Inhalable extracellular vesicle delivery of IL-12 mRNA to treat lung cancer and promote systemic immunity

Abstract

Lung carcinoma is one ofthe most common cancers and has one of the lowest survival rates in the world. Cytokines such as interleukin-12 (IL-12) have demonstrated considerable potential as robust tumour suppressors. However, their applications are limited due to off-target toxicity. Here we report on a strategy involving the inhalation of IL-12 messenger RNA, encapsulated within extracellular vesicles. Inhalation and preferential uptake by cancer cells results in targeted delivery and fewer systemic side effects. The IL-12 messenger RNA generates interferon-γ production in both innate and adaptive immune-cell populations. This activation consequently incites an intense activation state in the tumour microenvironment and augments its immunogenicity. The increased immune response results in the expansion of tumour cytotoxic immune effector cells, the formation ofimmune memory, improved antigen presentation and tumour-specific T cell priming. The strategy is demonstrated against primary neoplastic lesions and provides profound protection against subsequent tumour rechallenge. This shows the potential for locally delivered cytokine-based immunotherapies to address orthotopic and metastatic lung tumours.

Immune checkpoint blockade therapies hold promise for inducing tumour regression^1,2 but often fall short due to the immunosuppressive tumour microenvironment (TME)^3,4. This limitation has spurred the development of next-generation immunotherapies involving recombinant cytokines or chemokines. These agents could facilitate T cell recruiting and convert immunologically ‘cold’ tumours into ‘hot’ ones^4,5. Notably, interleukin-12 (IL-12) stimulates interferon-γ (IFNγ) production^6,7 and augments the cytolytic potential of immune cells⁸, bridging innate and adaptive immune responses⁷. Despite the immune response of these proteins’ monotherapy or the combination with other immunotherapies^9,10, systemic cytokine administration can induce severe side effects¹¹, as observed in early clinical trials^8,12. Therefore, localized TME delivery routes are imperative.

Leveraging messenger RNA (mRNA) ensures local translation¹³ and in situ vaccination^14,15 without risks associated with DNA, such as genomic integration¹⁶. Intratumoral (i.t.) injection of IL-12-encoding mRNA^8,11,14, a promising local delivery strategy, has been translated into clinical trials¹⁷. However, direct lesion delivery in internal organs remains a challenge¹⁸. Various mRNA delivery strategies, from lipid^14,19 to polymeric nanoparticles¹⁸, have been probed, but off-target effects and toxicity persist as concerns²⁰. Extracellular vesicles, especially exosomes, are increasingly harnessed as natural mRNA delivery systems^21–24. Pulmonary delivery through inhalation presents an effective, non-invasive local administration route²⁵, as we demonstrated previously for treating pulmonary fibrosis²⁵ and COVID-19 (refs. 26,27).

In this study, we developed inhalable extracellular vesicles loaded with IL-12 mRNA to address lung cancer and bolster systemic immunity in lung tumour-bearing mouse models. IL-12 mRNA was loaded into human embryonic kidney cell-derived exosomes (HEK-Exo) through electroporation²⁰, yielding IL-12 mRNA-loaded exosomes (IL-12-Exo). On inhalation by mice with lung tumours, IL-12-Exo outperformed IL-12 mRNA-loaded liposomes (IL-12-Lipo) in the TME biodistribution and minimized systemic toxicity. These inhaled IL-12-Exo promoted IFNγ-mediated immune activation, systemic immunity and immune memory, culminating in lung tumour suppression and heightened resistance against tumour rechallenges.

Preparation and characterization of IL-12-Exo

On the basis of The Cancer Genome Atlas database, we first explored the relationship between IL-12 expression and prognosis for patients with cancer through bioinformatics analysis. Our findings highlighted the positive correlation between elevated IL-12 expression and enhanced survival outcomes, especially in patients with lung cancer (P = 0.019) (Supplementary Fig. 1), suggesting the favourable modulating of IL-12 on antitumour treatment. Here, we used the LL/2 lung carcinoma cell line to establish an orthotopic lung cancer model (Supplementary Fig. 2a)²⁸. We noted an increase in macrophage number in lung TME (Supplementary Fig. 2b,c), which may shape immunosuppressive TME²⁹.

HEK-Exo with minimal immunogenicity in mice³⁰ serve as mRNA loading platforms^22,31–33. We prepared IL-12 mRNA³⁴ and developed IL-12-Exo using electroporation (Fig. 1a)³³. Concurrently, IL-12 mRNA (Supplementary Fig. 3a) loaded liposomes (IL-12-Lipo) were prepared for comparative analysis. Exosomal proteins were characterized through western blot (Supplementary Fig. 3b). Moreover, both IL-12-Exo and IL-12-Lipo exhibited analogous nanoscale distributions (Fig. 1b), and their morphologies were corroborated by transmission electron microscopy (Fig. 1c). Both formulations showed negative surface charges (−9.6 ± 1.0 mV for IL-12-Exo and −2.3 ± 1.1 mV for IL-12-Lipo) (Supplementary Fig. 3c). We measured the encapsulation efficiency^35,36 of IL-12 mRNA and found electroporation yielded similar encapsulation efficiency levels for both formulations (27.6% for IL-12-Exo and 28.8% for IL-12-Lipo) (Supplementary Fig. 3d).

Fig. 1 |. In vitro characterization of IL-12-Exo and in vivo distribution in LL/2 tumour-bearing mice.

