IL-33 is expressed from multiple promoters in chronic airway disease. We have previously shown that total IL33 and the truncated isoform IL33Δ34 are increased in COPD lung tissue specimens and airway basal cells (15, 25). Within the human IL33 gene, multiple accessioned transcripts are annotated that derive from distinct promoters (Figure 1A). Two close transcription initiation sites are approximately 25 kb upstream of the first coding exon 2, each containing a noncoding exon 1 and resulting transcripts that we denote as IL33A2 (NM_001314047.2) or IL33A1 (NM_033439.4). Another transcript is initiated roughly 5 kb upstream of exon 2 (IL33B, NM_001314045.2). A fourth transcript begins in the intron between exons 2 and 3 (IL33C, NM_001353802.2). These IL33 isoforms are distinguished by their unique 5′ untranslated (5′UTR) first exon. The IL33A1 transcript was the first to be accessioned and is generally regarded as the canonical or dominant promoter driving IL33 expression, largely based on studies in the mouse system (13). However, in some human cell lines, IL33B has been shown to be the dominant transcript (26). We were able to clone transcripts designated IL33A1, IL33A2, and IL33B from human COPD airway basal cells and verify the corresponding unique exons 1A1, 1A2, and 1B as depicted in Figure 1A. When cloning with 5′ primers corresponding to the first exon of IL33A2, we were only able to isolate a hybrid transcript that was spliced to incorporate both the 5′UTR exons 1A2 and 1A1, including an additional segment between them (Figure 1A and Supplemental Figure 1; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.174786DS1). We were unable to clone a transcript containing only exon 1A2 corresponding to NM_001314047.2. We were also unsuccessful in cloning the IL33C isoform from airway basal cells; it has a unique translation start site and protein N-terminus that lacks exon 2 and part of exon 3 (27).

IL-33 expression from multiple promoters in COPD and non-COPD lung tissue. (A) Human IL33 exon structure detailed for promoter-derived transcripts containing unique upstream 5′-noncoding exons 1A (IL33A), 1A2 (IL33A2), and 1B (IL33B) upstream of first coding exon 2 and 1C (IL33C) in the intron between exons 2 and 3. (B) Transcript mRNA levels measured in lung tissue using isoform-specific qPCR assays quantified from n = 18 non-COPD and n = 20 COPD specimens; each data point represents an average measurement from 4 different lung regions. (C) Transcript mRNA levels measured in cultured airway basal cells quantified from n = 12 non-COPD and n = 25 COPD specimens. Quantity is displayed as copy/mL and normalized to GAPDH. Statistical analysis included t test (B and C), Pearson’s correlation (C). *P < 0.05, **P < 0.01, ***P < 0.001. Data is representative of n = 2 technical replicates.

We generated isoform-specific quantitative PCR (qPCR) assays to determine the relative contribution of noncanonical transcripts to total IL33 expression in COPD tissues and airway basal cells. This analysis includes lung specimens from subjects without COPD and those with severe COPD who underwent lung transplantation (Supplemental Table 1). We validated isoform-specific assays using our cloned transcripts (Supplemental Figure 2) and found that the IL33A1 qPCR assay detects both IL33A1 and the IL33A2 hybrid transcripts based on plasmid standards. qPCR performed on non-COPD and COPD airway tissue demonstrates that all 4 IL33 transcripts are significantly increased in COPD tissue specimens with a wide range of expression levels observed (Figure 1B). While both IL33A1 and IL33B expression were similar in lung tissue, specific IL33A2 and IL33C expression was 2–3 orders of magnitude lower, indicating that these promoter-driven isoforms make a smaller contribution to total lung IL33 mRNA. We also quantified full-length IL33 (IL33full) and IL33Δ34 levels for this cohort using previously designed assays (25), which again demonstrated a significant increase in COPD relative to non-COPD controls. We then performed correlation analyses for the more abundant IL33A1 and IL33B transcripts, each with IL33Δ34 expression in lung tissue; this demonstrated significant association only for IL33B. These data show that multiple IL33 promoters are active in lung tissue with a range of expression levels that are overall higher in COPD specimens. Furthermore, the canonical IL33A transcript and noncanonical IL33B transcript appear to contribute similarly to total IL33 expression.

