclinical work with a murine surrogate antibody shows that LIF blockade can repolarize macrophages, increase CX3CR1‑positive CD8‑positive effector T cells via a CX3CR1/CX3CL1 axis, and sensitize tumors to anti‑PD‑L1 plus chemotherapy.
Beyond direct LIF neutralization, the data suggest adjacent modalities:
- Targeting the LKB1–SIK–CRTC2 pathway upstream of LIF, using SIK inhibitors or modulators of CRTC2 activity, to control LIF and related inflammatory outputs.
- Combining anti‑LIF therapy with checkpoint inhibition in LKB1‑mutant NSCLC, explicitly framed as ‘state editing’: anti‑LIF to dismantle the Sox17‑myeloid niche, PD‑1/PD‑L1 blockade to exploit the newly restored T‑cell competence.
- Integrating LIF blockade with other LKB1‑linked vulnerabilities—CD38/NAD metabolism, MAPK/mTOR dependence, STING silencing, glucocorticoid receptor modulation—in combinations that address both immune and metabolic dimensions of LKB1‑mutant disease.
Crucially, these modalities differ in proximity to the clinic. Anti‑LIF antibodies already exist and have safety data; upstream pathway modulators and metabolic strategies are earlier. LIF is therefore one of the few points where the mechanistic circuit and the clinical toolkit already line up.
Challenges and limitations of targeting LIF
It is just as important to be explicit about what these data do not yet establish as it is to highlight what they do show.
First, LKB1‑mutant NSCLC remains heterogeneous. Not all LKB1‑mutant tumors will express high LIF; not all will manifest the same Sox17‑high state; and co‑mutations in KRAS, TP53, KEAP1 and STING components modulate the immune landscape independently. In human cohorts, STK11 mutations are consistently associated with poor prognosis and primary resistance to PD‑1 axis inhibitors, but the additive value of LIF expression or Sox17‑state signatures as clinical biomarkers is still being defined.
Second, the preclinical data make a strong case that anti‑LIF therapy is T‑cell‑dependent, but they do not yet tell us how that dependence will play out in heavily pre‑treated, stromally complex human tumors. In mouse models, T cells regain clonality and effector function when the Sox17‑myeloid state is dismantled; in patients, prior chemotherapy, radiation, and immunotherapy may have already exhausted or reshaped T‑cell repertoires in ways that limit this rebound.
Third, cytokine signaling networks are redundant. LIF is part of the IL‑6 family, and other cytokines can partially substitute for its functions in stemness, EMT, and immune modulation. The data show that LIF is upstream of IL‑6, CSF3 and CCL2 in Lkb1‑mutant tumors, but they do not guarantee that blocking LIF alone will prevent all routes to an immune‑evasive state in the long term. Adaptive resistance via alternative cytokines, myeloid programs, or stromal interactions is a realistic concern.
Fourth, clinical trial design itself is a bottleneck. Many ongoing NSCLC trials still enroll ‘all‑comers’ or stratify lightly by PD‑L1 and histology. To test the LIF axis properly, trials will need richer enrichment for STK11/LKB1 mutations, careful definition of co‑mutational backgrounds, and incorporation of LIF‑state biomarkers—LIF expression, phospho‑STAT3, Sox17‑state scores, Arg1‑high myeloid signatures—as prospective endpoints rather than purely exploratory correlative studies.
Finally, the current data are strongest in LKB1‑mutant lung cancer. LIF has been implicated in glioblastoma, breast, ovarian, and pancreatic cancers, and AZD0171 is being tested in pancreatic adenocarcinoma combinations, but whether those tumors use LIF in the same ‘state‑organizing’ way, or whether LIF is one axis among many, remains an open question.
Where this leaves LIF mechanistically
Seen mechanistically, the appeal of the Pillai et al. study is that it takes a noisy, multi‑omic picture and distills it into a circuit that is both biologically satisfying and clinically reachable. It does not claim that LIF is the answer to LKB1‑mutant lung cancer, or that anti‑LIF therapy will erase all of the genotype’s liabilities. What it does claim, and support with credible data, is that LIF sits at the center of a specific, immune‑evasive tumor cell state and its myeloid niche, and that blocking LIF can erase that state and restore T‑cell surveillance.
In a field where many discoveries are data‑rich but translationally thin, that is a non‑trivial advance. It gives you a simple, testable hypothesis: if a tumor’s resistance circuit runs through LIF‑dependent Sox17‑high states and Arg1‑rich myeloid niches, then LIF should be both a biomarker and a lever. If it does not, then LIF is just another elevated cytokine in a noisy transcriptome.”
Other articles about Lung Cancer on OncoDaily.