Npm1 deficiency suppresses tumor growth in vivo

To examine whether there is a link between NPM1 expression and tumor progression, we first analyzed publicly available data in the TCGA database [17]. Compared to non-tumor samples, NPM1 mRNA was increased in many human tumor types, including COAD (colon adenocarcinoma), LIHC (liver hepatocellular carcinoma), HNSC (head and neck squamous cell carcinoma), LUAD (lung adenocarcinoma) and KIRP (kidney renal papillary cell carcinoma) (Fig. 1A). In addition, overall survival analysis demonstrated that high levels of NPM1 predict low survival rates in various human tumors (Fig. 1B). These results indicate that NPM1 promotes human tumor progression.

To characterize the effect of NPM1 on tumor progression and anti-tumor immune responses, we knockout the Npm1 gene in three mouse tumor cell lines (colon carcinoma MC38, melanoma B16F10 and lung carcinoma LLC) using the CRISPR-Cas9 system (Fig. S1A). The knockout efficacy was confirmed by immunoblotting (Fig. S1B). Wild-type or Npm1-deficient tumor cells were then subcutaneously injected into syngeneic wild-type mice, and tumor growth was monitored. We found that loss of NPM1 in MC38 (Fig. 1C and D), B16F10 (Fig. 1E and F) and LLC (Fig. 1G and H) tumor cells significantly reduced tumor growth and prolonged the survival of tumor-bearing mice compared to controls (Fig. 1I and J). Thus, Npm1 deficiency significantly suppresses tumor growth in multiple mouse tumor models, providing additional evidence that NPM1 promotes tumor progression in vivo.

Fig. 1

NPM1 is linked to increased human and mouse tumor progression. (A) Box plots showing the mRNA levels of NPM1 in different human tumors and the corresponding normal tissues using the TCGA database. (B) Kaplan-Meier survival curve for diverse human tumors based on NPM1 mRNA level using the TCGA database. (C and D) Wild-type and Npm1-deficient MC38 mouse colon cancer cells were subcutaneously injected into syngeneic mice C57BL/6. Tumor growth curve (C) and weight at the endpoint (D). n = 6, 5 mice, respectively in C. n = 9, 8 mice, respectively in D. (E and F) Wild-type and Npm1-deficient B16F10 mouse melanoma cells were subcutaneously injected into C57BL/6 mice. Tumor growth curve (E) and weight at the endpoint (F). n = 9, 10 mice, respectively in E. n = 10 mice per group in F. (G and H) Wild-type and Npm1-deficient LLC mouse lung carcinoma cells were subcutaneously injected into C57BL/6 mice. Tumor growth curve (G) and weight at the endpoint (H). n = 7 mice per group in G. n = 8 mice per group in H. (I) Overall survival of MC38 tumor-bearing mice. n = 7 mice per group. (J) Overall survival of B16F10 tumor-bearing mice. n = 15, 11 mice, respectively. Data are representative of three independent experiments (C-H); data are combined results from two separate experiments (I and J). Error bars in A, D, F and H indicate mean ± SD; error bars in C, E and G indicate mean ± SEM. P values in A, D, F and H were calculated by a two-tailed, unpaired Student’s t test; P values in C, E and G were calculated by two-way ANOVA; P values in B, I and J were analyzed by log-rank test. *P < 0.05; **P < 0.01; ***P < 0.001. n.s., not significant; N/A, not applicable; HR, hazard ratio. BLCA, bladder urothelial carcinoma; BRCA, breast invasive carcinoma; CESC, cervical squamous cell carcinoma and endocervical adenocarcinoma; CHOL, cholangiocarcinoma; COAD, colon adenocarcinoma; ESCA, esophageal carcinoma; GBM, glioblastoma multiforme; HNSC, head and neck squamous cell carcinoma; KICH, kidney chromophobe; KIRC, kidney renal clear cell carcinoma; KIRP, kidney renal papillary cell carcinoma; LIHC, liver hepatocellular carcinoma; LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma; PAAD, pancreatic adenocarcinoma; PRAD, prostate adenocarcinoma; PCPG, pheochromocytoma and paraganglioma; READ, rectum adenocarcinoma; SARC, sarcoma; SKCM, skin cutaneous melanoma; THCA, thyroid carcinoma; THYM, thymoma; STAD, stomach adenocarcinoma; UCEC, uterine corpus endometrial carcinoma. See also Figure S1