a, Schematic showing IL-12 mRNA loading into HEK-Exo (IL-12-Exo) or liposomes (IL-12-Lipo), followed by nebulized inhalation administration to LL/2 tumour-bearing mouse lungs. LL/2 tumour cells were i.v. injected into C57BL/6 mice for orthotopic lung tumour model establishment. UTR, untranslated region. b,c, NanoSight size distribution analysis (b) and transmission electron microscopy imaging (c) of IL-12-Exo and IL-12-Lipo (experiments were replicated in triplicate). Scale bars, 200 nm. d, Representative ex vivo images and quantitative analysis of mouse major organs that received fluorescence-labelled IL-12-Exo and IL-12-Lipo 24 h postadministration (n = 5 biologically independent mice per group). e, Representative immunostaining images of tumour-bearing mouse lungs. IL-12-Exo and IL-12-Lipo were stained with DiD (grey) before inhalation. LL/2 tumour cells and macrophages were stained with anti-luciferase antibody (green) and anti-F4/80 antibody (red), respectively. Nuclei were stained with DAPI (blue). Scale bar, 100 ¼m. White arrowheads indicated merged signals of IL-12-Exo or IL-12-Lipo with LL/2 tumour cells, yellow arrowheads indicated merged signals of IL-12-Exo or IL-12-Lipo with macrophages. Experiments were replicated in triplicate. f, Flow cytometry showing cellular uptake percentage of IL-12-Exo or IL-12-Lipo in total lung cells 24 h after drug administration. g–j, Flow cytometry showing the proportion of IL-12-Exo or IL-12-Lipo uptake by LL/2 tumour cells (g), macrophages (F4/80⁺CD11b⁺) (h), epithelial cells (CD45⁻CD31⁻EpCAM⁺) (i) and DCs (CD45⁺CD11c⁺CD24⁺) (j), each compared to total lung cells 24 h after drug administration. n = 9 biologically independent mice per group for f–h, n = 3 biologically independent mice per group for i,j. k,l, The uptake distribution of IL-12-Exo (k) or IL-12-Lipo (l) in various cell types. Experiments were replicated in triplicate. P values were determined by two-way ANOVA post-Bonferroni’s multiple comparison test (d) and two-tailed unpaired Student’s t-test (f–j) using GraphPad PRISM software. Exact P values are indicated. Results are presented as means ± s.d. Min., minimum; max., maximum. Schematic in a and mouse cartoon created with BioRender.com.

Superior lung distribution and tumour specificity of IL-12-Exo

Systemic toxicity associated with IL-12 remains a major hurdle for its therapeutic application in cancer. Previous research has reported elevated serum IL-12 concentrations leading to hepatic toxicity and weight loss in tumour-bearing mice on intravenous (i.v.) injection¹⁴. On the basis of previous studies demonstrating effective payload delivery via inhaled HEK-Exo to healthy lungs²⁵, we evaluated the targeting capability of inhaled IL-12-Exo to the lung TME. On inhalation of either dye-labelled IL-12-Exo or IL-12-Lipo, ex vivo imaging revealed their marked preference for lung enrichment over other organs 24 h postinhalation. IL-12-Exo presented a 1.54-fold higher mean fluorescence intensity than IL-12-Lipo (Fig. 1d), coupled with diminished distribution in off-target organs involving the liver and kidneys (Fig. 1d). Successful delivery of IL-12-Exo into tumour cells was further confirmed by immunofluorescent staining (Fig. 1e). Flow cytometry indicated a surpassed uptake of IL-12-Exo (24.80%) compared to IL-12-Lipo (16.71%) among total cells (Fig. 1f), with LL/2 tumour cells being the primary cells internalizing IL-12-Exo (67.74%, Fig. 1g–k). For IL-12-Lipo, LL/2 uptake was less than half of the total cells, with macrophages and epithelial cells being the main consumers for the remainder (Fig. 1g–j,l). Minimal uptake by dendritic cells (DCs) was observed for both particles (Fig. 1j). We found low levels of particles in circulation relative to lung enrichment (Supplementary Fig. 4).

We further investigated the in vitro uptake dynamics in LL/2 tumour cells and inflammatory macrophages³⁷. Both particles displayed time-dependent internalization, with IL-12-Exo presenting superior uptake in LL/2 tumour cells but comparable uptake in macrophages (Supplementary Fig. 5). A trans-well experiment further confirmed the ability of both particles to penetrate mucus³⁸, a key barrier for lung delivery (Supplementary Fig. 6). Notably, their demonstrated anionic surface (Supplementary Fig. 3c) minimized entrapment within the negative charged mucus.

IL-12-Exo safely encode IL-12 cytokine in the lung TME

We explored IL-12 protein expression (Fig. 2a) in LL/2 tumour cells, the primary target of IL-12-Exo (Fig. 1k). Notably, LL/2 cells transfected with IL-12-Exo displayed substantial IL-12 levels, in contrast to the insignificant cytokine presence with PBS or free HEK-Exo treatments (Fig. 2b). Turning to in vivo encoding of IL-12, LL/2 cell-induced tumours, known for poor immunogenicity³⁹, showed scarce IL-12 levels in the lungs post-treatment with PBS, Mock-Exo or Mock-Lipo. However, IL-12-Exo administration instigated a robust, sustained IL-12 expression for a minimum of 3 days, outperforming IL-12-Lipo (Fig. 2c). The latter, in contrast, showed a swift spread of IL-12 in the blood, liver and kidneys, with concomitant raised serum expressions of aspartate transaminase (AST) levels (Fig. 2d–g). All treatments did not affect the expression of alanine aminotransferase (ALT) (Supplementary Fig. 7). The clinical toxicity of IL-12 therapy is frequently associated with IL-12-driven IFNγ production⁴⁰. Our data revealed that localized IL-12 enrichment contributed to diminished serum IFNγ levels. Conversely, mice treated with IL-12-Lipo exhibited elevated serum IFNγ concentrations (Fig. 2h), possibly because leaked and dispersed IL-12-activated blood cells in circulation and triggered systemic IFNγ production^14,41. These results underscored IL-12-Exo’s proficiency in bolstering localized IL-12 expression in the TME, thereby evading off-target toxicity and improving systemic cytokine tolerability. After 3 days of IL-12 administration, immune stimulation within the lung TME (illustrated in Fig. 2i) revealed enhanced influx of CD8⁺ T cells with IL-12-Exo over IL-12-Lipo, accompanied by increased recruitment ofCD4⁺ T cells, natural killer (NK) cells and monocytes (Fig. 2j).

