Uncoupling the Senescence-Associated Secretory Phenotype from Cell Cycle Exit via Interleukin-1 Inactivation Unveils Its Protumorigenic Role

The IL-1 pathway is required for SASP expression but not for senescence-associated cell cycle exit.

Previous studies suggested that the IL-1 pathway acts as an upstream regulator of the SASP, but whether this signaling pathway is necessary for senescence-associated cell cycle exit has not been fully explored. To address this question, we introduced a tamoxifen (4OHT)-inducible oncogenic H-Ras^G12V transgene (here referred to as the Ras gene) into telomerase reverse transcriptase (hTERT)-expressing human IMR90 lung embryonic fibroblasts (IMR90T). We also introduced two independent doxycycline-inducible short hairpin RNAs (shRNAs) into these cells, which efficiently decreased IL-1R expression, as evidenced by quantitative PCR (qPCR) analysis (Fig. 1A). After 10 days of Ras activation, we observed a strong induction of the SASP factors IL-1α, IL-1β, IL-6, and IL-8 in control (shScr) cells but not in IL-1R knockdown cells, consistent with previous studies (21) (Fig. 1A). Enzyme-linked immunosorbent assays (ELISAs) confirmed the dampened Ras-induced secretion of IL-6 in IL-1R knockdown cells compared to levels in control cells (Fig. 1B). A similar IL-1R-dependent expression of SASP factors was observed when IMR90T cells were induced to senesce by exposure to etoposide (Etop), a DNA damage-inducing agent (Fig. 1G and H). These results confirm the requirement for IL-1 signaling in SASP induction in different senescence settings.

To determine whether the IL-1 signaling pathway also regulates cell cycle exit, we analyzed various hallmarks of senescence. We observed similar decreases in bromodeoxyuridine (BrdU) incorporation in Ras-activated control and IL-1R knockdown cells compared to that of ethanol (Et)-treated cells, indicative of decreased proliferation under both conditions (Fig. 1C). BrdU incorporation was also decreased in both Etop-treated control and IL-1R knockdown cells compared to that of their dimethyl sulfoxide (DMSO)-treated controls (Fig. 1I). In addition, we detected increased senescence-associated beta-galactosidase (SA-βgal) positivity in Ras-activated and Etop-treated cells compared to levels in their respective controls, regardless of IL-1R status (Fig. 1D and I). Senescence-associated heterochromatin foci (SAHF) were observed in both Ras-induced control and IL-1R knockdown cells, as well as in all Etop-treated samples (Fig. 1E and I). Finally, cytoplasmic chromatin fragments (CCFs), many of which were γ-H2AX positive, were present in all Ras-activated samples (Fig. 1F). Taken together, these results indicate that IL-1R and the IL-1-dependent SASP are not required for senescence-associated cell cycle exit or for the generation of SAHF or CCFs. Therefore, IL-1R inactivation can efficiently uncouple the SASP from other senescence-associated phenotypes.

The IL-1 pathway controls at least a majority of the SASP.

The SASP consists of a wide variety of different factors, including cytokines, growth factors, and proteases. To document the full extent of the IL-1-dependent SASP, we compared the senescence-associated transcriptome of IMR90T cells expressing shIL1R to that of control shScr cells by RNA sequencing. Cells were harvested on day 0 (d0), day 4 (d4), and day 10 (d10) of Ras induction. d0 corresponded to growing cells, d4 corresponded to the beginning of SASP induction based on a time course assay of IL-6 transcription (data not shown), and d10 corresponded to full senescence induction. As expected, the transcriptomes of shScr and shIL1R growing cells clustered close together, as evidenced by principal-component analysis (PCA) (Fig. 2A). Of note, while shScr and shIL1R transcriptomes clustered relatively close together at d4, they diverged at d10 of Ras activation, indicating that IL-1-dependent genes are expressed as part of the late arm of the senescence secretome (38). Using a log₂ fold change cutoff of 3 and an adjusted P value of <0.05, we identified highly upregulated and downregulated genes in shScr and shIL1R samples at d10 compared to levels at d0 of Ras activation (Fig. 2B; see also Tables S1 to S6 in the supplemental material). Gene ontology (GO) analyses revealed that inflammatory pathways were upregulated in shScr samples at d10 versus d0, while they were not affected in shIL1R samples (Fig. 2C). Consistent with our previous results indicating that inhibiting IL-1 signaling does not impair senescence-associated cell cycle exit, mitosis and DNA replication represented commonly deregulated pathways in both shScr and shIL1R samples (Fig. 2D).

