Does a CCL21 dendritic cell vaccine plus pembrolizumab shrink NSCLC?

A vaccine designed to pull T cells into tumors

The trial tested whether injecting a dendritic cell vaccine directly into a lung tumor could make pembrolizumab work in patients whose disease had already progressed on immune checkpoint inhibitors. Dendritic cells present antigens to T cells; here they were grown from each patient's own blood monocytes and transduced with an adenoviral vector carrying CCL21, a chemokine intended to attract T cells and bring them together with antigen-presenting cells. The hypothesis was that this local recruitment would strengthen tumor-antigen presentation and T cell activation, giving pembrolizumab more to work with.

The design was a single-center, non-randomized phase I study with dose escalation and expansion. The vaccine was injected into the same lung tumor on days 0, 21 and 42 by bronchoscopy or CT-guided transthoracic injection, while pembrolizumab was given intravenously at 200 mg every three weeks for up to one year or until progression or unacceptable toxicity. The primary objectives were the maximum administered dose during escalation and the objective response rate during expansion.

Patients who had already failed checkpoint blockade

Twenty-three patients with advanced or metastatic NSCLC were treated between July 2019 and January 2024. Eleven entered dose escalation, three at 1 × 10^7 cells per injection and eight at 3 × 10^7 cells per injection, and 12 were treated in expansion at the maximum administered dose of 3 × 10^7 cells per injection. Nineteen patients received at least two injections at that dose and formed the efficacy-evaluable subset.

This was a deliberately hard-to-treat population. Patients without EGFR or ALK alterations had to have progressed after at least one prior PD-1 or PD-L1 inhibitor, and those with sensitizing EGFR or ALK alterations had to have progressed on a tyrosine kinase inhibitor. In the full cohort, 21 of 23 patients (91%) had received chemotherapy. The paper notes that such patients would be unlikely to benefit from further checkpoint inhibitor treatment alone, so the trial was asking whether the vaccine could overcome that resistance.

No objective responses, but some stable disease

The trial did not meet its primary endpoint for objective response rate: among the 19 efficacy-evaluable patients there were no confirmed objective responses, with a 95% confidence interval of 0.0–17.6%. The disease control rate was 36.8%, meaning seven patients had stable disease as their best response, and no tumor shrank by 30% or more. After a median follow-up of 32.3 months, median progression-free survival was 2.1 months (95% CI 1.2–3.6) and median overall survival was 6.9 months (95% CI 4.0 months to not evaluable).

Stability was not evenly distributed. Tumor size stability was observed more often in lesions injected with the vaccine than in other target lesions, which is a local observation rather than a systemic response. The longest disease control lasted 12 months in a patient who completed the planned year of treatment; that patient had 80% tumor PD-L1 expression and a KRAS G12D mutation and had previously responded to single-agent pembrolizumab before progressing. Another patient with stable disease had an untreated 4 mm brain metastasis that was not visible on repeat MRI three months later, while a separate sacroiliac mass progressed. These are individual case descriptions, not evidence of a general response pattern.

Tolerability held up across the dose levels

No dose-limiting toxicities occurred during dose escalation, and the maximum administered dose of 3 × 10^7 cells per injection was carried into expansion. Adverse events of any cause occurred in 19 patients (83%); the most common any-grade events were dyspnea (35%), fatigue (30%) and cough (17%). Grade 3 or higher adverse events occurred in five patients (26%), including two grade 4 events and one grade 5 event of pneumonia.

Treatment-related adverse events occurred in four patients (17%). One had grade 1 flank pain associated with the injection, and three had events considered immune-mediated, including one grade 3 event (4%) of immune-related diabetes mellitus. The authors describe this as a favorable safety profile supporting the feasibility of the combination, but the small cohort limits how precisely any toxicity rate can be estimated.