Fig. 2 |. Inhaled IL-12-Exo safely stimulates local IL-12 expression in the lung TME.

a, Schematic depicting the assessment of IL-12 expression in vitro (24 h) and in vivo (1 and 3 days). C57BL/6 mice were injected with LL/2 tumour cells for tumour inoculation and various treatments 6 days post-LL/2 injection. ELISA was conducted 1 or 3 days, respectively, after treatments. b, IL-12 expressions in LL/2 tumour cells 24 h after incubation (n = 3 biologically independent samples). c, IL-12 expressions in tumour-bearing lungs (n = 7 for IL-12-Lipo and IL-12-Exo groups, n = 4 biologically independent mice for other groups). ‘Mock-Exo’ in b and c means HEK-Exo lacking IL-12 mRNA encapsulation, while ‘Mock-Lipo’ in c means liposomes lacking IL-12 mRNA encapsulation. d, IL-12 expressions in blood circulation after treatments. e,f, IL-12 expressions in organs, including the heart, liver, spleen, kidney, 1 day (e) or 3 days (f) after treatments. g, AST levels in blood circulations 3 days after treatments. h, IFNγ expressions in blood circulation 1 or 3 days after treatments. n = 4 biologically independent mice per group in d–h. ND, not detected. i,j, Schematic (i) and quantified flow cytometry analysis (j) of immune cell populations infiltrating lung tumours 3 days after treatments. These cells included MDSCs (CD45⁺CD11b⁺Gr-1⁺), macrophages (F4/80⁺CD11b⁺), monocytes (CD45⁺CD11b⁺MHCII⁻CD64^+/−Ly6C^low), CD8⁺ T cells (CD45⁺CD3⁺CD8⁺), CD4⁺ T cells (CD45⁺CD3⁺CD4⁺), NKT cells (CD45⁺CD3⁺NK1.1⁺), NK cells (CD45⁺CD3⁻NK1.1⁺), CD103⁺ DCs (CD45⁺CD11c⁺CD24⁺CD103⁺) and CD11b⁺ DCs (C D45⁺CD11c⁺CD24⁺MHCII⁺CD64⁻CD11b⁺) (n = 6 biologically independent mice per group). P values were determined by one-way ANOVA post-Bonferroni’s multiple comparison test (b,g), Two-way ANOVA post-Bonferroni’s multiple comparison test (c–f,h,j) using GraphPad PRISM software. Exact P values are indicated. Results are presented as means ± s.d. Schematic in a and mouse cartoons created with BioRender.com.

IL-12-Exo therapy effectively retards tumour growth

On confirming the preferential residence and sustained retention of IL-12 in the lung TME, we evaluated the antitumour capabilities of IL-12-Exo in LL/2 homograft lung tumour models (depicted in Fig. 3a). IL-12-Exo dramatically suppressed tumour growth in the lungs, exhibiting superior antitumour efficacy compared to IL-12-Lipo (Fig. 3b and Supplementary Fig. 8). Exo-treated mice displayed a moderate weight gain, contrasting with the noticeable weight loss seen in IL-12-Lipo group (Fig. 3c). With diminished tumour burden, IL-12-Exo-treated mice had correspondingly decreased lung weights compared to IL-12-Lipo counterparts (Fig. 3d). In terms of survival rates, the onset of mortality in the IL-12-Exo group was concomitantly delayed between days 30 and 50 after tumour challenge, whereas IL-12-Lipo showed onset between days 28 and 35. PBS-treated mice experienced earliest mortality, averaging a 21-day lifespan (Fig. 3e). Off-target effects of IL-12 include severe off-tumour toxicity and even mortality⁴². To determine whether cancer-related tumour growth was responsible for the observed mortalities of mice undergoing IL-12 treatments, we recorded lung weights at the endpoint of a survival study (illustrated in Fig. 3a). Throughout the observation period, almost all endpoint lungs exhibited severe tumour burden (Supplementary Fig. 9). While IL-12-Lipo led to a spike in serum AST expression, ALT levels remained consistent among treatments (Supplementary Fig. 10a,b).

Fig. 3 |. Local delivery of IL-12-Exo retards tumour growth in orthotopic lung tumours.

a, Schematic illustrating the antitumour efficiency assessments after treatments. b,c, Statistical tumour growth (b) and body weight (c) over time for each treatment group. The normalized tumour growth rate was determined by comparing the fold-changes over initial of the tumour volume (V/V₀), as characterized by the bioluminescence intensity from LL/2 tumour cells (n = 11 biologically independent mice for PBS group, n = 9 biologically independent mice for IL-12-Lipo group and n = 12 biologically independent mice for IL-12-Exo group). d, Lung weights 30 days after tumour inoculation (n = 7 biologically independent mice for PBS group, n = 8 biologically independent mice for IL-12-Lipo group and n = 9 biologically independent mice for IL-12-Exo group). e, Survival curves for mice in each treatment group (n = 10 biologically independent mice per group). f, Schematic showing the establishment of the B16F10 melanoma lung metastatic tumour model and anti-metastatic tumour assessments after mice were administered with different treatments. g,h, Statistical metastatic tumour growth (g) and body weight (h) over time for mice in each treatment group. The normalized tumour growth rate was calculated by comparing the fold-changes over initial of the tumour volume (V/V₀), as characterized by the bioluminescence intensity from B16F10 tumour cells (n = 6 biologically independent mice per group). i, Representative lung morphologies 21 days after B16F10 tumour inoculation. Black points indicate the metastatic foci. Results were gained from mouse studies where n = 6 biologically independent mice per group. j, Lung weights 21 days after B16F10 tumour inoculation (n = 3 biologically independent mice for PBS group, n = 5 biologically independent mice for IL-12-Lipo and IL-12-Exo groups). k, mRNA expressions of Typ1, a melanoma metastasis, 21 days post-tumour inoculation (n = 3 biologically independent mice per group, with three independent samples for each mouse). l, Survival curves of mice for each treatment group (n = 9 biologically independent mice per group). P values were determined by two-way ANOVA post-Bonferroni’s multiple comparison test (b,c,g,h), one-way ANOVA post-Bonferroni’s multiple comparison test (d,j,k), or one-side log-rank (Mantel–Cox) test (e,l) using GraphPad PRISM software. Exact P values are indicated. Results are presented as means ± s.d. Cartoon mice created with BioRender.com.

For metastatic lung tumours, we established a B16F10 cell-induced lung metastatic melanoma model and IL-12-Exo administration (outlined in Fig. 3f). In comparison to IL-12-Lipo, IL-12-Exo prominently suppressed tumour progression and displayed modest weight gain while reducing lung weights and metastatic loci (Fig. 3g–j). IL-12-Exo therapy further accompanied with serum chemistry normalization (Supplementary Fig. 10c,d). The expression of tyrosinase-related protein 1 (Tyrp1), a melanocyte-specific gene⁴³, reflects a quantitative measurement of B16F10 tumour metastasis and development in the lung⁴⁴. We observed a heightened mRNA expression of Tyrp1 in B16F10 tumour-infected lungs but mitigated by IL-12-Exo (Fig. 3k). Owing to the therapeutic benefits of IL-12-Exo, its treated mice outlived IL-12-Lipo recipients (Fig. 3l). To ascertain whether the tumour suppression induced by IL-12-Exo was specific to the B16F10 lung metastatic tumour, we established a breast cancer lung metastasis model by i.v. injecting 4T1 cells into mice. IL-12-Exo also demonstrated efficacy on these 4T1 tumour-bearing mice, prolonging survival (Supplementary Fig. 11a,b).