A noncanonical IL33 transcript contributes to alternative splicing. Given the observed correlation between IL33B and IL33Δ34 expression in lung tissue, we hypothesized that truncated spliced isoforms may be preferentially derived from the IL33B transcript. Precedent for this concept includes a previous report on bcl-X, which exhibits distinct splicing patterns based on interactions between alternate promoter–derived 5′UTR sequences with a common 3′UTR sequence (28). We similarly analyzed a cohort of COPD and non-COPD cultured airway basal cells by qPCR using the same assays as in Figure 1. We found that IL33A1 and IL33B transcripts were both abundantly expressed, with the noncanonical IL33B transcript expressed at slightly higher levels relative to the canonical IL33A1, and both IL33A2 and IL33C were again 2–3 orders of magnitude lower (Figure 2A). None of the transcripts were significantly increased in COPD basal cells compared with non-COPD in culture, and this may reflect altered promoter activity influenced by culture conditions or passaging of cells.

IL33 spliced isoform is derived from a noncanonical promoter. (A) Transcript mRNA levels measured in cultured airway basal cells using promoter-specific qPCR assays as in Figure 1, quantified from n = 12 non-COPD and n = 25 COPD specimens. (B) Nested PCR based on isoform-specific cDNA generation followed by amplification using primers encompassing the IL33 coding region. Primer sets are color coded according to location, and assay steps are as depicted in the schematic. Agarose gel analysis for a subset of COPD specimens demonstrating truncated bands observed only for PCR amplification on IL33B. (C) qPCR using primer-probe sets designed to detect the truncated IL33Δ34 transcript using random primer-generated cDNA template from A, quantified as copy/mL using plasmid standard and normalized to GAPDH. Statistical analysis included t test (A and C). Data in A are representative of n = 2 technical replicates; experiments in B and C were repeated in triplicate.

To test whether IL-33Δ34 was differentially generated from transcripts containing either exon 1A1 or 1B, we generated complimentary DNA (cDNA) from COPD basal cells using transcript-specific primer sets including unique 5′ exons 1A1 or 1B and the same 3′ primer corresponding to the protein C-terminus in exon 8 (Figure 2B). We then amplified each reaction with cloning primers specific to the IL-33 coding region (exon 2–8) and analyzed by agarose gel, which demonstrated a single PCR product based on IL33A1 for the specimens tested but multiple truncated bands for IL33B (Figure 2B). We isolated and sequenced these PCR products and found that IL33A cDNA yielded only IL33full, while IL33B yielded IL33full, IL33Δ34, and IL33Δ345 spliced products. We then performed qPCR analysis on our cohort of COPD and non-COPD airway basal cells shown in Figure 2A using qPCR assays designed to detect the truncated IL33Δ34 isoform in the context of IL33A1 or IL33B transcripts (Figure 2C). Specificity was validated for these assays, as shown in Supplemental Figure 2. Similar quantities of IL33A and IL33B transcripts were amplified in the random-primer generated cDNA libraries per Figure 2A, but the IL33Δ34 specific assays only amplified on the B amplicon, no signal was detected from the A amplicon. We have previously shown the IL-33Δ34 isoform can be tonically secreted from airway basal cells (25), and these results provide insight into regulation of truncated spliced isoforms; among these, the IL-33Δ34 isoform appears to be preferentially generated through noncanonical IL33B promoter usage.

