Dual promoter architecture of NFE2L2 underlies isoform-specific translational control of NRF2

Abstract

Transcription factor NRF2 (encoded by the NFE2L2 gene) is a central signalling hub that coordinates antioxidant, detoxification, anti-inflammatory and metabolic programmes to maintain cellular homeostasis, with broad relevance to degenerative diseases, chronic inflammation and cancer. NFE2L2 is transcribed from two promoters: the canonical promoter (P1) and an alternative promoter (P2), generating transcripts that differ solely in their first exon. Exon 1 of P1-derived transcripts contains the AUG start codon, whereas the considerably longer exon 1′ of P2-derived transcripts forms an extended 5′ untranslated region (5′UTR). Consequently, the P1-derived NRF2 isoform 1 contains 16 additional N-terminal amino acids relative to the P2-derived isoform 2. Although transcripts from both promoters are expressed in human tissues, the existence of isoform 2 protein and the functional significance of alternative promoter usage have remained unresolved.

Here, we identify an endogenous NRF2 protein species consistent with isoform 2 using an isoform-specific siRNA strategy. Isoform 2 is negatively regulated by KEAP1 and retains the ability to activate canonical NRF2 target genes. Despite similar stability of the two isoforms, isoform 2 accumulates at substantially lower basal levels, which is associated with reduced translational engagement of P2-derived transcripts. Polysome profiling revealed that P1-derived transcripts are predominantly associated with polysomes, whereas P2-derived transcripts are distributed between polysomal and subpolysomal fractions. The extended P2-specific 5′UTR contains evolutionarily conserved structural features that may contribute to translational regulation. Finally, the electrophilic NRF2 activator omaveloxolone (RTA-408) induced translation of transcripts from both promoters, demonstrating that NRF2 stress response is coordinated through the combined regulation of mRNA translation and protein stability.

Highlights

• •Human NFE2L2 is transcribed from the canonical P1 and alternative P2 promoters, generating transcripts that differ only in their first exon.
• •P2-derived transcripts contain a fully untranslated exon 1′, whereas exon 1 of P1-derived transcripts is partially coding.
• •An endogenous NRF2 protein species consistent with isoform 2 is regulated by KEAP1, while its low basal abundance results from reduced translational efficiency.
• •The electrophilic NRF2 activator omaveloxolone (RTA-408) induces translation of transcripts from both NFE2L2 promoters, linking translational and post-translational regulation of NRF2.

1. Introduction

Nuclear Factor Erythroid 2-like 2 (NFE2L2, NRF2), first isolated in 1994 [1], was discovered as a transcription factor inducing expression of detoxification enzymes (GST, NQO1) and antioxidants (PRDX1, SRXN1, TXNRD1), which protect mice from chemically-induced tumors [[2], [3], [4]]. Subsequent work revealed that NRF2 controls a much broader cytoprotective program that includes ABC transporters, NADPH-generating enzymes, and heme and iron metabolism pathways [[5], [6], [7]]. Owing to its central role in adaptation to oxidative, inflammatory, and metabolic stress, maintaining cellular resilience and modulating ageing, the transcription factor NRF2 has emerged as an attractive therapeutic target in preventive and personalized medicine [8,9]. Hormetic NRF2 activation by electrophilic compounds representing various chemical classes is a basis for chemoprevention and treatment of inflammatory, neurodegenerative and metabolic diseases [10,11].

NRF2 activity is considered to be regulated predominantly at the post-translational level rather than transcriptionally. Accordingly, oxidative or electrophilic stress does not significantly induce NFE2L2 gene transcription, but rather increases NRF2 protein abundance by inhibiting its degradation [[12], [13], [14], [15], [16], [17]]. Under basal conditions, KEAP1 dimer anchors NRF2 in the CULLIN-3 ubiquitin ligase complex and mediates its ubiquitination and constitutive degradation [[18], [19], [20], [21], [22], [23]]. Upon stress, KEAP1 cysteines become oxidized or covalently modified by electrophiles, altering the conformation of the KEAP1-NRF2 binding interface, which impairs NRF2 ubiquitination [3,4,13,[24], [25], [26], [27], [28], [29], [30], [31], [32]]. As a result, newly translated NRF2 enters the nucleus, heterodimerizes with small Maf proteins (sMaf), and activates a transcriptional program essential for the environmental stress response [33]. Because NRF2 activation depends on the continuous synthesis of new protein molecules [25,34], translational control represents an important yet comparatively understudied layer of NRF2 regulation. Given the growing clinical interest in pharmacological NRF2 activation, understanding all regulatory mechanisms governing NRF2 abundance is of increasing importance.

The NFE2L2 gene is transcribed from two promoters that generate distinct transcript isoforms, of which the canonical NRF2 protein is encoded by the P1-derived transcripts. Although transcripts originating from both promoters are expressed under basal and stress conditions in primary lymphoblastoid cells [35], the functional significance of alternative promoter usage and the existence of a protein product encoded by the P2-derived transcripts have remained unexplored.

Here, we identify and characterize the promoter 2-derived NRF2 isoform 2. Although its mRNA abundance is comparable to that of canonical isoform 1, isoform 2 exhibits reduced translation under basal conditions and enhanced translation upon electrophilic stress. Collectively, our findings establish translational regulation as an important mechanism shaping NRF2 abundance and stress responses.

