hMENA^11a depletion activates type I IFN pathways and increases PD-L1 expression in NSCLC cell lines

hMENA isoform expression pattern has been proposed to be a prognostic factor, with low hMENA^11a expression, along with high overall hMENA, identifying early NSCLC patients with poor prognosis.21 23

To gain insights into molecular mechanisms accounting for the poor prognostic value of low levels of hMENA^11a, we sought to analyze molecular changes resulting from hMENA^11a depletion in NSCLC cell lines, by a whole transcriptome analysis by RNA-Seq approach of two lung adenocarcinoma (ADC) cell lines (H1650 and H2030) with an epithelial phenotype and overexpressing hMENA^11a isoform.

To this aim, we specifically silenced hMENA^11a in these cell lines (hereafter si11a cells) by siRNA approach and compared their transcriptomic profile with the control cells (transfected with non-targeting siRNA, hereafter sictr cells), and with the profile of cells silenced in total hMENA isoforms (obtained by using siRNA targeting all hMENA isoforms, hereafter sihMENA(t) cells) (figure 1A,B). Expression levels of different isoforms in silenced cells are shown in online supplemental figure S1A.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

hMENA^11a silencing upregulates transcripts of genes related to IFN-signaling pathways in NSCLC cell lines. (A, B) (left) Volcano plot showing statistical significance (qvalue, −log10 scaled) vs fold change (log2 scaled) for genes differentially expressed after deletion of hMENA^11a in H1650 and H2030: 196 and 930 downregulated (blue) and 158 and 822 upregulated (red) genes in si11a cells vs sictr cells, respectively (q value <0.05). (A, B) (right) Enrichment analyses (GO Biological Process) of upregulated genes in si11a cells, H1650 (up) and H2030 (down). (C) Upregulated gene network in H1650 si11a cells revealed by STRING analysis. (D) qPCR results of mRNA levels of selected genes in sictr, si11a and sihMENA(t) H1650, H2030 and HCC2935 cells. (E) Levels of IFNβ in supernatants evaluated by ELISA (left) and of IFNB1 mRNA by qRT-PCR (right) expressed as fold-change relative to control in H1650 sictr, si11a, sihMENA(t) cells. (D, E) P values were calculated by two-tailed Student’s t-test. *p≤0.05; **p≤0.01.

Surprisingly, Gene Ontology term enrichment analysis of RNA-Seq data indicated that a large fraction of genes that changed their expression profile in response to hMENA^11a silencing was mainly involved in type I IFN-related pathways and related cellular response, suggesting that, among the upregulated biological processes in si11a cells, the response to Type I IFN was dominant (figure 1A,B). In particular, in both cell lines, si11a cells showed the upregulation of key molecules involved in the antiviral interferon response, such as the IFN-stimulated genes (ISGs) OAS1, OASL, ISG15, ANKRD17, RNASEL, IRF9. Notably, STAT1 the crucial mediator of interferon transcriptional activity was among the upregulated genes in si11a cells compared with sictr cells and sihMENA(t) cells. Of note, STRING network analysis for predicting protein–protein interactions indicated ISG15 as a central hub node gene in si11a cells (figure 1C).

Validation of RNA-Seq results by RT-PCR in three epithelial cancer cell lines confirmed that specific hMENA^11a silencing, but not hMENA(t) silencing, increases ISG mRNA levels (figure 1D and online supplemental figure S1B).

Since RNA-Seq analysis indicated that the specific hMENA^11a depletion activates IFN-related pathways resembling a viral defense response, which typically culminates with IFNβ production, we analyzed IFNβ levels in cell CM. We found that, while H1650 and HCC2935 sictr cells secrete barely detectable levels of IFNβ, si11a cells produce relevant amount of this cytokine, indicating that hMENA^11a silencing per se increases IFNβ secretion (figure 1E and online supplemental figure S1C, left). Noteworthy, cells silenced for total hMENA isoforms do not produce IFNβ. As expected, we found IFNB1 upregulation also at transcriptional level in si11a cells (figure 1E and online supplemental figure S1C, right). Of note, H1650 sictr cells, as well as sihMENA(t) cells, showed barely detectable levels of IFNα mRNA, and undetectable IFNγ. Neither IFNα nor IFNγ levels were increased by hMENA^11a silencing (online supplemental figure S1D).