Phorbol esters induce noncanonical IL33 promoters through PKC. In order to test factors that influence expression from noncanonical IL-33 promoters, we first characterized 3 cell lines that approximate airway basal cells in vitro; HBE-1 (HBE), Beas2B (B2B) and 16HBE14o- (16HBE) (29–31). We observed that HBE expressed high levels of IL33A and IL33B transcript at baseline, where B2B expressed only IL33B at 1,000-fold lower levels, and 16HBE exhibited no detectable IL33 expression; these results are quantified and summarized with a heatmap in Figure 3A. To identify stimuli capable of inducing IL33 expression from the 4 promoters, we curated the literature to assemble a panel of reported stimulants of type 2 inflammation or airway basal cell activation, proliferation, or differentiation (Supplemental Table 3). We performed a 12-hour expression screen for HBE and B2B cell lines along with 3 non-COPD low-passage primary airway basal cell specimens. The cDNA library was generated using a cells-to-cDNA kit, and qPCR for promoter-driven isoforms was performed. Neither baseline or induced IL33A1, IL33A2, or IL33C were detectable due to low cDNA yield (not shown); however, IL33B expression was sufficient to be analyzed for all 5 specimens. A heatmap representing fold-change IL33B expression relative to PBS for each cell type is shown in Figure 3B, which demonstrates that most conditions exhibited variability across specimens. Importantly, comparison of HBE and B2B lines illustrates that the fold-change IL33B expression induction was related to baseline expression levels, with high baseline–expressing cells (HBE; primary basal cells) demonstrating lower-magnitude induction relative to low baseline cells (B2B). The only stimulant in this screen that consistently induced IL33B in all cell lines and the 3 primary cell specimens was PMA. Follow-up time course analysis performed in HBE, B2B, and 16HBE cell lines at 6 and 12 hours demonstrated reproducible induction after PMA treatment (Figure 3, C and D). In HBE, all 4 IL33 transcripts were induced by PMA at 6 and 12 hours, though with somewhat different kinetics. We observed IL33B expression increased 4-fold at 6 hours and began to decay by 12 hours, while IL33A1 was increased by 2-fold at 6 hours and increased further at 12 hours after PMA treatment. The kinetics of IL33Δ34 expression mirrored IL33B, consistent with our observations that this spliced isoform appears to be preferentially derived from the IL33B transcript. The IL33A2 and IL33C transcripts were also induced by PMA in HBE but with expression levels 2–3 orders of magnitude lower and with kinetics similar to IL33A and IL33B, respectively. In the B2B line, expression of IL33B was at a low level at baseline but increased 50-fold following PMA treatment. In 16HBE cells, IL33B only reached detectable levels with PMA treatment. To test promoter responses in other cell types, we also treated Jurkat cells with PMA (Supplemental Figure 3). We observed IL33A expression detectable at very low levels in Jurkat cells and found that it was not PMA responsive; however, induction of IL33B and IL33A2 was observed and appeared to be PMA specific, as a similar response was not induced with LPS treatment.

PMA induces IL33 expression in airway basal cells and lines. (A) Baseline expression of IL33 isoforms as defined in Figure 1; summary heatmap of relative expression is shown to the left. White indicates no detection. (B) IL33B expression screen (represented as fold change ΔΔCt normalized to GAPDH) in HBE and B2B cell lines and 3 non-COPD airway basal cell specimens, demonstrating consistent induction across cell types with PMA (20 ng/mL). (C) Time-course PMA treatment in HBE cells demonstrating induction of all promoter-driven isoforms and IL33Δ34. (D) Time course as in C performed for B2B and 16HBE cells. (E) Protein induction measured by anti-IL33 Western blot (C-terminal domain, R&D systems; Supplemental Table 2) performed on HBE, B2B, and 16HBE cell lysates at 6 and 12 hours after PMA. (F) Expression analysis of IL33A1 and IL33B for PMA induction without and with PKC inhibitor Go6983 (500 nM), shown for both HBE cells and non-COPD airway basal cells. Statistical analysis included 1-way ANOVA (C, D, and F),. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. A and C–F were repeated in triplicate. B is representative of duplicate technical replicates.

On the protein level, the Western blot demonstrated induction of IL-33 protein at 6 and 12 hours after PMA treatment in both HBE and B2B cells (Figure 3E). No detectable IL-33 protein induction was observed for 16HBE. Protein expression levels were higher in HBE versus B2B, reflecting IL33 expression differences observed by qPCR. Of note, the magnitude of protein induction with PMA at 12 hours appears greater than 2-fold in HBE, suggesting that PMA may exert additional posttranscriptional effects on cellular IL-33 protein levels in airway epithelial cells.

To investigate the mechanism of PMA-induced IL-33 expression, we examined the potential role of PKC, given that phorbol esters can activate both PKC and cAMP pathways (32). We treated HBE, B2B, and primary airway basal cells with the PKC inhibitor Go6983 during PMA stimulation and found that this kinase inhibitor blocked PMA induction of IL33 expression in both airway cell lines and primary cells, with a return to baseline expression levels (Figure 3F and Supplemental Figure 3B).