Npm1 deficiency reprograms the immunogenic tumor microenvironment

To explore the potential function of NPM1 in anti-tumor immune responses, we deciphered the immune atlas of inoculated wild-type and Npm1-deficient MC38 tumors using 36-parameter mass cytometry by time of flight (CyTOF) (Fig. 2A). 22 distinct immune cell subsets were identified and the proportion of each immune cell type was calculated using the dimensionality reduction algorithm t-SNE and the clustering algorithm PhenoGraph (Fig. 2B). We found a significant increase in CD8+ T cells (cluster 4) and CD4+ T cells (cluster 7) in Npm1-deficient tumors; meanwhile, the macrophage (cluster 1 and 19) and myeloid-derived suppressor cell (MDSC) (cluster 17) populations were reduced (Fig. 2C).

Since the recruitment and activation of cytotoxic T cells in the TME is essential for eliciting an effective anti-tumor response [4], we next investigated the expression profiles of each T cell cluster using CyTOF analysis (Fig. 2D). Specifically, CD8+ T cells in cluster 2, a group that was increased in the Npm1-deficient TME, expressed the highest levels of the cytotoxic effector molecules perforin, granzyme B and IFN-γ, as well as the co-stimulatory molecules CD28 and CD86. This profile indicates the enhanced infiltration of effector CD8+ T cells in the Npm1-deficient TME. CD8+ T cells in cluster 4 showed high expression of Ly6C, CD28 and CD86 and were also significantly increased in the Npm1-deficient TME. Notably, Ly6C+ CD8+ T cells have been reported to exhibit a more activated state and correlate with enhanced cytotoxic activity [18]. CD4+ T cells in cluster 6 were defined as the regulatory T cells (Tregs) since they co-express FoxP3 and CD25. There was no significant difference in this cluster between the wild-type and Npm1-deficient TME. CD4+ T cells in cluster 7 were significantly increased in the Npm1-deficient TME. However, we could not define the specific subtype represented by cluster 7 due to antibody panel limitations.

In addition to T cells, we defined cluster 1 and cluster 19 as the inhibitory macrophages since they express high levels of F4/80 and Arginase 1. Meanwhile, cluster 10 and cluster 17 were characterized by high expression of Ly6C and thus defined as MDSCs. All of these myeloid cells, which play roles in immunosuppression in the TME, were significantly decreased in Npm1-deficient tumors (Fig. 2C). Furthermore, Npm1 deficiency resulted in reduced expression of genes associated with immune suppression in myeloid cells, such as CD11b, F4/80, Arginase 1 and Ly6C (Fig. 2E).

In summary, CyTOF analysis of MC38 tumors indicates that Npm1 deficiency inflames the TME to be immunogenic by promoting the infiltration, activation of CD8+ T cells, CD4+ T cells and inhibiting immunosuppressive phenotypes, thereby boosting anti-tumor immune responses.

Fig. 2

Npm1 deficiency in tumor cells leads to an inflammatory and immunogenic profile in the TME. (A) Schematic diagram of CyTOF experiments. MC38 tumor cells were subcutaneously injected into C57BL/6 mice. Tumor tissues were collected on day 25 and digested into a single-cell suspension. CD45+ cells were isolated using magnetic beads and labeled for CyTOF analysis. n = 8 mice per group. (B) t-SNE plot of 180,000 CD45+ singlets collected and analyzed as in (A). (C) The frequencies of each PhenoGraph cluster to CD45+ immune cells as in (A). (D) Heatmap showing the normalized expression profiles of the T cell PhenoGraph clusters as in (A). (E) Normalized expression of indicated myeloid markers on the t-SNE plot as in (A). Data are representative of two independent experiments. Error bars in C indicate mean ± SD. P values in C were calculated by a two-tailed, unpaired Student’s t test. *P < 0.05; **P < 0.01