To mimic a viral infection, able to trigger type I interferon pathway, we treated H1650 cells with polyinosinic acid-polycytidylic acid Poly(I:C), the synthetic dsRNA analog. As a result, hMENA^11a expression was reduced in parallel with an increase of STAT1 expression and activation, along with the production of IFNβ (online supplemental figure S1E).

Treatment of a panel of NSCLC cell lines with IFNβ upregulated PD-L1 expression (figure 2A), indicating that IFNβ is a critical regulator of PD-L1 expression levels in NSCLC cells, as previously reported for IFNγ in melanoma cells.28

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

hMENA^11a silencing increases PD-L1 at mRNA and protein levels via JAK/STAT1/IRF1 axis in NSCLC cell lines. (A) PD-L1 expression levels (MFI) assessed by flow cytometry analysis in a panel of NSCLC cells, untreated or treated with IFNβ (50 ng/mL) for 24 hours. (B) qRT-PCR results of CD274 analyzed in sictr, si11a and sihMENA(t) H1650 cells. (C) PD-L1 expression levels assessed by flow cytometry analysis in a panel of sictr, si11a, and sihMENA(t) NSCLC cells. (D) CD274 mRNA expression levels evaluated by qRT-PCR in A549 transfected with empty vector (pMSCV) or hMENA^11a (pMSCV11a), untreated or treated with IFNβ (50 ng/mL) for the indicated hours. (E) qRT-PCR results of CD274 analyzed in sictr, si11a and sihMENA(t) H1650 cells untreated or treated with JAK inhibitor (5 µM) for 72 hours. (F) Western blotting of cells treated as in (B). (G) Representative images of immunofluorescence of sictr, si11a and sihMENA(t) H1650 stained with anti-IRF1 antibody (green, left panel). Nuclei were stained with DAPI. Scale bar: 10 µm. (H) Quantitative analysis of nuclear IRF1 levels. Mean fluorescent intensities were determined by ImageJ and reported as fold changes. More than 100 cells were counted in three independent experiments. (I) Western blotting of nuclear and cytoplasmic fractions of sictr, si11a and sihMENA(t) H1650 cells. Anti-Lamin A/C and anti-tubulin antibodies were used as nuclear and cytoplasmic markers, respectively. Immunoreactivity was determined by ImageJ and numbers indicate the fold changes of IRF1 vs Lamin optical density values. For all western blots, one representative of at least three experiments is reported. (L) Luciferase assays of IRF1 transcriptional reporter activity in sictr, si11a and sihMENA(t) H1650 cells. Cells were cotransfected with non-targeting siRNA (sictr), or hMENA^11a specific siRNA (si11a) or siRNA targeting total hMENA (sihMENA(t)) and with inducible IRF1 responsive construct expressing Renilla luciferase. 48 hours later a luciferase assay was performed. Graphs represent mean±SEM of three independent experiments. (A–C, E, H, L). P values were calculated by two-tailed Student’s t-test. (D) P value calculated by repeated measure ANOVA. *P≤0.05; **P≤ 0.01; ***P≤0.001. ANOVA, analysis of variance.

Given that hMENA^11a depletion induced IFNβ production, we investigated whether hMENA^11a silencing affects PD-L1 expression. PD-L1 transcript levels increase in NSCLC cell lines after hMENA^11a depletion (figure 2B) as well as protein levels, as evidenced by flow cytometry in si11a cells (figure 2C). It is noteworthy that the overexpression of hMENA^11a in A549, expressing low basal levels of PD-L1, counteracts PD-L1 increase mediated by IFNβ treatment (figure 2D and online supplemental figure S2A).

To be noticed, ATAC-Seq analysis revealed the existence of an open chromatin region in the CD274 locus in H1650 si11a cells, further supporting a key role of hMENA^11a in the regulation of PD-L1 expression (online supplemental figure S2B). This evidence warrants further investigation in future studies.

Then, to identify paracrine mechanisms of hMENA^11a-mediated PD-L1 regulation, we treated the NSCLC cell line H1299, not expressing hMENA^11a (online supplemental figure S2C), with the CM of H1650 cells, silenced or not (online supplemental figure S2D), and looked at STAT1 activation, considering its role in IFN-mediated PD-L1 regulation.28 We observed that only the si11a cell-derived CM, but not sictr nor sihMENA(t), activates STAT1, as revealed by the increased phosphorylation in tyrosine 701, indicative of IFN pathway activation in the recipient H1299 cell line (online supplemental figure S2E). Importantly, in parallel we observed an increase of PD-L1, as detected by flow cytometry (online supplemental figure S2F) thus indicating that si11a cell-derived CM is able to induce activation of IFN pathway and to upregulate PD-L1 expression levels in a paracrine fashion.