Together, these results demonstrate that expression activity from all 4 IL33 promoters can be induced by PMA through a PKC-dependent pathway. The magnitude of induction appears to be, in part, related to baseline expression level, as seen for the fold-change induction in HBE (high baseline) compared with B2B (low baseline). However, there do also appear to be cell-specific differences in induction behavior for low-baseline cells — for example, in 16HBE, where PMA induction was observed but much more limited compared with B2B, with no protein induction observed. Such factors may also be relevant to the wide variability in expression response observed for the 3 primary basal cell specimens. Last, IL-33 protein levels are robustly increased by PMA in HBE cells, which may be due to combined transcriptional and translational effects.

AP TFs coordinate noncanonical IL-33 expression. To investigate the TFs responsible for PMA-induced IL-33 expression, we performed a competition-based TF binding array (Signosis). We amplified the A1, A2, and B core promoter regions (600 bp upstream of exons 1A1, 1A2, and 1B) from HBE cells and verified by Sanger sequencing that they matched the hg38 reference genome. We then used these PCR products as competitive DNA in the TF array to assess binding to a panel of 48 TFs in HBE cells following 6 hours of PMA treatment (Figure 4A); readout is the percent inhibition of binding to consensus DNA. Results demonstrate Transcription Factor IID (TFIID, part of RNA polymerase II pre-initiation complex) positive control binding to each A1, A2, and B promoters. The array also yielded several TFs that exhibited binding to 1 or more IL33 promoter sequences, with some apparently shared and some unique to a given promoter. Among the TFs included in the array, AP-1 and AP-2 TFs were of particular interest because they are known to be activated downstream of PMA and PKC (32, 33).

Transcription factor array and follow-up analyses in HBE cells. (A) Transcription factor (TF) binding array results for IL33 A1, A2, and B promoter sequences (600 bp upstream of transcription start site) for HBE cells treated with PMA for 6 hours. Percent inhibition is calculated relative to no-input control luciferase signal, expressed as percent reduction in relative luminescence units. Red indicates decreased luciferase signal, and bule represents increased signal compared with no-input control. TFIID is a positive control for RNA-polymerase transcription initiation complex indicating promoter function. Solid red boxes indicate shared TF hits; dashed red boxes indicate apparent unique hits. (B) Follow-up IL33 expression analysis with 6-hour treatment for candidate transcription factor activators identified from array; see Supplemental Figure 4 for cross-referenced TF expression. Activator concentrations are reported in Supplemental Table 4. (C) Expression for IL33A1, IL33B, IL33A2, and IL33C with lentiviral mediated RNA interference for TFAP2A, TFAP2C, and JUN in HBE cells. See Supplemental Figure 5 for knockdown validation. (D) HBE IL33A1 and IL33B expression in cells treated with AP-1 inhibitor T-5224; concentration in Supplemental Table 4. (E) Endogenous IL-33 protein secretion and total protein levels measured under RNA interference conditions with 12-hour PMA treatment, quantified by ELISA. (F) Anti–IL-33 Western blot performed under conditions of RNA interference and 12-hour PMA treatment, using both N-terminal domain (NTD) and C-terminal domain (CTD) targeting antibodies. Bands consistent with full-length (IL-33full) and IL-33Δ34 are labeled; MW fragment ~24 kDa is only reactive with CTD antibody. Markers were run on the same gel for each Western blot; they were cropped for the figure and are indicated by a white line with black border. Statistical analysis included 1-way ANOVA (B, C, and E) and t test (D). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. A represents a single experiment; B–F are representative of triplicate repeats.

To identify candidates for follow-up analysis, we cross-referenced TF array hits to airway basal cell expression in reference to other human lung structural cells, using available single-cell RNA-Seq data from the LungMAP consortium (https://app.lungmap.net). Expression analysis revealed that AP-1 and AP-2 family members were among the TFs with the highest expression in airway basal cells of those included in the array (Supplemental Figure 4). To comprehensively address both shared- and differential-binding TFs, we focused follow-up analysis on TFs that exhibited more than 30% inhibition in the assay with binding to 2 or more promoters (AP-1, AP-2, HIF, IRF, NF-κB, TCF/LEF, and TR; red boxes in Figure 4A) or apparent preferential binding to an individual promoter (GAS/ISRE, GR/PR, STAT1; dashed red boxes in Figure 4A). Of note, potent and highly specific inhibition of estrogen receptor (ER; ESR1) binding was observed only for the A1 promoter (Figure 4A). This may reflect the female source of the HBE cell line, while the B2B and 16HBE lines are derived from male donors. This observation was not pursued in the current follow-up assays because expression of ESR1 in airway basal cells could not be confirmed based on LungMAP single-cell data (Supplemental Figure 4) or qPCR in HBE cells (not shown).