2. Results

To understand the transcriptional control of NRF2, we interrogated the NFE2L2 gene locus. Transcripts generated from two NFE2L2 promoters differ exclusively in their first exons, which are entirely distinct in sequence, while exons 2–5 are shared (Fig. 1). Transcripts produced from the first promoter (P1) initiate translation at an AUG codon within exon 1 (hg38 chr2:177,264,576), whereas P2-derived transcripts initiate translation at the beginning of exon 2 (hg38 chr2:177,234,269; transcript variants: 2, 3, 4, 5) or in the middle of exon 2 (hg38 chr2:177,234,087; transcript variant 8), due to the absence of an AUG codon in the exon 1’ (to differentiate between P1- and P2-derived first exons, we will use the exon 1’ name for the P2-derived transcripts). Consequently, the P2-derived NRF2 isoforms are predicted to be at least 16 amino acids (aa) shorter at the N-terminus than the canonical 605 amino acid isoform 1.

Fig. 1.

NFE2L2 is transcribed from two promoters.

(A) Left: Schematic representation of the human NFE2L2 gene and all annotated transcript variants originating from promoter 1 (P1; transcript names shown in blue) and promoter 2 (P2; transcript names shown in green), according to the NCBI RefSeq database. Right: Enlarged view of the promoter-specific first exons, P1-derived exon 1 (Ex1) and P2-derived exon 1′ (Ex1’), together with RNA-seq read coverage across these regions visualized using data from the UCSC Genome Browser.

(B) Evolutionary conservation across the NFE2L2 locus. PhyloP conservation scores from the 100-way vertebrate genome alignment were used to assess nucleotide-level conservation. Ex, exon; Ex1, P1-derived exon 1; Ex1’, P2-derived exon 1′. Data were obtained from the UCSC Genome Browser.

The first exons of the P1- and P2-derived NFE2L2 transcripts are the sole regions distinguishing between these two promoter-derived transcript classes. However, the P2-derived exon 1′ (Ex1’) consists exclusively of a long 5′ untranslated region (5′UTR), a transcript feature that is often represented by lower read coverage in RNA-sequencing (RNA-seq) datasets [36,37]. Consistent with this observation, transcriptomic datasets across human tissues show substantially fewer reads mapping to the P2-derived exon 1′ than to the P1-derived exon 1 (Fig. 1A, right panel). Moreover, even within the P1-derived exon 1, sequencing reads predominantly map to the coding sequence (CDS) rather than to the 5′UTR region (Fig. 2D). Interestingly, analysis of vertebrate sequence conservation using phyloP scores revealed that the non-coding exon 1′ is more highly conserved than exon 1 (Fig. 1B).

Fig. 2.

P1- and P2-derived NFE2L2 transcripts are expressed in human lung cells

(A) Schematic representation of the human NFE2L2 gene and the location of primers used for transcript detection. Arrows above the gene indicate primer pairs used for RT-PCR, whereas dashed arrows below indicate primers used for RT-qPCR. Blue arrows mark primers specific for P1-derived transcripts, green arrows indicate primers specific for P2-derived transcripts, and black arrows denote primers detecting all NFE2L2 transcripts. Ex, exon; P1, promoter 1; P2, promoter 2.

(B) RT-PCR analysis of NFE2L2 transcript expression in lung-derived cell models. NSCLC, primary non-small cell lung cancer cells; NLF, normal lung fibroblasts.

(C) Relative contribution of P1- and P2-derived NFE2L2 transcripts. P1- and P2-derived NFE2L2 transcripts were quantified by promoter-specific RT-qPCR using Tr167 and Tr238, respectively, and normalized to total NFE2L2 transcripts using the Nrf2 All assay. Transcript proportions were calculated using the correction-factor approach described by Vidal-Petiot et al.[38] with amplification-efficiency correction. Bars represent mean ± SD of three independent biological replicates (n = 3); individual dots represent values calculated independently for each biological replicate.

(D) Quantification of P1- and P2-derived NFE2L2 transcripts from RNA-seq data using RSEM and featureCounts. RSEM TPM and percentage values represent the median of three H1299 biological replicates (SRR24121380, SRR24121381, SRR24121382). RSEM percentages were calculated relative to the total TPM of all eight annotated NFE2L2 transcripts. FeatureCounts values were determined for the promoter-specific first exons and normalized for exon length and sequencing depth (RPKM). Percentages represent the contribution of each discriminating exon relative to the combined RPKM of all five NFE2L2 exons.

(E) RNA-seq read coverage across the promoter-specific first exons of NFE2L2 in H1299 cells. Sequencing reads from H1299 cells (SRA accession SRR24121380) were aligned to the human genome (hg38) and visualized using JBrowse.

To determine which NFE2L2 transcripts are expressed in our cellular models, the isoform-specific reverse-transcriptase PCR (RT-PCR) was performed. We used two primer pairs designed to amplify NFE2L2 transcripts originating from both promoters (Mat. & Met. Table 1, Fig. 2A), and resolved the PCR products by electrophoresis in 2% agarose gels to identify transcript variants based on amplicon size (Fig. 2B). Multiple human lung cell types have been analysed, including lung cancer cell lines (A549, RERF, H1299), patient-derived primary non-small cell lung cancer (NSCLC) cells, and normal lung fibroblasts (NLF). This analysis revealed that P1-derived transcript 1 (NM_006164.5) and P2-derived transcript 2 (NM_001145412.3) are the major NFE2L2 transcripts expressed in human lung cells (Fig. 2B).