Overall, these data show that hMENA^11a silencing per se activates IFN-related pathways, triggering a viral response program that culminates with IFNβ production and in turn PD-L1 upregulation.

hMENA^11a silencing increases PD-L1 at mRNA and protein levels via activation of JAK/STAT1 pathway in NSCLC cell lines

Interferon regulatory factor-1 (IRF1) is a key PD-L1 transcription factor,28 thus we investigated whether the PD-L1 expression mediated by hMENA^11a silencing may be linked to IRF1 activity, which is downstream to IFN/STAT/JAK axis.

To assess the involvement of JAK/STAT1 axis in PD-L1 increase induced by hMENA^11a silencing, we used an ATP-competitive JAK inhibitor. The inhibitor completely suppressed STAT1 activation induced by hMENA^11a silencing (online supplemental figure S2G) and significantly counteracted PD-L1 mRNA upregulation (figure 2E), clearly indicating that JAK/STAT1 activation participates in PD-L1 increase.

Moreover, we revealed that the depletion of hMENA^11a upregulates STAT1 phosphorylation in all tested NSCLC cell lines (figure 2F and online supplemental figure S2H), indicative of the activation of JAK/STAT1 pathway.

To gain deeper insight into the mechanism underlying PD-L1 increase, we analyzed intracellular localization of IRF1, the positive regulator of PD-L1 transcription.29 Immunoflorescence analysis revealed that IRF1 localizes in both nuclear and cytoplasmic compartments of sictr H1650 cells. Differently its nuclear localization appeared strongly increased in si11a cells, while sihMENA(t) cells showed IRF1 localization comparable to that observed in sictr cells (figure 2G,H). Accordingly, immunoblotting analysis indicated that nuclear localization of IRF1 increases in si11a cells (figure 2I). This was confirmed by evaluating luciferase activity in an inducible IRF1 responsive reporter assay, where we found higher IRF1 activity in cells depleted of hMENA^11a, compared with sictr and sihMENA(t) cells (figure 2L). Collectively, these data indicate that hMENA^11a affects IRF1 subcellular localization and transcriptional activity, modulates JAK/STAT1/IRF1 signaling pathway and in turn PD-L1 expression.

hMENA^11a depletion perturbs cell–cell junction integrity, activates NF-kB pathway, induces proinflammatory cytokines but differs from E-cadherin loss for Type I IFN pathway activation

hMENA^11a is expressed in tumor cells with epithelial phenotype and correlates with E-cadherin expression in breast, lung and pancreatic cancers.20–22 On the other hand, hMENA^11a silencing downregulates E-cadherin expression, weakening cell–cell adhesion.22

Accordingly, in the current study, we find a significant downregulation of CDH1 transcripts in H1650 si11a cells, compared with sictr and sihMENA cells (online supplemental figure S3A, left). The decrease of E-cadherin expression in si11a but not in sihMENA(t) cells also occurred at the protein level (online supplemental figure S3A, right) and was accompanied by a critical delocalization of E-cadherin from cell–cell junctions, as observed by confocal analysis (figure 3A, upper panel).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

hMENA^11a depletion perturbs cell–cell junction, induces proinflammatory cytokines via NF-kB. E-cadherin loss does not activate IFN-I pathway. (A) Representative immunofluorescence images of sictr and si11a, sihMENA(t) H1650 cells stained with anti-E-Cadherin antibody (green) (upper panel) or with phalloidin (red) and anti-ZO-1 antibody (green) (lower panel). Nuclei were stained with DAPI. Scale bar: 20 µm. (B) ELISA for the indicated cytokines in supernatants of sictr, si11a and sihMENA(t) H1650 cells. (C) NF-kB transcriptional reporter activity was assessed by luciferase assays in sictr, si11a and sihMENA(t) H1650 cells. Cells were cotransfected with specific siRNAs and with inducible NF-kB responsive construct expressing Renilla luciferase. Forty-eight hours later, a luciferase assay was performed. (D) Western blotting of sictr, si11a and siE-Cadherin H1650 cells. Anti-HSP70 was used as loading control. (E) IFNβ expression at mRNA level (IFNB1, qRT-PCR, left) and protein level detected by ELISA in supernatants (right) of H1650 sictr, si11a, sihMENA(t) and siE-Cadherin cells. Graphs represent mean±SEM of three independent experiments. (B, C, E). P values were calculated by two-tailed Student’s t-test. *P≤ 0.05; **P≤ 0.01.