When we tested a panel of TF activators in high-baseline HBE and primary basal cells, we found that, among the agents tested — dexamethasone (GR), IFN-β (STAT1, IRF, GAS/ISRE), LPS, poly I:C (NF-κB), T3 (TR), or CHIR99021 (TCF/LEF) — none increased IL33A or IL33B expression to the same degree as PMA, though a more modest IL33B induction was observed for polyI:C in HBE (Figure 4B). Some activators resulted in a significant decrease in expression for multiple promoter-derived transcripts in HBE cells, most notably CHIR99021 and dexamethasone. This effect was less pronounced when confirmed in non-COPD primary cells (Supplemental Figure 5). In an effort to identify possible disease-relevant upstream factors that may be a physiologic source for the PMA effect on IL33 expression, we tested Alternaria alternata extract, protease activated receptor (PAR) agonist SLIGKV-NH2, and IL-4 in HBE cells. None of these stimulants increased IL33 transcript expression (Supplemental Figure 5), suggesting that the relevant stimuli in vivo may be more complex and possibly involve multicellular or multi-hit mechanisms to drive IL-33 expression.

To test the effect of AP-1 and AP-2 TFs identified in the TF array, we conducted RNA interference experiments in HBEs (Figure 4C). Cells were transduced with lentiviruses expressing short hairpin RNA (shRNA) targeting AP-2 family members TFAP2A and TFAP2C and the required component of the AP-1 complex JUN. Expression of IL33 transcripts was measured under the knockdown conditions and demonstrated a modest effect of TFAP2A, with a more marked effect with TFAP2C and JUN. The effect of TFAP2A knockdown appeared to be greater for IL33A than the noncanonical transcripts, suggesting that hetero- or homodimers of AP-2 TFs may differentially affect promoter activity. To confirm AP-1 regulation using a complimentary approach, a specific AP-1 inhibitor T-5224 was also tested that demonstrated significant reduction of IL-33 expression from all 4 promoters (Figure 4D and Supplemental Figure 5).

To determine whether PMA-induced IL-33 protein could also be secreted, ELISA was performed on cell supernatants and lysates following 12 hours of PMA treatment (Figure 4E). For TFAP2A knockdown, there was a moderate reduction in IL-33 protein, while TFAP2C and JUN both demonstrated a more pronounced effect on both secreted and total cellular protein. To confirm that secreted protein measured by ELISA was not due to cell death induced by PMA, lactate dehydrogenase (LDH) release assays were performed under knockdown conditions and demonstrated minimal cytotoxicity (Supplemental Figure 5).

We also analyzed PMA induction of IL-33 protein by Western blot under knockdown conditions. This revealed a similar pattern of reduced protein expression with knockdown of AP-2 and AP-1 TFs, and it included both an apparent ~34 kDa band corresponding to IL-33full and with longer exposure a ~24 kDa lower band (Figure 4F). We attribute this lower band to protein expressed from a spliced isoform, most likely the exon 3–4 deletion variant (IL-33Δ34), as the protein was reactive only with a C-terminal domain (CTD) targeting antibody, not an N-terminal domain (NTD) targeting antibody, as previously described (25). Notably both the full-length and truncated bands show induction with PMA under control conditions, and this induction is diminished under knockdown conditions, most prominently for JUN.

These results show that PMA reproducibly induces IL-33 in HBEs through a mechanism that involves coordination of AP-1 and AP-2 TFs to drive noncanonical promoters and associated transcripts. Furthermore, PMA induction of IL-33 cellular protein allows for secretion that is, in part, dependent on AP TF activity. Western blot analysis comparing NTD and CTD reactive antibodies suggest that secretion could be related to increased production of IL-33Δ34 protein, which we have previously demonstrated is tonically secreted from airway cells through an extracellular vesicle–mediated (EV-mediated) pathway (25). Measurement of LDH release did not demonstrate substantial cellular toxicity under assay conditions, supporting a regulated secretion mechanism.