Table 1.

Primer sequences used for RT-PCR.

NFE2L2 transcript variants	Primer sequences	Product length
NFE2L2 P1 Transcripts 1,6,7	For: TCATGATGGACTTGGAGCTG Rev GCAATGAAGACTGGGCTCTC	Tr 1: 475 bp Tr 6: 385 bp Tr 7: 256 bp
NFE2L2 P2 Transcripts 2,3,4,5,8	For: TCCTGCTTTATAGCGTGCAA Rev: GCAATGAAGACTGGGCTCTC	Tr 2: 602 bp Tr 3: 581 bp Tr 4: 476 bp Tr 5: 568 bp Tr 8: 531 bp

Next, we quantitatively assessed the relative contribution of P1- and P2-derived NFE2L2 transcripts. Using promoter-specific primer sets (Materials and Methods, Table 2; Fig. 2A and Fig. S2A), we performed reverse transcription quantitative real-time PCR (RT-qPCR) and determined the relative proportions of P1- and P2-derived transcripts (Fig. 2C and Fig. S2B). Both transcript populations were readily detected across all examined cell lines, with P1-derived transcripts generally predominating. P2-derived transcripts nevertheless constituted a substantial fraction of the total NFE2L2 transcript pool, ranging from approximately 13% to 47% across the examined cell lines (Fig. 2C).

Table 2.

Primer sequences used for RT-qPCR.

Gene	Primer sequences	Product length
NFE2L2 P1 (1,6,7)	For: AACACACGGTCCACAGCTC	102 bp
	Rev: TCTTGCCTCCAAAGTATGTCAA
NFE2L2 P2 (2,3,8)	For: GACGGGATATTCTCTTCTGTGC	128 bp
	Rev: CATACTCTTTCCGTCGCTGA
NFE2L2 All (1,2,3,4,5,6,7,8)	For: GAGAGCCCAGTCTTCATTGC	104 bp
	Rev: TGCTCAATGTCCTGTTGCAT
HMOX1	For: AAGACTGCGTTCCTGCTCAA	136 bp
	Rev: GGGCAGAATCTTGCACTTTGT
ACTB	For: TCCCTGGAGAAGAGCTACG	131 bp
	Rev: GTAGTTTCGTGGATGCCACA

We next compared the relative abundance of P1- and P2-derived NFE2L2 transcripts in H1299 cells estimated by promoter-specific RT-qPCR and RNA-sequencing approaches. For RNA-seq, we used two complementary approaches: transcript-level quantification with RSEM, which employs an expectation-maximization algorithm to probabilistically assign reads among transcript variants [39], and exon-level read counting with featureCounts (v2.0.8) [40,41]. targeting the transcript-discriminating exon 1. As summarized in Fig. 2D, RSEM estimated P2-derived transcripts to account for approximately 4.7% of the P1/P2 transcript pool, whereas exon-level quantification using featureCounts estimated P2-derived transcripts at approximately 7.3%. In contrast, promoter-specific RT-qPCR estimated P2-derived transcripts to account for approximately 28% of the total NFE2L2 transcript pool in H1299 cells. This discrepancy is particularly relevant in the context of NFE2L2 transcript architecture, in which the extended P2-specific 5′ UTR is poorly represented in the RNA-seq data (Fig. 2E).

Having demonstrated the presence of P1- and P2-derived NFE2L2 transcripts, we next sought to investigate the existence of NRF2 isoform 2, the predicted protein product of the P2-derived transcripts. This isoform has likely escaped detection because the predicted molecular weight difference between isoform 1 (605 aa) and isoform 2 is only ∼1.77 kDa, which is below the resolving capacity of conventional SDS-PAGE. We selected A549 cells as a model system because P1- and P2-derived transcripts are expressed at comparable levels in these cells (Fig. 2C), maximizing the likelihood of detecting the P2-derived protein isoform. Moreover, A549 cells have higher steady-state levels of NRF2 due to KEAP1 mutation in the NRF2 binding site [[42], [43], [44]]. To distinguish between the two protein isoforms, we designed an siRNA targeting exon 1 of the P1-derived transcripts (siNRF2-1), thereby selectively depleting P1 transcripts while leaving P2-derived transcripts unaffected (Fig. 3A and B). To facilitate detection of NRF2 protein species, cells were treated with the proteasome inhibitor MG-132 for 4 h prior to harvest. A commercially available siRNA pool targeting all NFE2L2 transcripts served as a knockdown efficiency control. NRF2 knockdown was evaluated by RT-qPCR and western blotting (Fig. 3B–D). mRNA and protein levels of the NRF2 target gene heme oxygenase-1 (HMOX1, HO-1) were monitored in parallel (Fig. 3C and D).

Fig. 3.

NRF2 isoform 2 is regulated by KEAP1

(A) Experimental workflow for identification of NRF2 isoform 2 in A549 cells. Cells were transfected with either a commercial siNRF2 pool targeting all NFE2L2 transcripts or with siNRF2-1, designed to selectively target P1-derived transcripts. Forty-four hours after transfection, cells were treated with the proteasome inhibitor MG-132 for 4 h to stabilize NRF2 protein.