IL-12-Exo therapy-induced IFNγ drives TME regulation

IL-12 is known to activate cells within both innate and adaptive immune compartments, predominantly through IFNγ (ref. 8). In patients with melanoma who were treated with recombinant human IL-12, a sustained and potent IFNγ overexpression correlated with favourable clinical outcomes⁴⁵. We probed whether IFNγ was implicated in IL-12-Exo activity. In synchrony with IL-12 secretion (Fig. 2c), IFNγ was rapidly upregulated 1 day postadministration of IL-12-Exo and IL-12-Lipo and maintained high levels until day 3 (Fig. 4a). The IFNγ response with IL-12-Exo surpassed that of IL-12-Lipo, mirroring IL-12 production trends (Fig. 2c). Survival assessments¹¹, using an IFNγ-neutralizing antibody, highlighted the essentiality of IFNγ (illustrated in Supplementary Fig. 11c). Inhibiting IFNγ negated IL-12-Exo’s antitumour activity, equating survival rates of the PBS group to the IL-12-Exo plus antibody cohort (Fig. 4b). These results demonstrated the critical role of IFNγ in IL-12-Exo therapy.

Fig. 4 |. IFNγ is essential for IL-12-Exo-mediated antitumour effects.

a, IFNγ expressions in LL/2 tumour-bearing lungs after 1 or 3 days. C57BL/6 mice were injected with LL/2 tumour cells for tumour inoculation and treated with IL-12-Exo, IL-12-Lipo or PBS 6 days after LL/2 injection (schematic in Fig. 2a). ELISA was performed 1 or 3 days after different treatments (n = 4 biologically independent mice for PBS group, n = 7 biologically independent mice for IL-12-Lipo and IL-12-Exo groups). b, Survival curves for mice treated with of PBS, IL-12-Exo or IL-12-Exo in combination with an anti-IFNγ antibody. PBS or IL-12-Exo was inhaled 6, 9 and 12 days post-tumour inoculation, while the indicated depleting anti-IFNγ antibody was injected beginning 1 day before IL-12-Exo inhalation (n = 10 biologically independent mice for PBS group and n = 8 biologically independent mice for the other two groups). c, Schematic showing the TME alteration assessment after mice were administered with IL-12-Exo, IL-12-Lipo or PBS. d, Quantified flow cytometry analysis of CD8⁺ T cells (CD45⁺CD3⁺CD8⁺), NKT cells (CD45⁺CD3⁺NK1.1⁺), NK cells (CD45⁺CD3⁻NK1.1⁺), CD4⁺ T cells (CD45⁺CD3⁺CD4⁺), T_reg cells (CD45⁺CD4⁺Foxp3⁺) and MDSCs (CD45⁺CD11b⁺Gr-1⁺) in the TME 30 days after tumour inoculation (n = 6 biologically independent mice for PBS and IL-12-Lipo groups and n = 8 biologically independent mice for IL-12-Exo group). e, mRNA expressions in the TME 30 days after tumour inoculation (n = 6 biologically independent mice per group). f, Luminex analysis of protein levels in LL/2 tumour-bearing lungs 30 days after tumour inoculation (n = 5 biologically independent mice per group). Clustered protein levels are shown as a heat map. P values were determined by two-way ANOVA post-Bonferroni’s multiple comparison test (a), one-way ANOVA post-Bonferroni’s multiple comparison test (d,e) or one-side log-rank (Mantel–Cox) test (b) using GraphPad PRISM software. Exact P values are indicated. Results are presented as means ± s.d. Cartoon mouse created with BioRender.com.

IFNγ assists IL-12 in mediating immune responses. Then 30 days after tumour inoculation, we determined the immunomodulatory effects of IL-12-Exo in the poorly immunogenic lung TME seeded by LL/2 cells (illustrated in Fig. 4c)³⁹. The infiltration of antitumour lymphocyte is essential to transform this ‘cold’ tumour to a ‘hot’ one. IL-12 is known to provide pivotal signals for the functional activation of lymphocytes, especially T cells and NK cells⁴⁶, which was also corroborated in our study. Specifically, both IL-12-Exo and IL-12-Lipo treatments substantially increased CD8⁺ T cells, CD4⁺ T cells, NK cells and NKT cell populations while decreasing immune suppressive regulatory T (T_reg) cells and myeloid-derived suppressor cells (MDSCs) (Fig. 4d). Post-treatment TME changes included a rise in M1-like macrophages and a decline in M2-like macrophages (Supplementary Fig. 12a,b). Further analysis of the immune changes in the B16F10 lung metastatic TME revealed augmented infiltration of CD8⁺ T cells, CD4⁺ T cells and NKT cells following IL-12-Exo treatment (Supplementary Fig. 12c). Collectively, IL-12-Exo therapy drastically reshaped the TME, favouring cytotoxic T cells over exhausted counterparts. Mechanically, IL-12-activated lymphocytes induced a cascade of phenotypic alterations, typified by enhanced production of pro-inflammatory cytokines⁴⁷. Elevated mRNA expressions of IL-12, IFNγ, TNF and CD8 were detected post-IL-12-Exo treatment, while CCL2 was reduced (Fig. 4e). Moreover, IL-12 mRNA treatments also incited an upregulation of a broad spectrum of inflammatory cytokines and chemokines in the TME (Fig. 4f). High IFNγ levels persisted even 30 days after tumour inoculation. In addition, upregulated inflammatory effector proteins, such as CXCL9 and CXCL10, potentially aided in crafting a ‘hot’ TME⁴⁸.

Responses of CD8⁺ T cells and NKT cells in IL-12-Exo therapy

Following our identification of the crucial role of IFNγ in IL-12-Exo therapy (Fig. 4a,b,e,f), we investigated the primary cell types driving IFNγ expression in response to IL-12-Exo therapy 30 days after LL/2 tumour inoculation (depicted in Fig. 5a). IFNγ could be secreted by immune cells, especially T cells, NK cells or NKT cells in the TME^11,14. Effective mouse immunotherapy also relies on a strong collaboration between T cell-derived IFNγ and DC-produced IL-12 (ref. 14). Here, we discovered that IL-12-Exo treatment notably prompted enhanced IFNγ production by CD8⁺ T cells and NKT cells (Fig. 5b). CD8⁺ T cells acted as primary effectors due to their advantageous IFNγ production. To further scrutinize the effector function of CD8⁺ T cells and NKT cells, we used syngeneic LL/2 cell-induced lung orthotopic tumour models and assessed levels of granzyme B, an indicative marker of robust tumour cytotoxicity (Fig. 5c). Compared to the PBS group, both CD8⁺ T cells and NKT cells in the IL-12-Exo group expressed upregulated granzyme B levels, exceeding IL-12-Lipo (Fig. 5c). These results demonstrated their direct cytotoxic capability against tumour cells under IL-12-Exo modulation.