Furthermore, hMENA^11a silencing, along with a critically disorganization of F-actin, induces an important delocalization of zonula occludens-1 (ZO-1) at cell junctions, in contrast with the typical honeycomb membrane staining of ZO-1 in sictr cells (figure 3A, lower panel).

Cell–cell junction perturbation has been associated with the activation of NF-kB-driven inflammatory pathways in different tumors30 31 and accordingly, we found that hMENA^11a silencing robustly induces transcripts relative to NF-kB-dependent inflammatory cytokines such as IL6, IL8 and CXCL1 (online supplemental figure S3B).

Thus, to in depth define the pattern of cytokines modulated by hMENA^11a, we screened a panel of 40 cytokines/chemokines by a Bio-Plex assay in different NSCLC cell lines. H1650 cells produced high levels of different cytokines/chemokines, the highest being IL6. Importantly, the analysis of CM of si11a cells revealed a strong increase of IL6, IL8 and CXCL1 vs sictr H1650 cells (online supplemental figure S3C). We further confirmed by ELISA the increase of CXCL1, IL6 and IL8 levels in si11a cell-derived CM (figure 3B), compared with sictr and sihMENA(t) cells. Inflammatory cytokine regulation has been ascribed to NF-kB activity and accordingly we observed that hMENA^11a silencing induced a strong activation of NF-kB in H1650 cells, compared with sictr and sihMENA(t) cells as evidenced by luciferase assay (figure 3C).

Secretion of inflammatory cytokines and chemokines by si11a cells does not depend on cell senescence (senescence-associated secretory phenotype32). Long-term cultured cancer associated fibroblasts were used as senescence positive control (online supplemental figure S3D).

E-cadherin loss has been reported to be associated with NF-kB activation and inflammatory cytokine production.30 33 To exclude that the pathways activated by hMENA^11a silencing are not related to its effects on E-cadherin down regulation, we analyzed the direct effects of E-cadherin depletion in H1650. As expected we found that levels of mRNAs of the two NF-kB target genes, IL6 and IL8, were increased in H1650 cells depleted of E-cadherin compared with sictr cells (online supplemental figure S3E), but notably we observed that STAT1 phosphorylation at Tyr 701 is not affected by E-cadherin silencing in H1650 cells (figure 3D). More importantly, IFNβ levels were unchanged in E-cadherin depleted cells (figure 3E).

These data evidence that hMENA^11a depletion perturbs cell–cell junction integrity, activates NF-kB pathway, induces proinflammatory cytokines but differently from E-cadherin loss specifically activates Type I IFN pathway, indicating that this effect is not related to cell–cell junction deregulation but likely to actin cytoskeleton disturbance occurring in the absence of hMENA^11a.

hMENA^11a depletion induces an increase of RIG-I which sustains STAT1 activation, PD-L1 upregulation and IFNβ secretion

Then, since Type I IFN production is activated by nucleic acid sensors,34 we focused on the pattern recognition receptor RIG-I,35 associated with the actin cytoskeleton and reported as colocalized with ZO-1, which is fundamental for its localization and activation for innate immune signaling. On cytoskeleton rearrangements, RIG-I delocalizes and activates type I interferon.27

We observed that hMENA^11a silencing, along with a critically disorganization of F-actin, induced an important delocalization of ZO-1 at cell junctions (figure 3A, lower panel). We hypothesized a role for RIG-I in IFN signaling activation observed in hMENA^11a silenced cells, also indicated by increased DDX58 (RIG-I) transcript levels in si11a cell lines, compared with relative sictr cells (figure 1D). Furthermore, RIG-I expression also increases at protein level in all the si11a cell lines, compared with sictr cells (figure 4A,B and online supplemental figure S4A). It is important to note that, differently from hMENA^11a, E-cadherin silencing does not increase RIG-I levels (online supplemental figure S4B), indicating that a mechanism specifically related to actin dynamics, rather than cell junction integrity, may account for RIG-I signaling regulation.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

hMENA^11a depletion induces an increase of RIG-I which sustains IFNβ production, STAT1 activation and PD-L1 upregulation. (A, B) (Left) Western blotting analysis of sictr, si11a and sihMENA(t), siRIG-I and siRIG-I+si11a H1650 (A) and HCC2935 (B) cells. Anti-HSP70 was used as loading control. (A, B) (Right) IFNβ levels were measured by ELISA in supernatants of H1650 (A) and HCC2935 (B). Graphs represent the mean±SEM of three independent experiments. (A, B) P values were calculated by two-tailed Student’s t-test.