(B) RT-qPCR analysis of P1- and P2-derived NFE2L2 transcript levels in A549 cells following transfection with siNRF2 pool or siNRF2-1, with and without MG-132 treatment, as described in (A).

(C) RT-qPCR analysis of HMOX1 expression in A549 cells following transfection with siNRF2 pool or siNRF2-1, with and without MG-132 treatment, as described in (A).

(D) Western blot analysis of NRF2 and HO-1 protein levels in A549 cells following NFE2L2 knockdown, as described in (A). The residual NRF2 species consistent with isoform 2 is indicated by a red asterisk.

(E) H1299 cells were transfected with pcDNA3.1 constructs encoding human NRF2 isoform 1 (NM_006164.5) or isoform 2 (NM_001145412.3). Following treatment with emetine for the indicated time points, NRF2 protein levels were analysed by western blotting to compare the stability of both isoforms.

(F) Western blot analysis of NRF2 and NQO1 protein levels following knockdown of KEAP1 and P1-derived NFE2L2 transcripts in H1299 and A549 cells.

Statistical significance was assessed using a two-tailed Student's t-test assuming equal variances (B, C). *p < 0.05, **p < 0.01, and ***p < 0.001. Bars represent mean ± SD from three independent biological experiments (n = 3).

The siNRF2 pool efficiently reduced transcripts from both promoters (>99%), whereas siNRF2-1 selectively depleted P1-derived transcripts (>99%) while leaving P2-derived transcript levels intact (Fig. 3B). Consequently, the residual NRF2 protein detected following selective depletion of P1-derived transcripts is expected to originate from P2-derived transcripts and therefore represent NRF2 isoform 2 (Fig. 3D). This putative isoform 2 protein was detected at substantially lower levels than the canonical isoform 1 under both basal conditions and following MG-132 treatment. HMOX1 mRNA was strongly induced by MG-132 (∼80-fold) (Fig. 3C). Complete NRF2 knockdown reduced HMOX1 expression eightfold, whereas selective depletion of the P1-derived transcripts reduced HMOX1 mRNA levels threefold. These results suggest that both NRF2 isoforms contribute to HMOX1 transcriptional regulation, an effect that was also evident at the protein level (Fig. 3D).

To further validate the contribution of the P2-derived transcripts, we performed the reciprocal experiment using siRNAs targeting exon 1′. However, despite testing three independent siRNAs, the knockdown efficiency was moderate, reaching only ∼50% (Fig. S3). Although a reduction in total NRF2 protein was observed, the limited silencing efficiency precluded firm conclusions regarding the quantitative contribution of the P2-derived transcripts to total NRF2 abundance. The reduced silencing efficiency may reflect structural features of the targeted 5′UTR region and/or occupancy by RNA-binding proteins and translation-initiation complexes, which can limit RISC accessibility to target transcripts and reduce silencing efficiency [45,46].

To determine why the steady-state protein levels of the two NRF2 isoforms differ despite comparable transcript abundance, we first assessed their stability in cells (Fig. 3E). Isoform 1 and isoform 2 were ectopically expressed in H1299 cells and, following translation inhibition with emetine, were chased through five timepoints over 90 min. Both isoforms displayed similar turnover kinetics, reaching ∼20% of their initial abundance after 90 min of emetine treatment. KEAP1 is the main regulator of isoform 1 degradation and it binds two motifs within NRF2 N-terminus [17,30,32,47]. Although isoform 2 lacks the N-terminal region present in isoform 1, it retains the DLG and ETGE motifs required for KEAP1 binding, thus we next checked if KEAP1 is also responsible for degrading isoform 2. We knocked down KEAP1 in H1299 and A549 cell lines, together with isoform 1 (Fig. 3F). The remaining isoform 2 accumulated when KEAP1 levels were reduced, indicating that its degradation is mediated by KEAP1, similarly to isoform 1.

Because neither transcript abundance nor protein stability could explain the markedly lower abundance of isoform 2, we hypothesized that the two transcripts differ in translational efficiency. Translation initiation constitutes the major regulatory step controlling protein synthesis, as the scanning 40S ribosomal subunit interacts with cis-regulatory elements located within the 5′UTR [48,49]. Given that the two NRF2 isoforms are expressed from transcripts containing distinct 5′UTRs (Fig. 1), we hypothesized that their translation efficiencies differ.

To test this, we performed polysome profiling in A549 cells followed by NRF2 isoform-specific RT-qPCR in polysome fractions, under basal conditions and upon treatment with an electrophilic NRF2 activator, RTA-408 (omaveloxolone) (Fig. 4B and C). We first established the kinetics of NRF2 accumulation following RTA-408 treatment by analysing cells after 15 min and 1, 2, 3 and 4 h of exposure (Fig. 4A). 2 h treatment was chosen for the translation studies. Unexpectedly, 2 h treatment with 200 nM RTA-408 altered the global polysome profile, increasing the monosomal fraction and reducing polysome abundance, consistent with partial inhibition of global translation (Fig. 4B and Fig. S4B).

Fig. 4.

Promoter usage determines distinct translational properties of NFE2L2 transcripts.

(A) A549 cells were treated with 200 nM RTA-408 for the indicated timepoints and NRF2 and HO-1 protein levels were analysed by western blotting.