Npm1 deficiency promotes CD8+ T cell infiltration into tumors

Flow cytometry analysis confirmed that CD45+ immune cell infiltration and the proportion of CD4+ T cells and CD8+ T cells were increased in Npm1-deficient MC38 tumors (Fig. 3A) and LLC tumors (Fig. 3B) compared to controls. Meanwhile, the proportion of natural killer cells (NKs) and DCs showed no significant differences in MC38 tumors (Fig. S2A, S2B) and LLC tumors (Fig. S3A, S3B); only the proportion of MDSCs was decreased in Npm1-deficient LLC tumors (Fig. S3A) in flow cytometry analysis. Immunohistochemistry (IHC) staining confirmed a consistent and significant increase in CD8+ T cell infiltration into Npm1-deficient MC38 tumors (Fig. 3C).

To determine the contribution of enhanced infiltration of CD8+ T cells to tumor inhibition in Npm1-deficient tumors, we utilized an antibody-based CD8+ T cell depleting approach and found that CD8+ T cell depletion strikingly abolished the growth delay of Npm1-deficient tumors (Fig. 3D).

Furthermore, to account for deviations caused by tumor size, we subcutaneously injected wild-type and Npm1-deficient tumor cells on the same day but resected them five days apart to equalize the tumor size at the experimental endpoint (Fig. 3E). Subsequent CyTOF analysis recapitulated the differences in immune cell infiltration into the TME, including a significant enrichment of CD8+ T cells, CD4+ T cells and DCs but decreased numbers of several tumor-associated macrophage (TAM) subsets in Npm1-deficient tumors compared to controls (Fig. 3F and G).

These results further demonstrate that Npm1 deficiency reprograms the immunogenic TME by promoting the infiltration of CD8+ T cells, which then control the growth of Npm1-deficient tumors.

Fig. 3

Npm1 deficiency in tumor cells enhances the infiltration of CD8+ T cells in the TME. (A) Flow cytometry analysis of the percentage of CD45+ immune cells, CD4+ T cells and CD8+ T cells in MC38 tumors. Representative FACS plots (left) with quantification (right) were shown. n = 9, 5 mice, respectively. (B) Flow cytometry analysis of the percentage of CD45+ immune cells, CD4+ T cells and CD8+ T cells in LLC tumors. Representative FACS plots (left) with quantification (right) were shown. n = 7, 9 mice, respectively. (C) Immunohistochemistry (left) with quantification (right) of CD8+ T cells in MC38 tumors. n = 5 mice per group. Scale bar, 10 μm. (D) Schematic diagram of lymphocyte depletion experiment (left) and tumor growth curve of tumor-bearing mice (right). MC38 tumor cells were subcutaneously injected into C57BL/6 mice on day 0 when CD8+ T cells were depleted using 100 μg of intraperitoneally injected anti-CD8 antibody on days − 4, -1 and 2. n = 8 mice per group. (E) Schematic diagram showing the mouse tumor model in which wild-type or Npm1-deficient MC38 tumor cells were subcutaneously injected into C57BL/6 mice and resected five days apart to equalize tumor size. n = 8 mice per group. (F) t-SNE plot of 180,000 CD45+ singlets collected and analyzed as in (E). (G) The frequencies of each PhenoGraph cluster to CD45+ immune cells as in (E). Data are representative of three independent experiments (A and B). Error bars indicate mean ± SD. P values in A, B, C and G were calculated using a two-tailed, unpaired Student’s t test; P values in D were calculated by two-way ANOVA. *P < 0.05; **P < 0.01; ***P < 0.001. a.u., arbitrary unit; i.p., intraperitoneal; s. c., subcutaneous. See also Figure S2 and Figure S3

Npm1 deficiency enhances tumor cell MHC expression and immunogenicity

Next, we wondered how Npm1 deficiency reprograms the immunogenic TME. We determined the effect of tumor NPM1 on in-vitro CTL-dependent killing of OVA-presenting tumor cells by OT-I transgenic T cells which specifically recognize SIINFEKL OVA peptides. We found that Npm1-deficient tumor cells were significantly more sensitive than wild-type tumors to specific T-cell killing (Fig. 4A and B).

