The gene expression profiles of human paired psoriasis tissues were retrieved from Gene Expression Omnibus (GEO) database. After carefully screening the content, discarding the datasets with incomplete information and those lacking control patients, the 5 datasets with paired samples, GDS5392, GDS4602, GDS2518, GDS3539, and GDS4600, were obtained (18-22). R packages (version 3.5.2) were used to annotate the raw data and make the expression matrix (23). The median of expression level was chosen for genes matched by several probes, and paired t test was used to compare UPF1 expression. Student's t-test was used for the analysis of statistical significance between two groups, and Bonferroni test was used for multiple comparisons after the analysis of variance. Data for continuous variables are expressed as the mean ± SD. All statistical analyses were performed using SPSS 21.0 software (IBM Corp.). P<0.05 from a two-tailed test was considered to indicate a statistically significant difference.

A total of 10 patients with a dermatologist-confirmed diagnosis of vulgaris psoriasis, according to the 'Consensus on diagnosis and treatment of Chinese integrative medicine for psoriasis vulgaris', published in China (24,25), were recruited at the Affiliated Hospital of Ningbo University (Ningbo, China) between January 2016 and July 2017. The patient information is shown in Table SI. The lesion scales from 6 patients (P1-P6) were independent samples, while the lesion scales and adjacent healthy samples from 4 patients (PA-PD) were paired samples. All the psoriasis patients had no other autoimmune or systemic diseases, and required a typical lesion of ≥1 cm in size that was suitable for biopsy, and the target lesion and surrounding 5 cm area were not treated with any therapeutic measures for at least 2 weeks before the sampling. Skin scales or healthy cornified epidermal layer of patients with psoriasis vulgaris were scraped by blunt scalpels with patients' permission. The scales were promptly soaked in liquid-free nitrogen RNA sample storage solution (Beijing Bomaide Gene Technology Co., Ltd.).

To create the well-established psoriasis-like skin model (26), 8-11-weeks-old Balb/c mice provided by the Experimental Animal Center of Hangzhou Normal University were kept under standard laboratory conditions of 12-h light-dark cycles, 50% humidity and 24-26°C ambient temperature with free access to food and water. The mice (2 male and 2 female, 20-25 g) received a daily topical dose of 40 mg commercially available IMQ cream (5%; Zhuhai United Laboratories Co., Ltd.) on the shaved back, equivalent of a daily dose of 2.083 mg of the active compound. The control shaved skin nearby the IMQ-treated area (distance, 1.5-2 cm) was treated similarly with a control cream (Vaseline Lanette cream; Fagron). During the 7-day experiment, the topical IMQ treatment did not lead pain, and the health of all 4 mice was monitored daily. On day 7, all 4 mice were sacrificed by rapid cervical vertebra dislocation, and the skin specimens of mice were collected once the cervical tissue separation was ensured.

Total RNA from model mice, clinical tissues or cells was isolated using TRIzol^® reagent (Invitrogen; Thermo Fisher Scientific, Inc.), and first-strand cDNA for RT-qPCR was synthesized using the PrimeScript™ RT reagent kit with gDNA Eraser according to the manufacturer's protocol (Takara Bio, Inc.). qPCR analysis was performed using the QuantStudio™ 7 Flex Real-Time PCR system (Thermo Fisher Scientific, Inc.) using SYBR Premix Ex Taq (Takara Bio, Inc.) with gene-specific primers listed in Table SII. The following thermocycling conditions were used: Initial denaturation at 95°C for 1 min; followed by 40 cycles of 95°C for 15 sec and 60°C for 30 sec. The disassociation stage was added to check the amplicon specificity. For transcript detections in scales tissues, UPF1 was quantified using the allele-specific primers UPF1 insA for c.2935_2936insA mutant, or UPF1 del corresponding to the c.2030-2081del UPF1 mRNA (Table SII). Specificity and efficiency of allele-specific PCR were tested with corresponding UPF1 constructs, and further details are described in Data S1. Each mRNA quantification value represented an average of at least three measurements, and mRNA expression levels were calculated using the 2^−ΔΔCq method (27) after normalization to the endogenous housekeeping gene 18S for human samples or GAPDH for mouse samples.

For RT-PCR, RT was performed using 1 µg of total RNA with the PrimeScript™ 1st Strand cDNA Synthesis kit according to the manufacturer's protocol (Takara Bio, Inc.). PCR amplification was performed with nested primers (Table SIII) overlapping the coding sequence region of the UPF1 mRNA (NCBI Reference Sequence, NM_002911.3). The following thermocycling conditions were used: Initial denaturation at 96°C for 2 min; followed by 35 cycles of 96°C for 10 sec, 55°C for 30 sec and 72°C for 2 min. The sanger sequencing was performed at Tsingke Biological Technology.