Fig. 5 |. CD8⁺ T cells are crucial for IL-12-Exo-mediated antitumour effects.

a, Schematic illustrating the TME modulation assessment after mice were treated with IL-12-Exo, IL-12-Lipo or PBS. b, Quantitative flow cytometry analysis of IFNγ expressions on CD8⁺ T cells, NKT cells, and NK cells in the TME, respectively, 30 days after tumour inoculation (n = 6 biologically independent mice for PBS and IL-12-Lipo groups, n = 8 biologically independent mice for IL-12-Exo group). c, Quantitative flow cytometry analysis of granzyme B expressions on CD8⁺ T cells, NKT cells and NK cells in the TME, respectively, 30 days after tumour inoculation (n = 5 biologically independent mice per group). d, Schematic showing the memory cell and survival assessment after different treatment regimens. e,f, Flow cytometry analysis of memory T cells (CD44⁺CD62L⁻ for effector memory cells, CD44⁺CD62L⁺ for central memory cells) (e) and memory NK cells (CD45⁺CD3⁺NK1.1⁺Ly6C⁺KLRG1⁻ for NKT memory cells, CD45⁺CD3⁻NK1.1⁺Ly6C⁺KLRG1⁻ for NK memory cells) (f) in the TME 42 days after tumour inoculation (n = 5 biologically independent mice per group). g,h, Schematic (g) and result (h) showing the tumour rechallenge assessment after mice were administered IL-12-Exo or PBS. These mice rejecting tumour growth were re-injected with LL/2 tumour cells on day 42, with healthy mice injected with LL/2 tumour cells serving as a control PBS group (n = 6 biologically independent mice for each group). i, Survival assessment of mice given PBS, IL-12-Exo, IL-12-Exo in combination with an anti-CD8 antibody, IL-12-Exo in combination with an anti-NK1.1 antibody or IL-12-Exo in combination with an anti-CD4 antibody. Mice were treated with the indicated depleting antibodies beginning 1 day before IL-12-Exo inhalation (n = 8 biologically independent mice for each group). P values were determined by two-way ANOVA post-Bonferroni’s multiple comparison test (b,c,e,h), two-tailed unpaired Student’s t-test (f) or one-side log-rank (Mantel–Cox) test (i) using GraphPad PRISM software. Exact P values are indicated. Results are presented as means ± s.d. Cartoon mice created with BioRender.com.

IL-12-Exo therapy also elicited immune memory development. A considerable duration post-IL-12-Exo treatment revealed a surge in efficacious CD8⁺ and CD4⁺ memory T cells (Fig. 5d,e), as well as NK and NKT memory cells (Fig. 5f) in the lung TME. These stimulated memory cells facilitated tumour resistance on rechallenge (Fig. 5g,h). In detail, post-IL-12-Exo therapy mice that initially repelled tumours were re-exposed to LL/2 tumour cells 42 days after primary inoculation, while a control group of healthy mice received the same LL/2 cells. Rechallenged mice displayed retarded tumour growth for 21 days, in stark contrast to the rapid tumour growth in the control group (Fig. 5h). To confirm causative roles of CD8⁺ T cells and NKT cells, we executed in vivo depletion in orthotopic lung tumour-bearing mice and monitored survival rates (depicted in Fig. 5d). Depletion of CD4⁺ T cells was also investigated, given their TME presence and immunological role. Cellular depletions were verified in mouse lungs (Supplementary Fig. 13). CD4⁺ T cell depletion mildly augmented therapeutic outcomes of IL-12-Exo therapy (Fig. 5i), potentially associated with the elimination of regulatory T cells (Fig. 4d)¹⁴. Conversely, therapeutic effects were dependent on CD8⁺ T cells. CD8 absence resulted in constrained tumour rejection and early mortality in mice, while NK1.1 depletion still preserved a partial degree of tumour remission (Fig. 5i).

Systemic T cell protection post-IL-12-Exo therapy

We delved deeper into investigating T cell protection after IL-12-Exo treatment, due to the key effector role of CD8⁺ T cells. Initially, we evaluated the presence of tumour-specific antigen-expressed CD8⁺ T cells in blood circulation (depicted in Fig. 6a). In mice with LL/2-OVA tumours, we detected OVA expression on peripheral CD8⁺ T cells. The PBS-treated group displayed negligible OVA⁺CD8⁺ T cells in circulation, while IL-12-Exo therapy notably led to their presence, outperforming IL-12-Lipo outcomes (Fig. 6b). The antigen specificity was further validated in 4T1 tumour-bearing mice, in which gp70 serves as the distinct antigen. Consistent with the therapeutic effect on LL/2-OVA tumours, gp70⁺CD8⁺ T cells were observed post-IL-12-Exo treatment (Fig. 6c).

Fig. 6 |. The cDC1 cell, functioning as an antigen-presenting cell, plays key roles in the therapy of IL-12-Exo.

a, Schematic outlining the establishment of the LL/2-OVA tumour-bearing orthotopic lung tumour model and the 4T1 tumour-bearing breast cancer lung metastatic tumour model, respectively, followed by the circulating T cell detection after treatment with IL-12-Exo, IL-12-Lipo or PBS. b, Flow cytometry results showing OVA-positive T cells (OVA⁺CD8⁺) in blood circulation for LL/2 tumour-bearing mice (n = 5 biologically independent mice per group). c, Flow cytometry results showing gp70-positive T cells (gp70⁺CD8⁺) in blood circulation for 4T1 tumour-bearing mice (n = 5 biologically independent mice per group). d, Schematic showing the cDC1 cell analysis in lymph nodes for each treatment group and in two tumour-bearing mouse models. e,f, Flow cytometry results showing the cDC1 cell percentage in lymph nodes of LL/2 tumour-bearing mice (n = 6 biologically independent mice per group) (e) or 4T1 tumour-bearing mice (f) (n = 4 biologically independent mice per group). g,h, Schematic (g) and results (h) of survival curves for each treatment group. LL/2 tumour cells were injected into C57BL/6 wild-type or Batf3^−/− mice, respectively (n = 10 biologically independent mice per group). P values were determined by one-way ANOVA post-Bonferroni’s multiple comparison test (b,c,e,f) or by one-side log-rank (Mantel–Cox) test (h) using GraphPad PRISM software. Exact P values are indicated. Results are presented as means ± s.d. Cartoon mice created with BioRender.com.