We went further to conduct a transcriptomic analysis to comprehensively compare gene expression in wild-type and Npm1-deficient tumor cells freshly isolated from MC38 tumor tissues. Gene ontology (GO) analysis of the up-regulated genes in the Npm1-deficient tumor cells showed that three of the top ten biological processes were associated with antigen processing and presentation (Fig. 4C). Gene-set enrichment analysis (GSEA) showed that two gene sets, the antigen processing and presentation of exogenous antigen (GO: 0019884) and the antigen processing and presentation of peptide or polysaccharide antigen via MHC class II pathway (GO: 0002504), were significantly enriched in Npm1-deficient tumor cells (Fig. 4D).

The RNA-seq results were confirmed by Q-PCR analysis of MHC-I (H2-D1, H2-K1 and Tap1) and MHC-II (Cd74, H2-Eb1 and H2-Dmb1) transcripts. These mRNAs were significantly elevated in Npm1-deficient tumor cells freshly isolated from inoculated tumors (Fig. 4E). Flow cytometry analysis validated that both MHC-I and MHC-II expression were significantly increased in Npm1-deficient tumors relative to control tumors (Fig. 4F). To further verify the relationship between NPM1 and MHC expression, we pretreated MC38 tumor cells with mIFN-γ in vitro and found that Npm1 deficiency resulted in a marked increase in MHC-I and MHC-II expression at the mRNA and protein levels (Fig. 4G and H). Consistent results were obtained by analyses of B16F10 tumor cells (Fig. S4A, S4B) and LLC tumor cells (Fig. S4C, S4D), both were treated with mIFN-γ in vitro. Consistently, siRNA-mediated silencing of NPM1 significantly enhanced hIFN-γ-induced MHC-I and MHC-II expression in human lung cancer cells A549 (Fig. 4I and J) and human colon cancer cells HCT116 (Fig. S4E-S4G). Furthermore, analysis of TCGA data [19] revealed that MHC class I molecules (HLA-A, HLA-B, HLA-C and TAPBP) (Fig. S5A) and MHC class II molecules (HLA-DPB1, HLA-DQA1, HLA-DQA2 and HLA-DRB1) (Fig. S5B) were negatively associated with NPM1 mRNA levels in human COAD and SKCM (skin cutaneous melanoma).

Taken together, these results demonstrate that Npm1 deficiency leads to increased MHC-I and MHC-II expression, which may underline the increased antigen processing and presentation of tumor cells, and also promote tumor immunogenicity and support cytotoxic T cell-mediated elimination of tumor cells.

Fig. 4

Npm1 deficiency enhances specific T-cell killing and MHC expression in tumor cells. (A) LDH cytotoxicity assay of MC38 tumor cells incubated with activated OVA-specific T cells at decreasing E: T ratios (1:2, 1:1, 2:1 and 4:1) for 24 h. n = 3. (B) Flow cytometry analysis of the fraction of PI+ MC38 tumor cells incubated with activated OVA-specific T cells at decreasing E: T ratios (1:2, 1:1 and 2:1) for 24 h. n = 3. (C and D) Transcriptomic analysis of CD45− tumor cells isolated from wild-type and Npm1-deficient MC38 tumors. The top 10 enriched biological processes in GO analysis (C) and GSEA plots of antigen processing and presentation gene signatures (D) are shown. (E and F) Q-PCR analysis of the mRNA levels of MHC genes (E) and flow cytometry analysis of H2-Kb/Db and I-Ab levels (F) of wild-type and Npm1-deficient MC38 tumor cells isolated from subcutaneous tumors. n = 4, 5 mice, respectively in E. n = 9, 8 mice, respectively in F. (G and H) Q-PCR analysis of the mRNA levels of MHC genes (G) and flow cytometry analysis of H2-Kb/Db and I-Ab levels (H) of wild-type and Npm1-deficient MC38 tumor cells treated with mIFN-γ (10 ng/mL) for 48 h in vitro. n = 3. (I and J) Q-PCR analysis of the mRNA levels of MHC genes (I) and flow cytometry analysis of HLA-A/B/C and HLA-DR/DP/DQ levels (J) in A549 cells transfected with siNC or siNPM1 for 48 h and then treated with hIFN-γ (20 ng/mL) for 24 h. n = 3. Data shown are representative of three independent experiments (A, B and E-J). Error bars indicate mean ± SD. P values were calculated using a two-tailed, unpaired Student’s t test. *P < 0.05; **P < 0.01; ***P < 0.001. E: T ratio, effector: target ratio; NES, normalized enrichment score; MFI, mean fluorescence intensity; ut, untreated; NC, negative control. See also Figure S4 and S5