For immunohistochemical (IHC), skin tissue from the model mice was fixed in 10% formalin for 24 h at room temperature and embedded in IHC-grade paraffin. Formalin-fixed paraffin-embedded sections were cut into 4-6-µm sections with a microtome and deparaffinized two times in xylene, followed by serial dilutions of ethanol. After heat-induced antigen retrieval using the antigen unmasking solution (Vector Laboratories, Inc.; Maravai LifeSciences) for 30 min at 95°C, the internal peroxidase activity was quenched by incubation with 3% hydroperoxide in methanol for 15 min at room temperature, then sections were incubated in 3% bovine serum albumin (Beyotime Institute of Biotechnology) for 1 h at room temperature. The prepared sections were incubated overnight at 4°C with anti-UPF1 primary antibodies (cat. no. sc-390096; 1:100; Santa Cruz Biotechnology, Inc.). The secondary antibody (cat. no. GAMPO; 1:1; DAKO; Agilent Technologies, Inc.) was used according to the manufacturer's instructions for 50 min at room temperature. The intensity of IHC staining was measured using ImageJ software (version 1.52a; Media Cybernetics, Inc.).

Protein extraction for western blotting was performed using RIPA lysis buffer (Beyotime Institute of Biotechnology) with phenylmethylsulfonyl fluoride serine protease inhibitor (Beyotime Institute of Biotechnology). Protein extracts were quantified using a BCA Protein Assay kit (Beyotime Institute of Biotechnology) according to the manufacturer's instructions. Total proteins (30 µg) were resolved by SDS-PAGE (8% gel) and visualized using Odyssey^® imaging system with IRDye secondary antibodies (cat. no. 926-32210; 1:1,000; LI-COR Biosciences) for 15 min at room temperature. The following primary antibodies were used overnight at 4°C: Anti-UPF1 (cat. no. sc-166092; 1:200; Santa Cruz Biotechnology, Inc.), anti-AREG (cat. no. GTX100986; 1:200; GeneTex, Inc.), and anti-GAPDH (cat. no. M20006L; 1:1,000; Abmart Pharmaceutical Technology Co., Ltd.). Semi-quantification was performed using ImageJ software.

Primary cultured normal human keratinocytes (HEKα), immortalized nontumorigenic human keratinocyte-derived cell line (HaCaT) and 293T cell line were purchased from the Shanghai Institute of Biochemistry and Cell Biology. HaCaT and 293T cells were cultured in DMEM with 10% FBS (Biological Industries), and HEKα cells were cultured in EpiLife Medium with 60 µM calcium with Keratinocyte Growth Supplement (Gibco; Thermo Fisher Scientific, Inc.). All cells were incubated at 37°C in a humidified 5% CO₂ incubator.

For plasmid and small interfering (si)RNA transfection, 4×10⁴ cells/well in 6-well plates were cultured overnight and then transfected using the GeneTran Reagent (Biomiga, Inc.). For co-transfection targeting both UPF1 and AREG, constructs were transfected at the same time; however, when constructs targeting the same gene were co-transfected, construct for knockdown was transfected first, then the overexpression construct was transfected 24 h later. Total mRNA and protein of cells were collected 48 h after transfection.

A short hairpin RNA (shRNA) sequence specifically targeting UPF1 (Table SIV) was cloned into Phblv-U6-puro (provided by Professor Fan Handong, Hangzhou Normal University) by BamHI and EcoRI digestion. The siRNA targeting AREG (targeted sequence, CCACAAATACCTGGCTATA) and a scrambled sequence were purchased from Guangzhou RiboBio Co., Ltd. The UPF1 insA and UPF1 del plasmids were constructed based on the pCMV-MYC-UPF1 vector (provided by Professor Lynne E. Maquat, University of Rochester). Primers for site-directed mutagenesis (Table SIV) were designed by QuickChange Primer Design (Agilent Technologies, Inc.) and applied with the KOD-Plus-PCR enzyme (Toyobo Life Science). Residual templates were digested by DpnI (New England Biolabs, Inc.) at 37°C for 5 h. The AREG gene open reading frame (ORF) with or without the 3′ untranslated region (3′UTR), which was referred as AREG-3′UTR-pEGFP or AREG-ORF-pEGFP, respectively, was cloned into the pEGFP-N1 vector (provided by Dr Wang Miao, Hangzhou Normal University) using the HindIII and BamHI restriction sites. The primer sequences used are listed in Table SIV. The AREG 3′UTR sequence was amplified and cloned into the dual-luciferase reporter construct pEZX-FR02 (GeneCopoeia, Inc.) by double digestion with EcoRI and SpeI (the primer sequences used are listed in Table SIV) to generate the pEZX-AREG-3′UTR construct.