The induction of CD8⁺ T cell responses is primarily driven by cross-presenting type 1 DCs (cDC1 cells)¹⁴. Analysing DCs across the LL/2 orthotopic and the 4T1 lung metastatic tumour models (illustrated in Fig. 6d), an elevated number of migratory CD11c⁺CD11b⁻MHC-II⁺CD24⁺CD64⁻CD103⁺XCR1⁺ cDC1 cells in lymph nodes were evident 3 days after IL-12-Exo treatment (Fig. 6e,f and Supplementary Fig. 14). To further shed light on the role of cDC1 cells, the therapeutic effects of IL-12-Exo were assessed in Batf3^−/− mice, devoid of cDC1 cells. Consistent with the pivotal role of antigen presentation, the tumour rejection ability of IL-12-Exo treatment was significantly diminished in Batf3^−/− mice (Fig. 6g,h). In conclusion, IL-12-Exo therapy facilitated DC migration and subsequent cross-presenting in lymph nodes, thereby priming a systemic antitumour T cell response that was crucial for durable tumour suppression.

Long-term preclinical toxicity study of IL-12-Exo

The safety of drug candidates is paramount for clinical translation. To evaluate the long-term toxicity and safety of IL-12-Exo, we implemented a three-dose escalation preclinical trial (illustrated in Supplementary Fig. 15a). The mice demonstrated a similar percentage change in body weight, with IL-12-Exo not leading to rapid weight loss (Supplementary Fig. 15b). There was no discernible difference in normalized tissue weights between groups (Supplementary Fig. 15c). Hepatotoxicity remains a chief concern with IL-12 therapies⁴⁹. Nonetheless, our chemical analysis revealed physiological levels of AST and ALT in the IL-12-Exo group (Supplementary Fig. 15d,e). Additionally, haematoxylin and eosin staining indicated an absence of severe lesions in primary organs for both groups (Supplementary Fig. 15f). Collectively, our findings demonstrated the favourable safety profile IL-12-Exo, with all parameters fell within the normal range.

Conclusions

Anticancer cytokine therapies, including FDA-approved IL-12 and IFNγ (ref. 11), face challenges related to clinical translation and off-target toxicity⁴⁰. Here we report an inhalable therapeutic cancer vaccine, IL-12-Exo, based on exosomes loaded with IL-12 mRNA. In syngeneic mouse lung tumour models, IL-12-Exo enabled local and systemic antitumour immune responses without detectable toxicity over a 6-week study.

Systemic recombinant human IL-12 administration in early clinical trials caused severe toxicity via pro-inflammatory cytokine surge⁵⁰. Local delivery, as we use, curtails this ‘cytokine storm’ while preserving antitumour efficacy. Leveraging inhalation, a non-invasive method, we achieved direct therapeutic delivery to the lungs, superior to other delivery modes such as i.t. delivery^14,51. This study confirms successful local delivery of IL-12 mRNA to the lung TME, localizing the toxic cytokine to the tumour lesions.

Post-IL-12-Exo inhalation, enhanced IL-12 expression stimulated IFNγ production, primarily by CD8⁺ T cells, with NKT cells playing a lesser role. IFNγ-neutralization completely abrogated IL-12-Exo’s tumour suppression. In a LL/2 orthotopic lung tumour model, protective immune responses of IL-12-Exo were mainly relied on CD8⁺ T cells and partially on NKT cells, bypassing the need for CD4⁺ T cells. IL-12-Exo also encouraged memory immune cell proliferation, enabling tumour rejection on rechallenge. We observed activated populations of antigen-presenting cells in lymph nodes and tumour-specific antigen-expressing CD8⁺ T cells in circulation, potentially induced by IL-12 signalling or IFNγ-mediated reprogramming of DCs or monocytes^52,53. Another possible reason is indirect outcomes of tumour antigen and damage-associated molecular pattern release following tumour cell killing triggered by IFNγ and granzyme B producing T cells and NKT cells¹¹. In this study, we explained IL-12’s pleiotropic therapeutic mechanism.

Our study has constraints. Electroporation, while effective, is costly and limited in small scale for RNA encapsulation into exosomes. Inhalation may not suit patients with respiratory complications, and the transient nature of mRNA prompts the exploration of payloads such as self-amplifying RNA^14,54 for persistent cytokine expression.

In summary, IL-12-Exo stands out as a potent IL-12 mRNA delivery system to the lung TME, combining simplicity with efficacy against primary tumour and metastases. Compared to liposome controls, exosomes boost IL-12 expression with mitigated toxicity. As a non-invasive method, inhalation promises patient compliance superior to intratumoral injection. Exosomes, as biocompatible vesicles, provide versatile RNA delivery solutions. IL-12-Exo thus presents a prospective anticancer therapeutic vaccine, with exosome-based systems hinting at broad therapeutic potentials across clinical needs.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41565-023-01580-3.

Methods

Cell lines

HEK 293T cell line (CRL-3216) purchased from the American Type Culture Collection (ATCC) was cultured in minimum essential medium (Thermo Fisher Scientific) containing 10% FBS, 1% l-glutamine, 0.5% gentamicin and 0.18% 2-mercaptoethanol. Medium was refreshed every other day. HEK cells were allowed to reach 70–80% confluence before generating serum-free secretomes as previously described²⁵. The resulting secretomes were filtered through a 0.22-μm filter to remove cellular debris. LL/2 cell line (ATCC, catalogue no. CRL-1642-LUC2) and LL/2-OVA cell line (a kind gift from A. Beg, Moffitt Cancer Center, Tampa, FL, USA) were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (ATCC catalogue no. 30-2002) containing 10% FBS, 1% l-glutamine and blasticidin (final concentration of 2 μg ml⁻¹). B16F10 cell line (ATCC, catalogue no. CRL-6475-LUC2) was cultured in DMEM (ATCC catalogue no. 30-2002) containing 10% FBS, 1% l-glutamine and blasticidin (final concentration of 10 μg ml⁻¹). The 4T1 cell line (ACTT, catalogue no. CRL-2539-LUC2) was cultured in RPMI-1640 medium (Thermo Fisher Scientific) containing 10% FBS, 1% l-glutamine and blasticidin (final concentration of 8 μg ml⁻¹). Medium was changed every other day. These tumour cells were allowed to reach 70–80% confluence before cell passaging.