Npm1 deficiency increases the transcription of Nlrc5 and Ciita in tumor cells

NLRC5 and CIITA have been recognized as master transcription factors regulating the MHC-I and MHC-II genes, respectively [20, 21]. Consistent with this, we found a negative correlation between NPM1 mRNA levels and NLRC5 and CIITA in multiple human tumors including SKCM, TGCT (testicular germ cell tumors), GBM (glioblastoma multiforme) and LGG (brain lower grade glioma) in TCGA data (Fig. 5A). Consistently, we observed the in vivo expression of Nlrc5 and Ciita were significantly enhanced in Npm1-deficient tumors (Fig. 5B and C). Meanwhile, the mRNA levels of Nlrc5 and Ciita were also significantly increased in mIFN-γ-treated Npm1-deficient MC38 cells (Fig. 5D) and in hIFN-γ-treated NPM1-silenced A549 cells (Fig. 5E) in vitro. These results indicate that Npm1 deficiency upregulates MHC expression, possibly by affecting the abovementioned upstream transcription factors.

Since Npm1 deficiency resulted in increased mRNA levels of Nlrc5 and Ciita, we tested whether NPM1 acts at the transcriptional or post-transcriptional level. We treated tumor cells with IFN-γ and found that Npm1 deficiency significantly increased the preRNA levels of Nlrc5 and Ciita(Fig. 5F). However, by measuring the half-life of the Nlrc5 and Ciita mRNAs in tumor cells treated with actinomycin D, we found that the stability of these transcripts was unaffected by the presence or absence of NPM1 (Fig. 5G). Therefore, NPM1 regulates Nlrc5 and Ciita expression at the transcriptional level.

In addition, we explored the effect of the NPM1 inhibitor NSC348884, which disrupted the oligomerization of NPM1 but did not change NPM1 protein levels, on MHC expression of tumor cells [22]. Our data revealed that NSC348884 did not affect the MHC expression in MC38, B16F10, A549 and HCT116 tumor cells (Fig. S6). We therefore hypothesized that NPM1 oligomerization is not related to its regulation of MHC expression in tumor cells.

Fig. 5

Npm1 deficiency promotes the transcription of Nlrc5 and Ciita. (A) Correlation of NPM1 with CIITA (upper) or NLRC5 (bottom) in different cancer types. Data were obtained from TCGA and analyzed by the Spearman correlation test. (B) Q-PCR analysis of the mRNA levels of Nlrc5 and Ciita in wild-type and Npm1-deficient MC38 tumors cells isolated from subcutaneous tumors. n = 4, 5 mice, respectively. (C) Immunoblot analysis of NLRC5 expression in wild-type and Npm1-deficient LLC tumor tissues. Uncropped blots are shown in the source data. (D) Q-PCR analysis of the mRNA levels of Nlrc5 and Ciita in wild-type and Npm1-deficient MC38 tumor cells treated with mIFN-γ (10 ng/mL) for 48 h in vitro. n = 3. (E) Q-PCR analysis of the mRNA levels of NLRC5 and CIITA in A549 cells transfected with siNC or siNPM1 for 48 h and then treated with hIFN-γ (20 ng/mL) for 24 h. n = 3. (F) Q-PCR analysis of the preRNA levels of Nlrc5 and Ciita in mIFN-γ-treated (10 ng/mL) wild-type and Npm1-deficient MC38 tumor cells in vitro. n = 3. (G) Decay curves of Nlrc5, Ciita and Actb mRNAs in mIFN-γ-stimulated wild-type and Npm1-deficient MC38 tumor cells treated with actinomycin D (10 μg/mL) for 0, 3, 6 and 9 h. n = 3. Data are representative of three independent experiments (B, D-G). Error bars indicate mean ± SD. P values in A were calculated using the Spearman correlation test; P values in B, D-G were calculated using a two-tailed, unpaired Student’s t test. *P < 0.05; **P < 0.01; ***P < 0.001. ut, untreated; NC, negative control; Act. D, actinomycin D. See also Figure S6