(B) Representative overlaid polysome profiles of DMSO-treated (grey) and RTA-408-treated (red) A549 cells.

(C) Distribution of P1-derived (left) and P2-derived (right) NFE2L2 transcripts across ribosomal fractions following polysome profiling of A549 cells. Grey lines indicate DMSO-treated cells and red lines indicate cells treated with 200 nM RTA-408 for 2 h. Statistical significance between DMSO- and RTA-408-treated cells was evaluated using a one-tailed homoscedastic Student's t-test for subpolysomal (fractions 1–3), light polysomal (fractions 4–7), and heavy polysomal (fractions 8–11) regions. Shaded areas represent mean ± SD calculated from three independent biological experiments (n = 3).

(D) Predicted secondary structures of the alternative first exons of the P1-derived (left) and P2-derived (right) NFE2L2 transcripts generated using RNAfold. Nucleotide size corresponds to SHAPE-MaP reactivity, with larger circles indicating higher reactivity. Evolutionary conservation, calculated using UCSC PhyloP, is colour-coded: red indicates higher conservation and slower evolution, whereas blue indicates lower conservation and faster evolution. The grey-highlighted region within the P2-derived exon 1′ marks a highly structured and evolutionarily conserved RNA element that may contribute to the translational regulation of the P2-derived transcripts.

Under basal conditions, P1-derived transcripts were almost exclusively associated with polysomes (Fig. 4C, left panel, grey), consistent with efficient constitutive translation. In contrast, P2-derived transcripts, which contain an alternative 729-nt 5′UTR, were distributed between monosomal and polysomal fractions, indicating that a substantial proportion is translated inefficiently or remains translationally inactive (Fig. 4C, right panel, grey). A similar distribution of P1- and P2-derived NFE2L2 transcripts was observed in HEK293T cells following polysome profiling (Supplementary Fig. 4A). This differential partitioning suggests that translation of the P2-derived transcripts is subject to additional regulatory control.

RTA-408 treatment promoted translation of both P1- and P2-derived NFE2L2 transcripts (Fig. 4C–red). P1-derived transcripts shifted towards heavier polysomal fractions, whereas the translational response of the P2-derived transcripts was even more pronounced, resulting in their nearly complete association with polysomes. Together with the comparable abundance of P1- and P2-derived transcripts (Fig. 2C) and the similar stability and KEAP1-dependent degradation of the two protein isoforms (Fig. 3E and F), these findings indicate that differences in translational efficiency contribute to the lower abundance of isoform 2 under basal conditions. More broadly, they establish translational control as an important mechanism regulating NRF2 abundance in cells.

To explore the structural basis for the translational differences between the two NRF2 promoter-derived transcripts, we modelled the secondary structures of their alternative first-exon 5′UTRs (Fig. 4D). The 729-nt exon 1′ unique to the P2-derived transcripts contains a highly conserved region with elevated base-pairing probability, indicative of increased structural stability. These features suggest the presence of a cis-regulatory RNA element that may contribute to the differential polysome association and translational regulation of the P2-derived transcripts.

3. Discussion

Our study demonstrates that NFE2L2 is expressed as two promoter-derived isoforms with distinct translational properties. The canonical 605 aa isoform 1, generated from the upstream promoter (P1), is efficiently transcribed and translated in human cells. In contrast, the 589 aa isoform 2, originating from the alternative promoter (P2), exhibits markedly reduced translation under basal conditions. Consequently, NRF2 isoform 2 is substantially less abundant at the protein level than isoform 1. The electrophile inhibitor RTA-408 induced translation of transcripts from both promoters, although P1-derived transcripts remained more efficiently translated under stress conditions. Although both isoforms are subject to KEAP1-dependent degradation under basal conditions, this work shows that translational regulation of NFE2L2 transcripts constitutes an additional major mechanism shaping NRF2 activation dynamics in cells.

So far, NRF2 isoform 2 has remained largely unexplored, likely because the predicted molecular weight difference between isoform 1 and isoform 2 is only ∼1.77 kDa, preventing their reliable separation by conventional Tris-glycine SDS-PAGE. Even when using Bis-Tris gels and electrophoresis under near-neutral pH conditions, as applied in this study, the difference in migration between the two isoforms is minimal. This technical limitation, together with the relatively limited characterization of P2-derived NFE2L2 transcripts, may have contributed to the lack of studies addressing isoform 2. Our identification of isoform 2 protein therefore provides an important basis for investigating its regulation and functional relationship with the canonical NRF2 isoform.

By employing an isoform-specific siRNA that selectively depleted P1-derived transcripts, we identified an endogenous NRF2 protein species consistent with isoform 2 (Fig. 3D–F). Notably, isoform 2 protein was present at substantially lower levels than canonical NRF2, despite both P1- and P2-derived transcripts being readily detected, suggesting that differences in translation or degradation may contribute to the differential protein abundance. Since both isoforms displayed similar stability and accumulated following KEAP1 knockdown, consistent with shared KEAP1-dependent regulation [14,47], we next examined their translational efficiency using polysome profiling combined with isoform-specific RT-qPCR in A549 and HEK293T cells (Fig. 4, Fig. S4). This analysis suggested that differential translational efficiency contributes to the lower abundance of isoform 2. Under basal conditions, P1-derived transcripts were predominantly associated with polysomes, whereas P2-derived transcripts partitioned between polysomal and subpolysomal fractions, consistent with less efficient translational engagement (Fig. 4C, grey).