To further explore the transcriptomic differences between shScr and shIL1R samples, we compared differentially expressed genes using a log₂ fold change cutoff of 1 and an adjusted P value of <0.05. Only 32 genes were found to be differentially expressed between shScr and shIL1R samples at d4, while 359 genes were differentially expressed between shScr and shIL1R at d10 (Fig. 2E and Tables S7 and S8). Of the 359 differentially expressed genes, 203 genes were downregulated upon IL-1R knockdown. Downregulated pathways consisted of inflammatory and immune responses, consistent with the contribution of the IL-1 signaling pathway in SASP production (Fig. 2F). Chromatin immunoprecipitation enrichment analysis (ChEA) indicated that the vast majority of the corresponding downregulated loci could be bound by RelA, the DNA-binding subunit of NF-κB (Fig. 2G). Finally, virtually all genes previously reported to be SASP factors (27) and upregulated in shScr d10 versus d0 samples were downregulated in shIL1R d10 samples, albeit at varied levels (Fig. 2H). Taken together, these results strongly support the notion that the IL-1 pathway controls the vast majority of the SASP without affecting cell cycle exit.

IL-1α signals through IL-1R to activate the SASP.

To determine the mechanism of SASP activation via the IL-1 pathway, we tested the contribution of both IL-1R ligands, IL-1α and IL-1β, to SASP expression in senescence. We first introduced a doxycycline-inducible shRNA against IL-1α into IMR90T cells and induced them to senesce via Ras activation. Western blotting and qPCR analyses validated IL-1α knockdown and indicated that IL-1α knockdown reduced IL-1β protein levels and IL-1β, IL-6, and IL-8 mRNA levels upon senescence induction compared to levels for a scramble control (Fig. 3A and B). The reduction in IL-6 transcript levels was corroborated at the protein level, as shown by ELISA (Fig. 3C). These results were further validated in IL-1α CRISPR-deleted clones, where clones were selected based on the absence of both full-length and mature IL-1α protein as detected by Western blotting (Fig. 3D). Ras-induced senescence failed to induce IL-8 expression and IL-6 secretion in these knockout cells as measured by qPCR and ELISA, respectively (Fig. 3E and F). Moreover, we validated that additional IL-1R-dependent SASP factors found in our transcriptome analysis, including CCL-3, granulocyte-macrophage colony-stimulating factor (GM-CSF), and MMP-1, were also downregulated in senescent IL-1α-deleted cells compared to levels in wild-type (WT) cells (Fig. 3G).

While both full-length and C-terminal mature IL-1α (mIL-1α) have been reported to activate IL-1R signaling, much less is known about the activity of the N-terminal propiece (ppIL-1α). ppIL-1α contains a functional NLS, but its nuclear function is elusive. Based on reported evidence that ppIL-1α may contribute to SASP activation (39), we tested the activity of full-length IL-1α, ppIL-1α, and mIL-1α upon ectopic expression in IMR90T cells (Fig. 3H and I). Expression of full-length or mature IL-1α was sufficient to induce expression of the SASP factors IL-1β, IL-6, and IL-8 (Fig. 3I). However, ectopic expression of the propiece alone failed to induce SASP factor upregulation (Fig. 3I). From these results, we concluded that nuclear ppIL-1α activity likely does not contribute to SASP activation.

We next tested whether IL-1α levels influence other senescence-associated phenotypes. We observed that ectopic expression of IL-1α was not sufficient to induce senescence, as cells continued to incorporate BrdU and remain SA-βgal negative (Fig. 4A), while IL-1α knockdown did not affect the decreased BrdU incorporation or increased SA-βgal positivity induced by Ras (Fig. 4B and C). Similarly, SAHF were also detected in all Ras-activated cells, regardless of IL-1α status (Fig. 4D). These results further support the conclusion that the IL-1-dependent SASP is not required for cell cycle exit, and disruption of the IL-1α/IL-1R signaling pathway uncouples the SASP from senescence.