To determine whether RIG-I has a role in sustaining IFN activation downstream to hMENA^11a silencing, we first analyzed the effects of RIG depletion in our cell lines (figure 4A,B), then, to demonstrate that RIG-I signaling is functional in our cells, we treated cells with M8, RIG-I agonist,36 that, as expected, increased RIG-I levels and induced STAT1 activation (online supplemental figure S4C). We observed that si-RIG-I per se does not affect STAT1 expression and phosphorylation (figure 4A,B, online supplemental figure S4D and E). Differently, when we presilenced cells for RIG-I and then, 24 hours later, silenced also for hMENA^11a, we inhibited STAT1 phosphorylation, IFNβ and PD-L1 increase, indicating that RIG-I is crucial in sustaining the STAT1 pathway activation we identified in si11a cells (figure 4A,B, online supplemental figure S4D and E). Collectively these data indicate, in a panel of NSCLC cell lines, independently of their genomic alterations, that hMENA^11a regulates STAT1 activation, PD-L1 upregulation and IFNβ secretion via RIG-I.

IFNβ derived from si11a cells induces a unique macrophage phenotype and a paracrine loop which favors an EMT phenotype

Tumor cell intrinsic modifications may result in distinct immune responses, with macrophages being one of the key orchestrators of the TME. Type I IFN signaling activation characterizes a peculiar macrophage subtype, IFN-tumor-associated macrophage (TAMs), identified in different patients with cancer, including patients with lung cancer.37 We then reasoned that the secretoma and the signaling pathways activated by hMENA^11a downregulation may influence macrophage polarization. Thus, we treated human monocyte-derived M0 macrophages (MΦs) for 24 hours with the CM collected from sictr, si11a and sihMENA(t) H1650 cells. The exposure of MΦs to tumor CM induced a M2-like phenotype, as shown by the modulation of typical M2-like markers, such as CD163, CD206 and IL10 (figure 5A,B and online supplemental figure S5A) and their morphology (online supplemental figure S5B).38 Notably, macrophages treated with CM-si11a (thereafter si11a-MΦs) exhibited a unique phenotype compared with CM-sictr and CM-siMENA-polarized MΦs, with high levels of CD16, HLA-DR and CD163 and low level of CD206 receptor but also higher expression of the M1-like costimulatory molecules CD80 and CD86, (figure 5A,B and online supplemental figure S5A). Interestingly, we found higher levels of PD-L1 and PD-L2 in si11a-MΦs with respect to sictr and siMENA-MΦs and increased levels of IL10 and TGFβ cytokines (figure 5A,B and online supplemental figure S5A).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Conditioned medium derived from si11a NSCLC cells induces a unique macrophage phenotype. (A) Mean fluorescence intensity (MFI) expression levels of macrophage-specific markers shown in a representative flow cytometry analysis. Monocyte-derived M0-MΦs from healthy donors (HDs) were polarized for 24 hours with CM derived from sictr, si11a and sihMENA(t) H1650 cells (sictr-MΦs, si11a-MΦs, siMENA-MΦs, respectively). MFI values are indicated. (B) Levels of expression of macrophage markers by flow cytometry in MΦs treated as in (A) (n=9). Median value, first and third quartiles by box, minimum and maximum by whiskers. (C) Analysis of expression levels of selected macrophage markers by flow cytometry in M0-MΦs untreated or treated for 48 hours with anti-IFNAR Chain 2 Antibody (1 µg/mL), and then polarized with CM from sictr, si11a and sihMENA(t) H1650 cells (n=7). (D) (Left) Representative images of immunofluorescence analysis of A549 cells untreated (UNT) or treated with CM from M0-MΦs, CM-TUM/MΦ-sictr or CM-TUM/MΦ-si11a, stained with anti-Vimentin antibody (red). Nuclei were stained with DAPI. Scale bar: 20 µm. (Right) Quantitative analysis of Vimentin levels. MFI of vimentin/DAPI of three different fields per sample were reported as fold changes (n=7). P values were calculated by Wilcoxon rank test, with Bonferroni correction for multiple comparison. *p≤0.05. NSCLC, non-small cell lung cancer.