Epigenetic regulation of AP-mediated IL-33 expression. To further investigate how AP TFs regulate multiple promoters across the ~42 kb IL33 gene locus, we took advantage of the observation that HBE, B2B, and 16HBE cell lines exhibit marked baseline differences in IL33 expression from canonical and noncanonical promoters (Figure 3A). Because both AP-1 (33, 34) and AP-2 (35) TFs can be activated by phorbol esters, and AP-2 can function as a pioneer TF to regulate stem cell associated gene programs (36, 37), we examined IL33 expression in the context of multiple AP-1 and AP-2 responsive promoters and/or enhancers within the gene locus.

First, we measured relative expression levels of AP-1 and AP-2 TFs in the cell lines under study and primary cells according to differentiation state (Figure 5A). Analysis of cell lines demonstrated similar levels of TFAP2A and TFAP2C expression, although with a significantly lower level for TFAP2A in HBE and B2B relative to 16HBE. AP-1 TFs can function as homo- or heterodimers, with JUN being the required subunit for activity (38). We examined both FOS and JUN expression patterns in the cell lines, and this revealed a very large differential expression for FOS in HBE relative to B2B and 16HBE and a less prominent though significant pattern observed for JUN. We then analyzed a non-COPD specimen cultured in a polarized format as undifferentiated basal cells or differentiated at air-liquid interface (ALI) for 3 weeks (Figure 5B). Under differentiated conditions, basal cells would be reduced in number, and bulk mRNA expression should reflect increased presence of ciliated, mucus, and secretory cell populations. Comparison of differentiation states revealed that TFAP2A, TFAP2C, and JUN were significantly decreased when cells were differentiated at ALI (Figure 5B). There was also a corresponding decrease in noncanonical IL33 transcript expression, though canonical expression appeared preserved.

Epigenetic modulation of expression from IL33 locus. (A) Expression levels of AP-2 (TFAP2A, TFAP2C) and AP-1 (FOS, JUN) transcription factors measured in HBE, B2B, and 16HBE cell lines. (B) Expression analysis for AP-1 and AP-2 TFs and IL33 promoter–driven isoforms in undifferentiated primary airway basal cells and after 3 weeks in air-liquid interface (ALI) conditions, performed on the same non-COPD specimen. (C) ATAC-Seq performed on airway cell lines with chromatin-accessible peaks superimposed on the IL33 promoters region, also mapped for comparison: whole-lung ATAC-Seq, RAMPAGE (promoter mapping), cis-regulatory elements (CRE), and H3K27 acetylation marks on human hg38. Promoter-derived transcripts are labeled. Predicted JASPAR transcription factor binding sites for AP-1 and AP-2 are labeled green and blue, respectively; dashed boxes represent putative enhancers and are color coded based on activator protein transcription factor sites. Summary of ClinVar SNPs is displayed as PubMed SNPs. (D) ChIP-PCR analysis performed for each of the IL33 promoters and the labeled Intron 1 Enhancer (Int1 Enh). Immunoprecipitation was performed with AP-2α, AP-2γ, FOS, and JUN antibodies, as shown in the schematic. qPCR was performed with promoter-specific primer sets and reported as fold-enrichment (ΔΔCt) relative to bead-only control. Statistical analysis included 1-way ANOVA (A), t test (B), and multiple t test (D). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Experiments in A and B are representative of duplicate experiments, C is a single experiment, and D is representative of triplicate repeats.

Together, these findings provide support for the role of AP-1 and AP-2 TFs in regulating IL33 cellular expression, both in the high-baseline context of the HBE cell line and in basal progenitor cells. These results are consistent with prior reports implicating AP TFs in maintenance of stem cell programs and illuminate a potential basis for observed basal cell–restricted IL33 expression in human airways.