The 293T cells were transfected with or without shUPF1 and the transcription in cells was inhibited by incubation with 20 mg/ml 5,6-dichloro-1-β-D-ribofuranosylbenzimidazole (DRB; Enzo Life Sciences, Inc.) dissolved in DMSO. Total RNA was extracted at 0, 3, 5, 7, and 10 h following the inhibition of transcription, and AREG mRNA levels were measured by RT-qPCR, as described above. Quantity of RNA at each time point was determined by comparison with standard curves generated by amplification of the time-zero RNA sample.

The pEZX-AREG-3′UTR transfected 293T cells were co-transfected with or without shUPF1 using the GeneTran™ III reagent (Biomiga, Inc.). At 80% confluence and 48 h post transfection, cells were collected and the activities of both firefly luciferase (FLuc) and Renilla luciferase (RLuc) were measured according to the instructions of the dual luciferase reporter assay system (GeneCopoeia, Inc.). The internal standard for transfection efficiency was normalized to RLuc activity.

HaCaT and HEKα cells were plated at a density of 2×10⁴ cells/well in 96-well plates and incubated overnight. The proliferation index was measured using a Cell Counting Kit-8 (Dojindo Molecular Technologies, Inc.) at 48 h according to the manufacturer's instructions. For the cell wound healing assay, cells in the logarithmic growth phase were seeded in 24 well plates at a density of 5×10⁵ cells/ well. When the cells reached 100% confluence, a straight line was drawn in the cells in each well with a 10-µl pipette tip and the cells were serum-starved. Representative images (magnification, ×20) of wound closure were captured at 0, 6, 12, 24, and 48 h with an inverted microscope, and the wound area was measured using ImageJ software. The percentage of wound coverage was calculated by the difference in area between indicated time point and 0 h, divided by the area at 0 h.

The gene sequences used in the current study are publicly available at the GenBank [UPF1 mRNA (NM_002911.3) and AREG mRNA (NM_001657.4)]. Data on UPF1 mutations were deposited in the GenBank under accession numbers MH183202 (UPF1 c.2935_2936 insA) and MK089816 (UPF1 c.2030_2081 del 50 nts).

First, a paired samples test for UPF1 expression was concluded using GEO datasets GDS5392, GDS4602, GDS2518, GDS3539 and GDS4600. An analysis of 186 paired psoriasis and corresponding normal tissues showed that the UPF1 mRNA expression increased significantly in psoriasis (Fig. 1A). Increased UPF1 mRNA levels were also observed in sporadic patients' skin scales obtained in the current study, including four paired tissues (PA, PB, PC and PD; Fig. 1B), six independent scales and four independent normal tissues. The IMQ induced psoriasis-like skin showed morphological characteristics consistent with the symptoms of human psoriasis (28), such as skin thickening (Fig. S1 and Data S1) and hyperkeratosis (Fig. 1D), and RT-qPCR showed that the psoriasis-like skin expressed higher mRNA levels of proinflammatory cytokines and chemokines associated with psoriasis including AREG, interleukin (IL)-17A, IL-6, tumor necrosis factor-α and C-X-C motif chemokine 2 (data not shown). All 4 paired tissues from IMQ-induced psoriasis-like skin exhibited increased UPF1 mRNA levels (Fig. 1C), while protein levels were significantly decreased in 2/4 pairs, as detected by IHC (Fig. 1D and E). Abnormal UPF1 expression levels were also revealed by GEO database analysis for three other skin diseases tissues, including the Marfan syndrome, squamous cell carcinoma and melanoma tissues (Fig. S2 and Data S1). As UPF1 mRNA is a natural substrate of the NMD pathway, as well as a rate limiting element for NMD (6), the abnormal expression of UPF1 in those skin disease tissues was presumed to indicate a disturbed NMD pathway in the keratinocyte cells. According to the cell-type specific manner of NMD, multiple reported NMD substrates, which have been reported to be sensitive to UPF1 depletion in different cell lines (6-8), were tested by RT-qPCR in shUPF1 treated keratinocyte cells in a preliminary study. Three well-characterized endogenous NMD substrates, NIK, TBL2 and NAT9 mRNAs, were verified to increase in level in response to UPF1-depletion in the studied 293T and keratinocyte cells (data not shown), thus, the NIK, TBL2 and NAT9 mRNAs were selected to presented the NMD endogenous substrates. In the psoriasis-like mouse model, a disturbed NMD was hinted by the accumulation of NIK, TBL2 and NAT9 mRNAs in some paired lesions (Fig. 1F).