To determine how electrophilic stress affects translation of both isoforms, we treated cells with the synthetic oleanolic acid derivative RTA-408 (omaveloxolone), an NRF2 activator approved for the treatment of Friedreich's ataxia and currently being evaluated for additional therapeutic indications [[50], [51], [52], [53], [54], [55]]. Following 2 h treatment with 200 nM RTA-408, P2-derived transcripts shifted almost entirely to polysomal fractions, whereas P1-derived transcripts redistributed toward heavier polysomes compared to basal conditions (Fig. 4C–red). These findings indicate that translation of NFE2L2 transcripts originating from both promoters is dynamically regulated. However, for P2-derived transcripts, translational activation appears to be a major determinant of isoform 2 abundance in cells. Consistent with this observation, we identified within the P2-derived 5′UTR a conserved vertebrate RNA element predicted to form a stable secondary structure that could hinder ribosome loading or scanning under basal conditions and become functionally relieved during electrophilic stress (Fig. 4D). The mechanistic role of the P2 5′UTR in regulating translation remains to be experimentally defined.

4. Materials and methods

4.1. Cell culture

Human non-small cell lung cancer (NSCLC) cell line H1299 was purchased from Cytion (Eppelheim, Germany), A549, RERF and normal lung fibroblasts (NLF) were purchased from RIKEN BRC Cell Bank (Tsukuba, Japan). Cells were cultured in Dulbecco's modified Eagle's medium (Gibco, Thermo Fisher Scientific), supplemented with 8% Fetal Bovine Serum (Gibco, Thermo Fisher Scientific) and 1% Penicillin-Streptomycin (10,000 U/mL, Gibco, Thermo Fisher Scientific) at 37°C under humidified conditions with 5% CO2.

4.2. Drug treatments

Cycloheximide (CHX, Sigma-Aldrich) was dissolved in 100% ethanol to 10 mg/ml and diluted to 150 μg/ml in media for treatment.

Emetine dihydrochloride (Merck Millipore) was dissolved in dH2O to 100 mM and diluted to 20 μM in media for treatment.

MG-132 (Sigma-Aldrich) was dissolved in 100% DMSO to 50 mM stock and diluted to 30 μM in media for treatment.

RTA-408 (omaveloxolone, MedChemExpress) was dissolved in 100% DMSO to 100 mM stock. For treatment, 1 mM RTA-408 in 100% DMSO was freshly prepared and diluted in media to 200 nM.

4.3. Isolation of NSCLC cells and depletion of CD45⁺ cells

Cancer samples (0.25-1 g) were obtained from patients diagnosed with NSCLC. Patients were admitted to the Clinic of Thoracic Surgery of Medical University of Gdansk (MUG) University Clinical Centre for surgical resection of the primary tumor. The patients had not been previously treated with any anti-cancer therapy and no metastases were detected. Samples were cut into 1-2 mm fragments and washed 3 times with PBS to remove contaminating debris and erythrocytes. Subsequently, digestion solution (DS) in 1:1 of DS (ml): tissue mass (g) ratio was added. Tissue was gently agitated at 37°C until single-cell suspension was obtained. DS contained 5% collagenase type I (Sigma-Aldrich; 5 mg/ml), 19% of PBS and 1% of Fetal Bovine Serum (FBS). The collagenase was inactivated with an equal volume of 10% FBS-supplemented low glucose-DMEM medium (Gibco, Thermo Fisher Scientific), followed by filtration of the resulting cell suspension through a 100 μm nylon cell strainer. Subsequently, cells were washed with PBS and the erythrocyte lysis buffer was used to remove erythrocytes (10 min, RT). After centrifugation (600g, 10 min) cells were washed with 4% FBS-supplemented PBS and subjected to tumour-infiltrating leukocyte depletion. For this purpose, positive immunomagnetic selection of CD45⁺ cells was performed with EasySep Human CD45 Depletion Kit II (StemCell Technologies) according to the manufacturer instructions (post-isolation purity 96-99%). Thus, untouched NSCLC and lung-derived cells were isolated. Cells were plated into 75 cm2 culture flasks in TumorPlus263 medium without serum. Establishment of the primary cell line was confirmed with flow cytometry (FAP-CK19+ phenotype; FACS Aria Fusion, BD Biosciences) and by the continuous proliferation of cultured cells for 2 months in vitro. Cells that fulfilled both criteria were used for the experiments.

4.4. RNA sequencing analysis

Raw paired-end RNA-seq reads from H1299 cells taken from SRA project PRJNA954031 from three samples (SRR24121380, SRR24121381, SRR24121382; dataset SRP431844) were quality-trimmed using fastp (v0.23) [56] and aligned to the GRCh38.p14 human reference genome (GCF_000001405.40) using STAR (v2.7.11b) [57] within a Snakemake pipeline (v8) [58].

Transcript isoform expression was quantified using RSEM (v1.2.28) [39]. RSEM employs an expectation-maximization algorithm to probabilistically resolve reads mapping to multiple isoforms, which is particularly relevant for NFE2L2 transcripts that share exons 2–5. Expression values are reported as TPM.