IL-1β also affects the SASP without disrupting senescence-induced cell cycle exit.

We next determined whether IL-1β, which shares the same signaling pathway as IL-1α, contributes to SASP induction. After introducing an shRNA targeting IL-1β into IMR90T cells and inducing senescence via Ras activation, we performed Western blotting and qPCR analyses to confirm IL-1β knockdown (Fig. 5A and B). Surprisingly, IL-1β depletion also dampened the Ras-induced upregulation of IL-1α, IL-6, and IL-8 transcripts and IL-1α protein levels (Fig. 5A and B). The decreased expression of IL-6 was validated at the protein level by ELISA (Fig. 5C). As with IL-1α and IL-1R, IL-1β knockdown did not affect other senescence-associated phenotypes, as evidenced by decreased BrdU incorporation, increased SA-βgal positivity, and increased SAHF emergence in Ras-activated cells, all comparable to those of control senescent cells (Fig. 5D to F). From these results, we conclude that IL-1β is also required for SASP expression and can be used to uncouple the SASP from cell cycle exit.

Both IL-1α and mIL-1β can induce the SASP in the absence of the other.

We initially hypothesized that the requirement of both IL-1α and IL-1β for SASP expression reflected a hierarchical relationship between the two cytokines. As shown in Fig. 3I, overexpression of both full-length and mature IL-1α in IMR90T cells was sufficient to upregulate the SASP factors IL-1β, IL-6, and IL-8 (Fig. 6A and B). We also observed by Western blotting that mIL-1α overexpression induced endogenous full-length IL-1α in shScr control cells, consistent with IL-1α participating in a feed-forward loop (Fig. 6A). However, overexpression of IL-1α did not induce the upregulation of IL-1β, IL-6, or IL-8 upon IL-1R knockdown, further confirming that IL-1R lies downstream of IL-1α (Fig. 6A and B). Notably, overexpression of both full-length and mature IL-1α in IL-1β knockdown cells still induced IL-6 and IL-8 expression (Fig. 6B). Therefore, we concluded that IL-1β is not required for IL-1α to induce the SASP.

We next introduced full-length and mature IL-1β (mIL-1β) into IMR90T cells (Fig. 6C). Interestingly, we observed that although overexpression of mIL-1β resulted in the upregulation of IL-1α, IL-6, and IL-8 in shScr control cells, overexpression of the full-length form did not (Fig. 6D). This is consistent with previous studies demonstrating that IL-1β requires a cleavage event before it can activate IL-1R (34). Indeed, ectopic expression of full-length IL-1β does not result in the spontaneous generation of its mature form, unlike IL-1α (Fig. 6A and C). Similar to IL-1α, overexpression of IL-1β in IL-1R knockdown cells was not sufficient to induce IL-1α, IL-6, or IL-8 upregulation, confirming that IL-1R acts downstream of both IL-1α and IL-1β (Fig. 6D). Surprisingly, overexpression of mIL-1β is sufficient to induce the upregulation of IL-6, IL-8, and IL-1β even in IL-1α knockdown cells (Fig. 6C and D). From these results, we concluded that, once cleaved, IL-1β does not require IL-1α to induce the SASP.

Taken together, these results argue against the hypothesis that a hierarchical relationship between IL-1α and IL-1β dictates SASP production. Rather, our observations point to a model where a specific amount of either IL-1α or mIL-1β is sufficient to signal through IL-1R and activate the SASP. Thus, reducing the levels of either IL-1α or IL-1β would impact SASP activation.

The IL-1-dependent SASP drives pancreatic cancer progression.

Using a mouse model of pancreatic cancer, we previously demonstrated that inactivation of the senescence pathway through genetic deletion of the chromatin modifier Sin3B correlates with impaired SASP production and delayed tumor progression (28). Thus, we investigated whether ablation of the SASP through the genetic inactivation of IL-1α could recapitulate the delayed cancer progression in this mouse model of pancreatic cancer. Mice harboring a Lox-STOP-Lox-K-Ras^G12D allele were crossed with mice expressing Cre under the control of the pancreas-specific p48 promoter (here referred to as KC) (Fig. 7A). These mice develop PanINs, a precursor to malignancy, as early as 2 weeks of age, and senescent cells can be detected within low-grade PanIN lesions (40, 41). KC mice were bred in an IL-1α-deleted background and sacrificed at 12 weeks of age (Fig. 7A). Upon examination of pancreata from KC WT mice, we confirmed the presence of PanIN lesions among normal acini by H&E staining (Fig. 7B). Strikingly, KC IL-1α-deleted pancreata displayed a significantly lower percentage of neoplastic lesions (Fig. 7B). We also observed decreased desmoplastic tissue, a hallmark of pancreatic lesions, in KC IL-1α-deleted pancreata compared to that in KC WT tissue, as measured by Masson trichrome staining (Fig. 7C).