Animals

Female C57BL/6, BALB/c (Charles River Laboratory) and Batf3^−/− C57BL/6 (JAX stock no. 01375) mice, aged 8 weeks, were used in the study. All animal studies complied with the regulations and guidelines relating to the use of animals of the Institutional Animal Care and Use Committee of the North Carolina State University (protocol number 22-411).

Exosome isolation

HEK-Exo were isolated from secretomes using ultracentrifugation (130,000g) (Beckman Coulter Ultracentrifuges) and Ultra-15 centrifugal filter units (100 kDa cutoff) (MilliporeSigma)²¹. Pegylated Remote Loadable liposomes were purchased from Avanti Polar Lipids (Avanti Polar Lipids). Concentrations of HEK-Exo or liposomes were quantified by NanoSight (catalogue no. NS3000, Malvern Panalytical). Successful exosome isolation was confirmed by immunoblotting for exosome markers CD81, CD9 and CD63 at 20–24 °C

Vector constructs and mRNA loading

The murine single chain IL-12 sequence was previously described³⁴, which was prepared and purified by BOC Science company. This sequence encodes a fusion IL-12 protein-encoding single chain construct (IL12.p40.L.p35), consisting of a mouse IL12p40 subunit-encoding complementary DNA (cDNA) without a stop codon, a linker (15Aa encoding (Gly4Ser)3) sequence and a mouse IL12p35 subunit-encoding cDNA with no initial 22Aa codifying codons. IL-12 mRNA was loaded into HEK-Exo and liposomes, respectively, through electroporation, yielding IL-12-Exo or IL-12-Lipo (ref. 21). Two billion particles of HEK-Exo or liposomes were diluted in Gene Pulser Electroporation Buffer (Bio-Rad) at a 1:9 ratio of particles to buffer, followed by the addition of 10 μg of IL-12 mRNA. The mixture was transferred into a 4 mm Gene Pulser/MicroPulser Electroporation Cuvette (Bio-Rad) that was prechilled on ice. The cuvette was then inserted into the Gene Pulser XcellTM Total System (Bio-Rad) and electroporated under the following conditions: pulse type, square waveforms; voltage, 200 V; pulse length, 10 ms; and number of pulses, five and pulse intervals of 1 s. Ultrafiltration method was used to remove electroporation buffer and concentrated exosomes to the expected volume using Ultra-15 centrifugal filter units (100 kDa cutoff).

Characteristics of IL-12-Exo and IL-12-Lipo

Size distributions of IL-12-Exo or IL-12-Lipo were measured by NanoSight. Their zeta potentials were measured by dynamic light scattering analysis using a Malvern ZEN 3600 Zetasizer. Morphologies of samples were evaluated by TEM (Talo JEOL JEM-2000FX). Samples were incubated with DiD labelling solution (catalogue no. V22889, Thermo Fisher Scientific) or PKH26 (catalogue no. MINI26-1KT, Sigma) according to the manufacturer’s instructions.

Encapsulation efficiency

The encapsulation efficiency (EE%) of IL-12 mRNA was determined using the Quant-iT RiboGreen RNA Assay (Invitrogen)^35,36. RNA standards of high and low concentrations were prepared using Tris-EDTA buffer. IL-12-Exo was eluted in Tris-EDTA buffer, while additional samples were made in Tris-EDTA with Triton-X100 surfactant. Subsequently, 100 μl of either RNA standards or experimental samples were mixed with 100 μl of RiboGreen dye. This blend was allowed to incubate in darkness at room temperature for 5 min before the measurement of fluorescence intensity at an excitation of 485 nm and emission of 535 nm. The fluorescence deriving from IL-12-Exo dispersed in both Tris-EDTA and Tris-EDTA/Triton-X100 theoretically attributed to free (unencapsulated) and total mRNA, respectively. Free exosomes dispersed in Tris-EDTA/Triton-X100 indicated the existing RNA in free exosomes, which would also be subtracted from the total RNA in the calculation. The concentration (Conc.) of mRNA was converted from fluorescence intensity by using standard curves plotted from standard RNA solutions. Ultimately, the EE% was calculated using the formula: $\text{EE\%=}\left({\text{Conc}}_{·\text{total}\text{RNA}}{-\text{Conc}}_{·\text{free}\text{RNA}}{-\text{Conc}}_{·\text{exosome existing RNA}}\right){\text{/Conc}}_{·\text{total}\text{RNA}}$. The quantification of mRNA loaded in liposomes followed an analogous process, without the need to account for RNA in free liposomes.

Mucus penetration and cellular uptake studies

In vitro uptake of IL-12-Exo and IL-12-Lipo was investigated in LL/2 cells and IL4 pre-activated raw264.7 cells. In detail, these two types of cells (2 × 10⁵ per well, DiO labelling) were incubated with PKH26-labelled IL-12-Exo or IL-12-Lipo for 2, 6, 12 and 24 h, respectively, for the assessment of uptake efficacy. Cells were washed three times with PBS, followed by 4,6-diamidino-2-phenylindole (DAPI) staining and observation through a confocal microscope (Philip, FLUOVIEW). Mucus penetration was assessed using trans-well experiments. Consistent with our previous study³⁸, we used porcine stomach mucus due to the gene similarity and volume limitations in porcine tracheal tubes. Briefly, LL/2 cells were seeded in the bottom well of a 24-well plate (5 × 10⁴ per well) and left to adhere overnight. Next, 500 μl of mucus was added to the upper well for 1 h to form mucus layer. Following this, PKH26-labelled IL-12-Exo or IL-12-Lipo was added to the upper well and incubated for 6 h, respectively. On removal of the upper well, cells were rinsed three times with PBS, followed by DAPI staining and observation using a fluorescence microscope (Revolve, ECHO).