NPM1 associates with IRF1 and inhibits IRF1-mediated Nlrc5 and Ciita transcription

It is well established that IRF1 and STAT1 are key factors required for the transcriptional induction of MHC antigen-presenting genes upon IFN-γ stimulation [2, 3]. Previous researches show that NPM1 interacts with IRF1 and STAT1 [23] and NPM1 inhibits the DNA-binding and transcriptional activity of IRF1 [24]. Indeed, immunoprecipitation assays confirmed that NPM1 could associate with IRF1 and STAT1 as reported (Fig. 6A). We examined the subcellular location of NPM1 protein and found that NPM1 was mainly localized in the nucleolus (Fig. 6B). Subcellular fractionation analysis revealed that NPM1 was widely distributed in the nucleus, including nucleoplasm and chromatin (Fig. 6C). Notably, NPM1 was abundant in chromatin fraction, implying NPM1 might regulate gene transcription. Therefore, we constructed luciferase reporter plasmids bearing the core promoter region of either the Nlrc5 or the Ciita gene. Using the dual-luciferase reporter assay, we discovered that the activity of both the Nlrc5 and Ciita promoter was increased in Npm1-deficient cells compared to controls (Fig. 6D). By performing chromatin immunoprecipitation (ChIP) assays, we noticed that Npm1 deficiency greatly enhanced the binding of IRF1 to the promoter of Nlrc5 and Ciita in IFN-γ stimulated tumor cells (Fig. 6E). However, we observed no significant effect of NPM1 on STAT1 binding (Fig. 6F). Our data indicate that NPM1 interacts with IRF1 to block its binding to the Nlrc5 and Ciita promoters and thus inhibit their transcription, leading to MHC downregulation and, ultimately, immune evasion (Fig. 6G).

Fig. 6

NPM1 inhibits MHC expression by interacting with IRF1 and blocking IRF1 binding to Nlrc5 and Ciita promoters. (A) Immunoprecipitation analysis of the interaction between NPM1 and IRF1 and STAT1 in MC38 tumor cells treated with mIFN-γ (10 ng/mL) for 24 h. Uncropped blots are shown in the source data. (B) Immunofluorescence analysis of the subcellular location of NPM1 in MC38 tumor cells. Scale bar, 10 μm. (C) Subcellular fractionation of MC38 tumor cells with or without mIFN-γ treatment (10 ng/mL) for 24 h. Each fraction was loaded in equal proportions. Uncropped blots are shown in the source data. (D) Dual-luciferase reporter assay of Nlrc5 (right) and Ciita (left) promoter activity driven by IRF1 and STAT1 transfection in wild-type and Npm1-deficient MC38 tumor cells. Data are relative to Renilla luciferase activity. n = 3. (E and F) ChIP analysis of the DNA binding of IRF1 (E) or STAT1 (F) to the Nlrc5, Ciita and Actb promoters in wild-type and Npm1-deficient MC38 tumor cells with mIFN-γ stimulation (10 ng/mL) for 24 h. Data are relative to control immunoprecipitation with immunoglobulin G (IgG) antibody. n = 3. (G) The proposed model for the previously undescribed role of NPM1-IRF1-NLRC5/CIITA-MHC axis in suppressing tumor antigen presentation and reprogramming the immunosuppressive TME. Data are representative of three (A-D) or two (E and F) independent experiments. Error bars indicate mean ± SD. P values in D-F were calculated using a two-tailed, unpaired Student’s t test. **P < 0.01; ***P < 0.001. NC, negative control; WCL, whole cell lysates; TF, transcription factor; ut, untreated; n.s., not significant