To address the role of altered chromatin accessibility in AP TF activity and IL33 expression patterns, we analyzed the 3 human cell airway cell lines with distinct baseline IL33 expression levels by ATAC-Seq (Figure 5C). We identified multiple chromatin-accessible peaks within the IL33 gene locus and compared them with whole-lung ATAC-Seq data and RNA Annotation and Mapping of Promoters for the Analysis of Gene Expression (RAMPAGE) data available through ENCODE (https://www.encodeproject.org; ref. 39), with custom tracks displayed on the genome in Figure 5C. Whole-lung ATAC-Seq performed on normal human specimens revealed prominent ATAC-Seq and RAMPAGE peaks within the IL33A2 and IL33A1 promoters, which may reflect a predilection for canonical IL33 expression in multiple lung cell types including endothelial cells (40). In contrast, the HBE cell line exhibited prominent ATAC-Seq peaks corresponding to the noncanonical B promoter but not A2 or C promoters (highlighted by solid boxes), consistent with our noncanonical expression data for these transcripts. Of note, HBE cells still express IL33A, despite no prominent peak corresponding to the A1 promoter, as present in whole lung data. The B2B line demonstrated peaks only at the B and C promoters, while near-background peaks were observed across all promoters in 16HBE cells, consistent with undetectable expression in this cell line. In addition to core promoter regions, there are several putative enhancers throughout the IL33 locus, highlighted by ENCODE cis-regulatory elements (CRE) and histone lysine acetylation (H3K27Ac) marks in Figure 5C. Multiple ATAC-Seq peaks are observed for the cell lines within the exon 1A1-1B intronic region corresponding to CRE and H3K27Ac marks (highlighted by dashed boxes; Figure 5C), which could function as enhancers and potentially explain the observed IL33A1 expression in HBE cells. We also analyzed JASPAR-predicted (41) AP-1 and AP-2 TF binding sites in the IL33 locus, which were located in multiple promoters and putative enhancer regions as highlighted in Figure 5C. The A2, B, and C promoters all contain a proximal AP-1 site with adjacent AP-2 sites for B and C, while A1 has neither of these sites in the core promoter region. Of note, there are additional AP-2 sites located in 3 putative enhancers within the exon 1A1-1B intron, which correspond to our ATAC-Seq peaks in HBE as well as CREs and H3K27 acetylation marks. The putative enhancer region closest to the A1 promoter has been designated Intron 1 Enhancer (Int1 Enh) in Figure 5, C and D.

Given the presence of IL33A1 expression in HBE despite the absence of an ATAC-Seq peak in the promoter region, we performed ChIP-PCR analysis using unstimulated and PMA-stimulated HBE cells with IP for antibodies to AP-2α, AP-2γ, Fos, and Jun. Results are graphed as fold-change enrichment relative to bead control in Figure 5D. This analysis revealed that there was relatively low-level enrichment of the A1 promoter for AP-2 targets, but there was moderate baseline enrichment for AP-1 targets, which decreased in response to PMA. In contrast, the A2 promoter was enriched at baseline for AP-2 and AP-1 targets, with both Fos and Jun ChIP demonstrating a high level of baseline enrichment — 100- and 1,000-fold, respectively. In response to PMA treatment, AP-1 factors did not further enrich A2, but there was a dramatic effect for AP-2 TFs (100-fold increase), suggesting part of the PMA effect on IL33 expression is mediated through the region of the A2 promoter. Likewise, AP-2 and AP-1 TFs mildly enriched the B promoter, which was augmented by PMA for AP-2 TFs. The pattern observed for the B promoter was mirrored at the C promoter, with mild enrichment for AP-2 TFs with PMA. Interestingly, analysis of the first putative Int1 Enh located ~1 kb distal to the A1 promoter demonstrated high level enrichment with AP-1 targets at baseline and a strong induction of AP-2 enrichment with PMA. This result, in combination with observations for the A2 promoter, suggests that AP TFs could modulate IL33A1 and IL33B expression through nearby enhancers.