More CD4+ T cells and new clones, briefly

Post-hoc analyses of biopsies taken on days 0, 21 and 42 examined what the vaccine actually did inside tumors. Single-cell RNA sequencing combined with T cell receptor sequencing profiled 65,344 cells from 41 samples across 17 patients, and 40,005 T cells were subclustered into 12 groups. The therapy increased tumor infiltration by CD4+ T cells and by novel T cell clones, and the increase in T cell infiltration within a tumor correlated with the amount of CCL21 secreted by the specific vaccine batch used for that tumor.

Disease stability tracked with two immune features. Tumors from patients with local disease stability showed a decrease in mutant-allele tumor heterogeneity, a score derived from whole exome sequencing that the authors interpret as more effective immunoediting and clonal pruning. T cell receptor diversity, measured by the inverse Simpson index, rose significantly in stable-disease samples but not in progressive-disease samples. Among clones that persisted to day 42, clonal proportion rose a mean 5.2-fold, and this expansion was larger in stable-disease patients (7.1-fold) than in progressive-disease patients (4.6-fold, p = 0.0001).

New clones appeared but mostly did not persist

The central tension is that immune activity was measurable while clinical benefit was not. Many clones that emerged on day 21 were absent by day 42; clones found on day 42 were 2.7 times more likely to have been present on day 21 than the reverse, which the authors say is consistent with clonal loss. Persistent clones were more likely than transient ones to express gene signatures of neoantigen-specific T cells and to shift toward effector or exhausted phenotypes, but they remained low in number overall.

The authors' explanation is specificity. T cells migrating along a CCL21 gradient would not necessarily be tumor-specific, so the infiltrate may have been dominated by bystander clones that were not selected for tumor antigens and were therefore vulnerable to neglect. Consistent with that, an initial rise in T cell infiltration did not produce a persistent cytotoxic CD8+ T cell signature in many patients. Immunosuppressive myeloid signaling signatures were also identified across time points, and peripheral blood showed minimal changes in T cell receptor diversity or transcriptional profiles, which the authors read as further evidence that robust tumor-specific clones were not generated systemically.

Correlative immune findings without a control arm

The immune comparisons come from post-hoc analyses of a small, single-arm study. The authors state that without a control arm the associations between clinical outcomes and immunologic findings remain correlative rather than mechanistic, and short-term disease control may instead reflect underlying disease biology. Tissue quality and quantity varied, limiting statistical power, and the number of patients contributing to each analysis was smaller than the full cohort: 15 patients for the heterogeneity analysis, 10 for T cell receptor diversity, and 12 across 28 samples for the myeloid analysis.

Two further limits matter for interpretation. The dendritic cell product varied between patients, particularly in CCL21 secretion, which could introduce variability in immune responses. And the study did not directly confirm that any T cell clone was neoantigen-specific; the neoantigen signature analysis assigns probabilities rather than proving specificity. The authors propose that future work validate T cell specificity, for example by testing whether cloned T cell receptors respond to predicted neoantigens.

Stable disease clusters with KRAS mutations

The most useful split in this paper is not response versus no response, since no objective responses occurred, but stable disease versus progression. The stable-disease group was enriched for KRAS-mutant tumors, including co-mutations in STK11 or KEAP1, while the progressive-disease group most commonly carried TP53 mutations. The patients with the most striking clinical courses, the 12-month disease control and the untreated brain lesion that resolved, also had the largest shifts in immune metrics such as T cell receptor diversity.

That pattern points to a specific priority rather than a general call for more vaccine studies. If the vaccine's effect depends on recruiting T cells that are already tumor-relevant, then the variable worth measuring is the specificity and persistence of the induced clones, not simply the amount of infiltration. The trial shows that intratumoral CCL21-DC plus pembrolizumab can increase CD4+ T cell infiltration and perturb clonal dynamics in a population that had progressed on checkpoint blockade, but those changes were largely transient and did not translate into tumor shrinkage. The next test is whether the minority of persistent, neoantigen-signature-bearing clones can be made the majority.

Sources used in this answer