We next assessed whether immune cell infiltration was affected in KC pancreata by the absence of IL-1α. Correlating with the decrease in PanIN lesions, we observed a reduction in the number of infiltrating T cells and macrophages, as evidenced by decreased CD3 and F4/80 staining, respectively, in KC IL-1α-deleted pancreata compared to that of their KC WT counterparts (Fig. 7D and F). However, when we compared the numbers of immune cells specifically within or surrounding low-grade lesions, we did not detect any difference in CD3 staining, suggesting that the reduced amount of T cells in KC IL-1α^−/− pancreas reflected the decreased number of lesions and not reduced T cell recruitment (Fig. 7E). Strikingly, similar analyses revealed decreased amounts of macrophages per lesion in KC IL-1α^−/− pancreas, suggesting that IL-1α inactivation leads to reduced macrophage recruitment in senescent precancerous lesions (Fig. 7G).

Furthermore, we assessed whether IL-1α affected cell cycle exit and the SASP in an in vivo setting. We detected a slight but nonsignificant increase in the percentage of SA-βgal-positive senescent cells in PanINs of IL-1α-deleted mice, indicating that IL-1α deletion does not affect all senescence-associated phenotypes, consistent with our in vitro results (Fig. 7H). This increase correlated with a trending decrease in the number of proliferating cells in low-grade lesions of KC IL-1α-deleted mice compared to that of KC WT mice as evidenced by Ki-67 staining, although it failed to reach statistical significance (Fig. 7I). These results suggest that decreased proliferation in IL-1α-deleted pancreatic lesions contributes to the delayed cancer progression we observed (Fig. 7B). Lastly, we detected a decrease of the SASP factors IL-6, IL-1β, and Gro-α in KC IL-1α-deleted whole pancreas extracts compared to levels in KC WT pancreas, suggesting that SASP production requires IL-1α in this in vivo model, consistent with our in vitro data (Fig. 7J).

Together, these observations indicate that the presence of IL-1α correlates with the development of PanIN lesions and immune cell infiltration and suggest that IL-1α and the associated SASP serve as a therapeutic target to delay pancreatic cancer progression.

Cells.

IMR90 primary lung embryonic fibroblasts expressing hTERT (IMR90T) were obtained from S. Smith (NYU School of Medicine, New York, NY). Cells were cultured in minimum essential medium Eagle (MEM; Corning) supplemented with 10% fetal bovine serum (Atlanta Biologicals) and 1% penicillin-streptomycin (Cellgro). HEK293T cells (ATCC) were used to generate retro- and lentiviruses and were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Corning) supplemented with 10% donor calf serum (Atlanta Biologicals) and 1% penicillin-streptomycin. IMR90T cells were maintained in 6% O₂ and 5% CO₂ at 37°C, while 293T cells were maintained in 5% CO₂ at 37°C.

Viruses and infections.

Oncogenic ERT2-Ras^G12V was cloned into pQCXIN. Inducible shRNAs against IL-1α and IL-1R were purchased from Dharmacon, and shRNA against IL-1β was purchased from Sigma. The IL-1α coding sequence (Origene) was cloned into either MSCV or pLB(N)CX. The IL-1β coding sequence was cloned into pLB(N)CX. ppIL-1α, myc-tagged mIL-1α, and mIL-1β constructs were generated using Q5 site-directed mutagenesis (NEB) by following the manufacturer’s protocol. Retro- and lentiviruses expressing the constructs described above were generated via calcium phosphate transfection using a standard procedure. Cells were infected for 2 consecutive days and selected with 400 µg/ml G418 (ERT2-Ras) for 1 week, 1.5 µg/ml puromycin (IL-1α constructs, shRNAs) for 3 days, or 10 µg/ml blasticidin (IL-1α and IL-1β constructs) for 3 days.