Exon-level read counts were obtained using featureCounts (v2.0.8) [40]. Counts for the discriminating exon 1 of NM_006164.5 (chr2:177,264,532–177,264,727; 196 bp) and NM_001145412.3 (chr2:177,263,403–177,264,131; 729 bp) were normalized to RPKM to account for differences in exon length. The percentage of exon 1 RPKM was calculated relative to the sum of RPKM values across all five exons of each transcript.

4.5. Evaluation of transcripts expression by RT-qPCR

Total RNA was isolated lines using the RNeasy Mini Kit (Qiagen) according to the manufacturer's instructions. RNA concentration and purity were assessed using NanoDrop. Two micrograms of RNA were transcribed (Applied Biosystems High-Capacity cDNA Reverse Transcription Kit). For each qPCR reaction 25 ng of cDNA was mixed with 0.5 μM primers (listed in Table 2) and master mix (FastStart Essential DNA Green Master, Roche). qPCR reaction was performed as follows: 95°C for 10 min, (95°C for 20s, 60°C for 20s, 72°C for 20s) × 45 cycles on Rotor-Gene Q (Qiagen) with melt curve analysis. Gene expression (HMOX1, NFE2L2) was normalized to ACTIN with comparative Ct method. Statistical significance was assessed using a two-tailed Student's t-test assuming equal variances (B, C). p < 0.05, **p < 0.01, and ***p < 0.001. Bars represent mean ± SD from three independent biological experiments (n = 3).

4.6. Identification of NFE2L2 transcript variants with RT-PCR

Total RNA was isolated from six lung cell lines using the RNeasy Mini Kit (Qiagen) according to the manufacturer's instructions. RNA concentration and purity were assessed using NanoDrop. One microgram of total RNA was reverse transcribed using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems), and the resulting cDNA was diluted 1:2. One microliter of diluted cDNA was used as template for each 20-μL PCR reaction containing 0.02 U HotStarTaq DNA polymerase (Qiagen), 1× PCR buffer, 0.2 mM dNTP mix, and 200 nM of each primer. Primers were designed to enable length-based discrimination of NFE2L2 transcript variants expressed in the examined cells (Table 1). PCR was performed with an initial activation step at 95°C for 15 min, followed by 45 cycles of 95°C for 20 s, 62°C for 20 s, and 72°C for 60 s, and a final extension at 72°C for 3 min. PCR products were separated by electrophoresis on a 2% agarose gel and analysed. The identity of the PCR products was confirmed by Sanger sequencing.

4.7. Quantification of P1- and P2-derived NFE2L2 transcripts with RT-qPCR

The relative contribution of NFE2L2 transcripts originating from the P1 and P2 promoters was determined by RT-qPCR using promoter-specific primer pairs and a primer pair targeting a sequence common to all NFE2L2 transcripts (primer sequences are listed in Table 2). Transcript proportions were calculated according to the method of Vidal-Petiot et al. [38] using assay-specific correction factors. An additional efficiency-corrected analysis was performed using standard curves generated from the same serially diluted cDNA stock for all three assays. P1 and P2 transcript proportions were calculated for each biological replicate and expressed as percentages of total NFE2L2 transcripts. Data are presented as mean ± SD of three biological replicates.

4.8. siRNA-mediated gene knockdown

Reverse transfection was performed in 12-well plates. 100,000 (H1299) or 130,000 (A549) cells/well were mixed with 20pmoles of each siRNA: siNRF2, siKEAP1, non-targeting control (all from ON-TARGETplus™ SMART pool, Dharmacon™) or with siRNA targeting P1-derived NFE2L2 3’[(CAC ACG GUC CAC AGC UCA U)TT]5′, alongside 3 μl/well of Lipofectamine 3000 reagent (Invitrogen, Thermo Fisher Scientific), according to manufacturer's instructions. 48h after transfection, cells were collected for analysis of protein and mRNA expression.

4.9. Transient transfection with plasmids and emetine chase

pcDNA3.1(+)-based vectors encoding N-terminally HA-tagged human NRF2 isoform 1 or isoform 2 were purchased from GenScript. An empty pcDNA3.1(+) vector was used as a transfection control. Reverse transfection was performed in 12-well plates. H1299 cells (150,000 cells per well) were transfected with 1 μg plasmid DNA using 2 μL Lipofectamine 3000 and 1.7 μL P3000 reagent (Invitrogen, Thermo Fisher Scientific), according to the manufacturer's instructions. After 48 h, cells were treated with 20 μM emetine for 15, 30, 60, or 90 min, as indicated. Cells were subsequently collected and analysed by Western blotting.

4.10. Western blotting

Total protein was acquired by lysing cells in RIPA buffer. 10μg of proteins were electrophoretically separated in 4-12% gradient NuPAGE gels (Thermo Fisher Scientific) in MOPS buffer and transferred to the nitrocellulose blotting membrane (Amersham Protran®) via wet transfer. For blocking, 5% non-fat milk in Tris-buffered saline was applied at room temperature for half an hour. Membranes were incubated overnight in 1:500 dilution of antibodies: anti-NRF2 (E5F1A, 20733; Cell Signaling Technology), anti-HMOX-1 (MA1-112, Thermo Fisher Scientific), anti-KEAP1 (OTI1B4, AB119403, Abcam), anti-NQO1 (A180, NB200-209, Novus Biologicals) and anti-β-actin (A2228, Sigma-Aldrich) at 4°C. Subsequently, membranes were washed three times in TBST followed by incubation for 1 h with HRP-labeled goat anti-rabbit/mouse IgG (Jackson ImmunoResearch) at a 1:3000 dilution and washed in TBST again. Bands were visualized using chemiluminescent substrate (Clarity Max™ Western ECL Substrate, BIO-RAD).