Considering that IFNβ has been reported to upregulate CD80 and CD8639 40 as well as IL10,41 we hypothesized that IFNβ, found at high levels in CM-si11a, may have a role in this unique MΦ polarization.

IFNβ treatment (24 hours) on monocyte-derived M0-macrophages determined a MΦ polarization status that significantly resembles the phenotype observed after CM-si11a exposure (online supplemental figure S5D). To investigate whether IFNβ signaling is required for the CM-si11a-mediated polarization of MΦs, we blocked IFN-αβ-receptor (IFNAR) for 48 hours with anti-IFNAR Chain 2 Antibody (figure 5C and online supplemental figure S5C). Noteworthy, the upregulation of IL10, TGFβ, PD-L1 and PD-L2 is abrogated in si11a-MΦs. Conversely, we observed that IFNAR blockade counteracts the CM-si11a-mediated CD206 downregulation (figure 5C). Overall, secretome of hMENA^11a depleted cancer cells, via IFNβ signaling, forces the macrophages toward a peculiar polarization, characterized by co-expression of both M1-like and M2-like markers.

To identify whether hMENA^11a loss activates a paracrine loop between tumor cells and macrophages, favoring an EMT phenotype in cancer cells, we treated A549 cells with CM derived from macrophages cultured with tumor supernatants (CM-TUM/MΦ-s). Interestingly, tumor cells treated with CM-TUM/MΦ-si11a expressed higher levels of the EMT marker vimentin, compared with cells cultured with CM-TUM/MΦ-sictr, indicating that soluble mediators produced by macrophages polarized by cancer cells depleted of hMENA^11a can favor EMT of the surrounding cells, likely affecting tumor cell invasiveness (figure 5D).

These data indicate that the secretoma and signaling pathways activated by hMENA^11a downregulation activates a paracrine loop resulting in a peculiar polarization of macrophages that favors an EMT phenotype in cancer cells.

hMENA^11a expression, IFN and macrophage scores stratify GR or PR ICB-treated NSCLC patients

To translate our experimental findings into the clinical setting, we profiled tumor tissues of 15 NSCLC patients treated with ICB therapy (nivolumab or pembrolizumab) by Nanostring nCounter PanCancer IO 360 Panel, integrated with custom probes for hMENA splicing variants. We compared the gene expression profile of seven patients classified as PRs, with eight patients classified as GRs.

The differential gene expression analysis between GR and PR patients evidenced that 318 genes were modulated in the two groups (58 genes upregulated and 260 downregulated in GRs vs PRs) (online supplemental file 2). In line with our previous data on the role of hMENA^11a as prognostic factor in NSCLC, we found that hMENA^11a transcripts (ENAHa) were downregulated in PR patients, suggesting that low expression of hMENA^11a identify patients with poor response to ICB. In line with our experimental results, genes belonging to the Type I IFN pathway, namely IRF1, IRF2, STAT1, OASL, were upregulated in PR patients (figure 6A), again supporting that hMENA^11a loss activates IFN signaling pathway. Noteworthy, we found statistically significant higher levels of ‘IFN Response Signature’ included in IO 360 Panel in PRs than GRs (figure 6B, upper panel, left), indicating that IFN-related gene expression correlates with poor clinical response to checkpoint therapy in our patient setting. This was confirmed by the analysis of ‘Interferon Response’ module expression, recently identified among recurring gene modules across diverse cancer types42 (figure 6B). Notably, we observed that this module is more expressed in our PR patients, further supporting that interferon signaling may be linked to a poor prognosis. Considering our data on the paracrine-loop between cancer cells and macrophages, we then analyzed the NanoString macrophage score (CD68, CD84, MS4A4A, CD163) in our patients and we observed an enrichment in PR patients as indicated in figure 6B, upper panel, left. Recently the IFN-TAM subset, resembling M1-like macrophages, but with immunosuppressive functions, has been reported.37 Notably this gene signature was higher in our non-responder patients (figure 6B, upper panel, right), further supporting our experimental data (figure 5) and suggesting a crucial role of IFN-TAMs in ICB response in NSCLC patients. Then, to exclude the involvement of dendritic cells (DCs) as other immune cells related to ICB response mediated by hMENA^11a downregulation and Type I IFN activation, we looked at NanoString DC score (CCL13, CD209, HSD11B1) and, differently from macrophages, we did not evidence significant differences between GR and PR patients (online supplemental figure S7A).