CRISPR/Cas9 editing.

Lentiviral Cas9 WT construct was a gift from E. Hernando (NYU School of Medicine, New York, NY). A guide RNA against IL-1α (5′GGTAGTAGCAACCAACGGGA-3′) was cloned in pLKO-GFP. Both constructs were infected into IMR90T cells, and clones were generated via single-cell cloning in 96-well plates. Successful deletion was confirmed via Western blotting.

Senescence induction.

For Ras-induced senescence, cells were treated with 200 nM tamoxifen (Sigma) continuously for 10 days. Fresh media and tamoxifen were added every 2 to 3 days. Cells were treated with an equal volume of ethanol as a control. For etoposide-induced senescence, cells were treated with 50 µM etoposide (Sigma) or an equal volume of DMSO as a control for 48 h. Etoposide-containing medium was then replaced by normal culture medium for 7 days before harvest.

shRNA knockdown.

To induce IL-1α (no. V2THS_111503; Dharmacon) or IL-1R (no. V2THS_131081 and V2THS_131083; Dharmacon) knockdown, cells were cultured in the presence of 0.5 µg/ml doxycycline (Sigma) for 24 h before the start of senescence induction. Medium was changed and fresh doxycycline was added every 48 h until the end of the experiment to maintain knockdown. shIL-1β (no. TRCN0000058384; Sigma) was constitutively active. The levels of knockdown were determined by Western blot analysis and qPCR.

Western blot analysis.

Western blotting was performed as previously described (28). The following primary antibodies were used: mouse antitubulin (T9026; Sigma) at 1:2,000 dilution, rabbit anti-IL-1α (ab9614; Abcam) at 1:500 dilution, and anti-mouse IL-1β (MAB201; R&D) at 1:1,000 dilution.

Real-time qPCR.

Total RNA was extracted using TRIzol (Life Technologies) according to the manufacturer’s instructions. One microgram of total RNA was used to generate cDNA with oligo(dT). Real-time qPCR was performed using Maxima SYBR green (Fisher Scientific), and samples were run on a Bio-Rad iCycler MyiQ. Expression levels were normalized to those for tubulin. Primer sequences are available upon request.

BrdU incorporation.

Cells were plated in duplicate on coverslips and were pulsed with 30 µM bromodeoxyuridine (BD Biosciences) for 2 h. Cells were fixed for 10 min with 4% paraformaldehyde–phosphate-buffered saline (PBS) and incubated in 4 N HCl for 10 min to denature DNA. Cells were neutralized with 0.1 M sodium tetraborate, pH 8.5, for 7 min and then permeabilized with 0.1% Triton X-100–3% bovine serum albumin (BSA)–0.1% Tween 20–PBS for 5 min. Cells were incubated with 1:100 mouse anti-BrdU (BD Biosciences) for 75 min and 1:500 Alexa Fluor 488–anti-mouse secondary antibody (Life Technologies) for 1 h. Coverslips were then mounted on slides using mounting medium with 4′,6-diamidino-2-phenylindole (DAPI) (Vectashield). Slides were examined on a Zeiss AxioImager A2 microscope. A total of 200 cells per coverslip were counted.

SA-βgal assay and SAHF quantification.

Cells were plated in duplicate on coverslips 48 h before harvest. Cells were fixed with 2% formaldehyde–0.2% glutaraldehyde in PBS for 10 min and stained at 37°C overnight in 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) solution (1 mg/ml X-Gal, 5 mM potassium ferrocyanide, 5 mM potassium ferricyanide, 150 mM NaCl, 2 mM MgCl₂, 25.2 mM sodium phosphate, 7.36 mM citric acid in H₂O at pH 6.0). Coverslips were washed and mounted on slides as described above. Slides were examined on a Zeiss AxioImager A2. UV light was used to visualize SAHF, and bright-field microscopy was used to visualize SA-βgal positivity. A total of 200 cells per coverslip were counted.

Immunofluorescence.