4.11. Polysome fractionation

Polysome fractionation was performed in steady-state HEK293T and A549 cells and upon 2 h treatment with 200 nM RTA-408 to assess the translational status of NFE2L2 isoforms. Three biological repeats of control and treated cells were incubated for 10 min at 37°C with 10 μg/mL cycloheximide (CHX) to arrest ribosomes, and rapidly chilled on ice. After washing with ice-cold PBS containing CHX, cells were lysed in polysome lysis buffer (10 mM HEPES, 62.5 mM KCl, 5 mM MgCl₂, 1% Triton X-100, 400 U/mL RiboLock RNase inhibitor, 100 μg/mL CHX, 10 mM DTT) and incubated on ice for 40 min with intermittent vortexing. Lysates were clarified by centrifugation (15,000 rpm, 10 min, 4°C), and equal amounts of cleared lysate were layered onto 10%–60% linear sucrose gradients prepared in 25 mM Tris-HCl (pH 7.4), 25 mM NaCl, 5 mM MgCl₂, and 2 mM DTT. Gradients were centrifuged (SW41 rotor, 2 h 30 min, 37,000 rpm, 4°C) and fractionated on a BioComp Gradient Station. Fractions (250 μL) were immediately mixed with TRIzol, and RNA was isolated using the Direct-zol RNA Miniprep Kit (Zymo Research) with on-column DNase digestion. RNA (100–500 ng) from each fraction was reverse-transcribed, and cDNA was quantified by isoform-specific qPCR to determine the distribution of NFE2L2 transcripts across monosomal and polysomal fractions. Student's t-test (one-tailed, homoscedastic) was performed to evaluate statistical significance of NFE2L2 expression in fractions 1-3 (sub-polysome), 4-7 (light polysome) and 8-11 (heavy polysome) between DMSO control and RTA-408-treated A549 cells.

4.12. RNA structure modeling

RNA secondary structure for the first exon sequences of NRF2 isoforms was predicted using the RNAfold web server (ViennaRNA Package, version X.X) with default parameters. Dot-bracket notation and corresponding nucleotide coordinates were obtained by downloading the JSON output from the forna visualization tool. Per-nucleotide phylogenetic conservation scores and SHAPE-MaP reactivity values were obtained from UCSC genome browser. These values were merged with the RNAfold coordinates for downstream visualization.

RNA structure plots were generated in R (version 4.4.0) using the ggplot2 package. Conservation scores were transformed using 2score2^{\text{score}}2score to emphasize differences and mapped to node size, while SHAPE-MaP reactivity values were encoded as a colour gradient (red for low reactivity, dark blue for high reactivity). Each nucleotide was plotted as a filled circle at its predicted 2D coordinate, with edges representing RNA backbone connectivity. Every 10th nucleotide position was labeled for reference, and the final structures were rendered with a fixed aspect ratio to preserve spatial geometry. Plots were exported as high-resolution PDF files for inclusion in figures.

CRediT authorship contribution statement

Alicja Sznarkowska: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Zuzanna Urban-Wójciuk: Investigation. Alicja Dziadosz-Brzezińska: Investigation. Maciej Cieśla: Writing – review & editing, Writing – original draft, Visualization, Resources, Methodology, Investigation, Funding acquisition. Sara Kusiński: Investigation. Kevork Wakimian: Investigation. Alicja Szabelska-Beręsewicz: Visualization, Software, Methodology, Investigation. Elżbieta Chruściel: Resources. Ines Papak: Resources. Natalia Marek-Trzonkowska: Resources, Methodology, Funding acquisition. Artur Piróg: Formal analysis. Witold Rzyman: Resources.

Ethics declaration

Written informed consent to take part in the study and to publish the article has been obtained from all participants or their legal representatives. The privacy rights of participants have been observed.

This study included organ or tissue donors. This study includes human biological material and consent was obtained by donors, or their next of kin or legal representatives, for use in this study and for publication of the article. The samples used in this research were not sourced from executed prisoners or prisoners of conscience.

This study was performed in compliance with relevant laws, regulatory frameworks and guidelines where the research took place. This study was approved by the Independent Bioethics Committee for Scientific Research at the Medical University of Gdańsk. (Approval No. NKBBN/880-329/2016)

Funding sources

This research was funded by the National Science Centre grant Sonata nr UMO-2021/43/D/NZ1/02059 (ASz). MC is beneficiary of EMBO Installation Grant, Sonata Bis (UMO-2022/46/E/NZ3/00141) and OPUS (UMO-2021/43/B/NZ3/01177) grants from National Science Center in Poland.

ICCVS is funded by the FENG International Research Agenda's Program of the Foundation for Polish Science (MAB/2017/03), and the Priority FENG.02 Innovation-friendly environment, Measure FENG.02.01 International Research Agendas in the frame of the project “Science for Welfare, Innovations, and Forceful Therapies (SWIFT)” (FENG.02.01-IP.05-0031/23).

Declaration of competing interest

None.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Data availability

All data are included in the manuscript