• Download figure
• Open in new tab
• Download powerpoint

Figure 6

Low hMENA^11a expression, high IFN pathway and Macrophage score identify poor responder ICB treated NSCLC patients. Differential Gene Expression Analysis (DGEA) between seven poor responders (PRs) and eight good responders (GRs) ICB treated NSCLC patients, by Nanostring nCounter PanCancer IO 360 Panel and custom probes for hMENA splicing variants (ENAHa for hMENA^11a). (A) Volcano plot showing statistical significance (q value, −log10 scaled) vs Fold Change (log2 scaled) for genes differentially expressed in good responder cohort (GRs) vs poor responder cohort (PRs). Genes with −log10 (qvalue) <0.1 were considered significantly modulated. Log2 (Fold Change) positive values indicate higher expression in GRs (orange), log2 (Fold Changes) negative values indicate higher expression in PRs (blue). Selected genes are reported. The complete list of differentially expressed genes between GRs and PRs is reported in online supplemental file 2. (B) Boxplots showing the comparison of scores (log2 scaled) of four reported signatures between GRs and PRs in our internal cohort. IFN-Nano: IFN Response Signature Pathway included in IO 360 Panel by NanoString (t-test p<0.01); IFN-Barkley: Interferon Response Module reported by Barkley et al42 (t-test p<0.01); Macro-Nano: Macrophage signature genes included in IO 360 Panel by NanoString (t-test p<0.05); Macro-Ma: signature genes of IFN-TAM cluster reported by Ma et al37 (t-test p<0.05). (C) Boxplots showing the comparison of scores (log2 scaled) of the signatures reported in (B) between GRs and PRs in the OAK (39 vs 35 patients) and POPLAR (11 vs 7 patients) datasets. Shown in the boxplots are the medians (horizontal lines), 25th–75th percentiles (box outlines), and highest and lowest values within 1.5 times the interquartile range (vertical lines). T test OAK dataset: IFN-Nano: P=0.0024; IFN-Barkley: P=0.0021; Macro-Nano: P<0.00001; Macro-Ma: P<0.00001. T test POPLAR dataset: IFN-Nano: P=0.0035; IFN-Barkley: P=0.0025; Macro-Nano: P=0.003; Macro-Ma: P=0.0018. * P ≤ 0.05; ** P ≤ 0.01; **** P ≤ 0.0001. ICB, immune checkpoint blockade; NSCLC, non-small cell lung cancer.

To validate findings from our discovery cohort we evaluated the IFN and Macrophage gene signatures in two large cohorts of advanced NSCLC patients treated with ICB in second or third line, the OAK (NCT02008227)43 and POPLAR (NCT01903993)44 datasets. Since the expression levels for transcript variants are not available in these datasets, we used ESRP1, the splicing regulator of 11a exon inclusion, and ENAH gene expressions as proxy to estimate hMENA^11a expression45 after validating its robustness on TCGA Lung data as shown in online supplemental methods and figure S6. Notably, we were able to confirm that patients classified as having low expression of the hMENA^11a transcript and showing poor response exhibited elevated levels of both IFN and Macrophages gene signatures (figure 6C).

A mechanistic link sustaining tumor immune evasion has been reported between TAM abundance and CD8⁺ T cell exhaustion, on prolonged residence in the TME.46 Thus, we have looked at CD8⁺ T cells and CD8⁺ exhausted cells in the three datasets of ICB-treated patients (online supplemental figure S7A–D). Of note, PRs (hMENA^11a low, Type I IFN high) in Nanostring cohort display a higher exhausted CD8⁺ T cell score. This was confirmed also in validation cohorts (online supplemental figure S7D), suggesting that T cell exhaustion is linked to macrophage abundance, as recently reported.46

Similarly to our discovery cohort, DC abundance obtained by deconvolution analysis performed with two different computational tools, did not differ among GR and PR patients in OAK and POPLAR cohorts (online supplemental figure S7B,C).

Collectively these data strongly suggest that actin cytoskeleton modification related to hMENA^11a loss may influence the interferon response, activate a loop with macrophages, likely contributing to an immunosuppressive TME which hampers the response to ICB.