Tumour inoculation and inhalable drug delivery

Lung orthotopic and metastatic tumour models were established through i.v. injection of luciferase-expressing LL/2, B16F10 and 4T1 cells (1 million) into C57BL/6 mice. Tumour signals were quantified via luciferase by intraperitoneally (i.p.) injecting d-luciferin bioluminescent substrate (PerkinElmer) followed by quantification using the IVIS Spectrum imaging system. Nebulized treatment started 6 days post-tumour injection. IL-12-Exo, IL-12-Lipo or PBS control were given into mice using a nebulizer (Pari Trek S Portable Compressor Nebulizer Aerosol System, catalogue no. 047F45-LCS). Dose was standardized by the amount IL-12 mRNA (2.3 μg per mouse with 20 g of body weight, approximately 2 billion particles). For the distribution study, animals were euthanized 24 h post-treatments, followed by the analysis of IL-12-Exo-positive or IL-12-Lipo-positive cells. For IL-12 protein expression after animal euthanasia, blood and organs were collected for flow cytometry and histological examination. For rechallenge experiments, LL/2 tumour cells (1 million) were i.v. injected into the C57BL/6 mice that rejected primary tumours.

Depletion studies

Depletion of cellular subsets in vivo was conducted using in vivo antibodies against IFNγ (clone XMG1.2, catalogue no. BE00055, BioXCell, 400 μg, i.p. twice weekly), CD8α (clone 53-6.7, catalogue no. BE0004-1, BioXCell, 400 μg, i.p. twice weekly), NK1.1 (clone PK136, catalogue no. BE00036, BioXCell, 400 μg, i.p. twice weekly) or CD4 (clone GK1.5, catalogue no. BE0003-1, BioXCell, 400 μg, i.p. twice weekly) as previously described⁹. No dilution was needed for these nebulization antibodies used for cell depletion.

ELISA and Luminex analysis

LL/2 tumour-bearing lungs were collected and ground in tissue protein extraction reagent containing 1% proteinase and phosphatase inhibitors (Thermo Fisher Scientific, catalogue no. 78442). The lysates were incubated for 30 min at 4 °C with slow rotation and then centrifuged to remove debris. The supernatants were collected and transferred to a clean tube for enzyme-linked immunosorbent assay (ELISA) or Luminex analysis. Protein levels of IL-12 or IFNγ in cells, lung tissue supernatants or serum were measured by ELISA kits from Thermo Fisher Scientific (catalogue no. BMS616 for IL-12 and catalogue no. BMS606-2 for IFNγ), following the manufacturer’s instructions. Portions of lung tissue supernatants were sent to Eve Technology for Luminex analysis of cytokine, chemokines and growth factors. Furthermore, cell supernatants of LL/2 cells that were transfected by IL-12-Exo or IL-12-Lipo for 24 h were also collected for protein detection using an ELISA method.

Antibodies, staining and flow cytometry

Staining antibodies were summarized in Supplementary Table 1, with dilution details. Cell viability was distinguished using the live or dead dye Zombie Violet (BioLegend, catalogue no. 423114) or Lime (506) (Thermo Fisher Scientific, catalogue no. L34990). Following the euthanization of tumour-bearing mice, their lungs were sliced and digested by DNase I (100 μg ml⁻¹, Invitrogen) and collagenase type IV (200 U ml⁻¹, Invitrogen) for 60 min at 37 °C with slow rotation. These digested lung tissues were then filtered by 70-μm nylon strainers and ground into single-cell suspensions. Lymph nodes were also ground into single-cell suspensions. After removing red blood cells with ACK Lysis Buffer (Gibco), the suspensions were stained. Intracellular staining of FoxP3, IFNγ, TNF and granzyme B was performed using the FoxP3 Transcription Factor Buffer Set (eBioscience, catalogue no. 00-5523-00). Stained cell samples were analysed using a fluorescence-activated cell-sorting analyser (BD Biosciences, LSR-II), with data analysed using FCS Express v.7 (De Novo software) and FlowJo software (Flowjo, LLC). We used the following markers to identify cell populations: CD8⁺ T cells (CD45⁺CD3⁺CD8⁺), CD4⁺ T cells (CD45⁺CD3⁺CD4⁺), T_reg cells (CD45⁺CD4⁺FoxP3⁺), NK cells (CD45⁺CD3⁻NK1.1⁺), NKT cells (CD45⁺CD3⁺NK1.1⁺), MDSCs (CD45⁺CD11b⁺Gr-1⁺), effector memory T cells (CD44⁺CD62L⁻),central memory T cells (CD44⁺CD62L⁺), NK memory cells (CD45⁺CD3⁻NK1.1⁺Ly6C⁺KLRG1⁻) and NKT memory cells (CD45⁺CD3⁺NK1.1⁺Ly6C⁺KLRG1⁻). The flow cytometry results are presented as a percentage oftotal cells.

Immunofluorescence, histology, imaging and processing

Immunofluorescence staining was performed on tissue slides fixed in 4% paraformaldehyde (Electron Microscopy Sciences, catalogue no. 15710). These slides were then permeabilized and blocked using Dako Protein blocking solution (DAKO, catalogue no. X0909) containing 0.1% saponin (Sigma-Aldrich, catalogue no. 47036). Immunofluorescence images were captured through a confocal microscope (Philip, FLUOVIEW) and analysed using NIH ImageJ software. Haematoxylin and eosin staining (Sigma-Aldrich, catalogue nos. HSS16 and 318906) was performed on paraffin-embedded formalin fixed tissues from mouse major organs (spleen, heart, liver, lung, kidney).

Blood biochemistry

Blood biochemistry tests were conducted by North Carolina State University Veterinary Hospital’s clinical pathology laboratory. Analysis was performed according to the International Federation of Clinical Chemistry and Laboratory Medicine on the Cobas C501 instrument by Roche. AST and ALT were determined by International Federation of Clinical Chemistry and Laboratory Medicine methods without pyridoxal-5-phosphate.

Statistical analysis

All quantitative experiments were presented as means ± standard deviation (s.d.). A two-tailed unpaired Student’s t-test was used to analyse differences between two groups. One- or two-way analyses of variance (ANOVA) were used for comparisons among three or more groups, supplemented with a post hoc Bonferroni correction. Differences were analysed using GraphPad PRISM software and exact P values were documented in the figures or figure captions. No exclusion criteria were incorporated in the design of experiments for this study.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary Material

The online version contains supplementary material available at https://doi.org/10.1038/s41565-023-01580-3.

Data availability

All data supporting the conclusions of this study are presented in the article and the Supplementary Information. The Cancer Genome Atlas database (https://www.cancer.gov/ccg/research/genome-sequencing/tcga) and Kaplan–Meier plot database (http://kmplot.com) are used for the results of Supplementary Fig. 1.