Cells were plated in duplicate on coverslips 48 h before harvest. Cells were fixed with 4% paraformaldehyde–PBS for 15 min and blocked with 0.3% Triton X-100–5% goat serum–PBS for 1 h. Cells were then incubated with rabbit anti-γ-H2AX (9718; CST) at 1:400 dilution in 0.3% Triton X-100–1% BSA–PBS overnight at 4°C. The following day, cells were incubated with 1:500 Alexa Fluor 488–anti-rabbit secondary antibody (Life Technologies) in 0.3% Triton X-100–1% BSA–PBS for 1 h. Coverslips were then washed and mounted as described above. Slides were examined on a Zeiss AxioImager A2 microscope. A total of 100 cells per coverslip were counted.

ELISA.

Conditioned medium (CM) was prepared by washing cells once in PBS and incubating in serum-free MEM. CM was collected after 24 h, and cell numbers were determined. ELISA kits for detecting IL-6 were purchased from BioLegend, and ELISAs were performed according to the manufacturer’s instructions.

RNA sequencing.

RNA was harvested from 2 biological replicates using the RNeasy microkit (Qiagen). RNA quality assessment, library preparation, and sequencing were performed at the NYU School of Medicine Genome Technology Center. Strand-specific libraries were prepared using the TruSeq RNA Library Prep kit, and libraries were sequenced on an Illumina HiSeq2500 using 50-bp paired-end reads. Sequences were mapped to the hg19 genome, and analysis was done as previously described (60).

Mice.

LSL-KRas^G12D mice were gifts from T. Jacks (Massachusetts Institute of Technology, Cambridge, MA). D. Bar-Sagi (NYU School of Medicine, New York, NY) provided the p48-Cre mice. IL-1α^−/− mice were generated as described previously (61). The strains were mated to obtain mice with the correct genotypes. All animals were maintained on a C57BL/6 background, and both males and females were used. All experiments were approved by the IACUC.

Histological and immunohistochemical staining.

Pancreata from 3-month-old mice were fixed for 48 h in zinc fixative (BD Biosciences) and processed for paraffin embedding. Sectioning and histological, trichrome, and immunohistochemical staining were performed at the NYU Langone Experimental Pathology Research Laboratory as previously described (28). The following antibodies were used: anti-F4/80 (clone ND; Cell Signaling), anti-CD3 (clone SP7; Spring Biosciences), and anti-Ki-67 (clone SP6; Abcam). Slides were scanned into SlidePath and examined there. The percentage of neoplastic lesions was calculated by dividing the number of pixels that constituted normal acinar area by the total number of pixels that constituted tissue and subtracting this number from 1. The percentage of trichrome area was calculated by dividing the number of blue pixels by the total number of tissue pixels. The number of CD3-positive cells was calculated by counting the horseradish peroxidase (HRP)-positive brown cells per field of view and averaging the number over 15 fields or by counting the HRP-positive brown cells in low-grade lesions and dividing by the total number of lesions. The percentage of F4/80-positive cells was calculated by dividing the number of HRP-positive brown pixels by the total number of tissue pixels or by dividing the number of brown pixels in or surrounding low-grade lesions by the total number of lesion pixels. Representative images were taken with a Zeiss AxioImager A2.

Tissue SA-βgal staining.

Pancreata from 3-month-old mice were freshly frozen in optimal cutting temperature (OCT) compound. Frozen sections of pancreatic tissue were fixed with 2% formaldehyde–0.2% glutaraldehyde in PBS for 5 min, washed with PBS, and stained at 37°C overnight in X-Gal solution as described above. After counterstaining with Nuclear Fast Red solution (Ricca), slides were subjected to an alcohol dehydration series and mounted with Permount (Fisher). Slides were examined on a Zeiss AxioImager A2. Percent SA-βgal positivity was calculated by dividing the number of blue pixels by the total number of tissue pixels and averaging the number over 10 fields.

Statistical analysis.

Results were compared statistically using GraphPad Prism software. Values were subjected to unpaired two-tailed t tests, multiple t tests, one-way analysis of variance (ANOVA) followed by Dunnett’s multiple-comparison test, or two-way ANOVA followed by Tukey’s multiple-comparison test. All data are presented as means ± standard errors of the means.

Accession number(s).

Raw transcriptomic data were deposited at GEO and are available under accession number GSE108278.