Multiple prior reports have implicated SNPs within the IL33 locus with disease-associated changes in IL33 expression (42–44). Based on our results above, we cataloged SNPs reported in PubMed from ClinVar database (https://www.ncbi.nlm.nih.gov/clinvar) that were inclusive of the IL33 promoter and enhancer regions (chr9: 6193455–6248408) and overlayed on our ATAC-Seq data in Figure 5C. While none of these SNPs localized to the core IL33A1, IL33A2, or IL33B promoter regions, some overlapped with ATAC-Seq peaks in the C promoter region (rs79981454, rs7037276, rs10975516, rs11792633, and rs76864631). Some clinically relevant SNPs were also located approximately 1,500 bp proximal to the A2 promoter (rs928413 and rs7848215) and others 400 bp distal to the A1 promoter (rs1157505 and rs11791561), but none of these corresponded to ATAC-Seq peaks in our data set. Multiple SNPs were located within ATAC-Seq peaks in the 1A1-1B intron region that correspond to putative enhancers (dashed boxes, including rs72614080, rs16924159, rs16924161, and rs12551256). Among these, rs72614080 is associated with prognosis in osteosarcoma (45), rs16924159 with vascular disease (46), and an off-peak SNP rs1891385 with COPD (47). Other SNPs associated with airway disease (42) are located ~6 kb proximal to the A2 promoter (rs992969 and rs3939286) and do not correspond to ATAC-Seq peaks observed in this data set; however, an AP-1 binding site was reported within the region encompassing these SNPs.

These data indicate that IL33 expression from canonical and noncanonical promoters is coordinated by interaction of AP-1 and AP-2 TFs with IL33 promoters and enhancer regions. This regulatory program appears to be active in airway basal progenitors and may be lost as cells become differentiated. Expression from multiple noncanonical IL33 promoters can be partly explained by differences in chromatin accessibility at core promoters and may also include long-range effects mediated through enhancers in the 1A1-1B intron. Polymorphisms within the IL33 locus have been studied; however, based on the ATAC-Seq data from airway cell lines, these do not correspond to peaks within IL33 core promoter regions. Instead, they appear to be associated with enhancer regions, as an AP-1 site is located in an upstream regulatory region and a SNP associated with COPD is located near the 1A1-1B intronic enhancer region. There also remains potential for numerous uncharacterized SNPs within IL33 enhancers to regulate expression from canonical or noncanonical promoters in airway cells.

AP TFs are increased in COPD. Given our findings that AP-1 and AP-2 TFs regulate canonical and noncanonical IL33 promoters, we examined TF expression patterns in COPD and non-COPD lung tissue. On the mRNA level, all 4 AP-1 and AP-2 TFs examined were significantly increased in COPD relative to non-COPD control lung tissue (Figure 6A). A similar pattern was observed in primary airway basal cells, but only the AP-2 TF TFAP2A was significantly increased in COPD compared with non-COPD cells (Figure 6B). Both FOS and JUN were markedly and significantly upregulated in lung tissue and airway cells. On the protein level, tissue immunostaining demonstrated that AP-2α and AP-2γ were both enriched within COPD airways and staining colocalized at the base of the epithelium within IL-33+ basal cells (Figure 6C). To evaluate potential AP-2γ activation in lung tissue, sections were stained for phosphorylated protein, which revealed focal nuclear staining in IL-33+ basal cells. These data support a role for increased AP-1 and AP-2 TF activity in COPD disease, which we have shown functionally induces IL33 expression from both canonical and noncanonical promoters that influence alternative splicing patterns.

AP transcription factors in COPD. (A) TFAP2A, TFAP2C, FOS, and JUN expression in COPD lung tissue demonstrates significantly increased expression for all transcription factors in COPD relative to non-COPD, represented as fold change ΔΔCt normalized to GAPDH. Cohort includes n = 18 non-COPD and n = 20 COPD specimens; each data point represents an average measurement from 4 different lung regions. (B) Analogous expression levels in airway basal cells for n = 12 non-COPD and n = 25 COPD specimens. (C) IHC staining of non-COPD and COPD lungs for IL-33 costaining with AP-2α or AP-2γ demonstrates enrichment of TF signal in COPD airways. Costaining of IL-33 and phospho–AP-2γ demonstrates enrichment of nuclear AP-2γ signal in COPD tissue. Images taken with 40× magnification. Scale bar: 10 mm. Statistical analysis included t test (A and B). *P < 0.05, **P < 0.01, ***P < 0.001. Experiments in A and B are representative of duplicate technical replicates; C staining was repeated in triplicate.