The skin scales are mostly composed of corneocytes (29); thus, the tissues may have been fragile for RNA sequencing. Five out of ten psoriasis scales from sporadic psoriasis patients were demonstrated to be reliable sources of RNA, as PCR fragments of UPF1 and AREG mRNAs were amplified successfully in these tissues (Fig. S3 and Data S1). Among the 5 available scales for RNA sequencing, 2 aberrant UPF1 transcripts were identified in patients A and B, and neither mutant was detected in the corresponding healthy epidermal tissue. The heterozygous mutation c.2935_2936insA (Fig. 2A) identified in patient A (a 36-year-old man with an affected buttock) generated an in-frame PTC located 136 nts upstream of the exon 22-23 junction (Fig. 2B). The heterozygous mutation c.2030-2081del (Fig. 2A) in patient B (a 48-year-old woman with affected arms) was spliced by the use of noncanonical splice donor/acceptor within exon 15 sharing a GGCC sequence, which generated a PTC 93 nts upstream of the exon 16-17 junction (Fig. 2B). There was no detectable mutation in the corresponding DNA or flanking introns, and, to the best of our knowledge, these two mutations had not been reported prior to the present study. The canonical ORF of UPF1 was disrupted in both aberrant transcripts, and the mutants were predicted to produce truncated proteins losing serine-glutamine cluster domain or the helicase domain (Fig. 2B).

To analyze the expression of UPF1 transcripts in psoriasis scales, specific primers were designed, according to the aberrant sequences (Fig. 2C). The specificity and efficiency of RT-qPCR for UPF1 insA were tested with allele-specific amplification. Transcript analysis demonstrated that the aberrant mutants were only detectable in psoriasis scales, whereas the wild-type (WT) transcripts were expressed in both the scale lesions and healthy cornified epidermal layer, showing a decrease in lesions (Fig. 2D).

The selective accumulation of the PTC-harboring transcripts in lesions but not in the adjacent normal tissues raised the possibility that the NMD pathway was perturbed in the two scales. Consistent with this hypothesis, it was found that the expression level of preselected NMD substrates, NIK, TBL2 and NAT9, was elevated in the psoriasis scales compared with normal control samples (Fig. 3A).

Successful overexpression and knockdown of UPF1 was confirmed (Fig. S5A and Data S1). The subsequent transfection of the UPF1 insA, UPF1 del or UPF1 WT constructs into UPF1-depleted 293T cells revealed that the mutated transcripts were expressed at a slightly lower steady-state level than UPF1 WT at both the mRNA and protein levels (Fig. 3B and C; lanes 3-5). As predicted above, the mutants generated minor protein bands that were predicted to be ~114 kDa for UPF1 insA and 78 kDa for UPF1 del (Fig. 3C; lanes 4-5), as compared with the 128 kDa full-length UPF1 protein (Fig. 3C; lane 3). The truncated UPF1 protein in psoriasis scales was not confirmed, as no protein sample could be collected. Subsequently, the expression levels of NMD substrates, which depended on mutated UPF1, were examined. Unlike UPF1 WT, which could rescue NMD activity in UPF1-depleted cells to decrease the level of NMD substrates, UPF1 insA and UPF1 del did not impact the NIK, TBL2 or NAT9 mRNA expression levels, indicating that the two mutants lacked a detectable NMD function (Fig. 3D).

After verifying the disruption of the NMD pathway by UPF1 mutants, gene regulation under a deficient NMD pathway was explored. As NMD targets diverse substrates and governs RNA homeostasis in a cell-specific manner (6), it was hypothesized that NMD deficiency would allow partial targets to be stabilized, leading to a disordered gene expression in a keratinocyte-specific manner. As the most abundant EGFR ligand gene in keratinocytes (30), AREG mRNA was identified as a candidate for the NMD pathway, since it harbors an EJ in its 3′UTR, which could trigger moderate NMD (Fig. 4A) (1). The expression level of AREG mRNA was increased in psoriasis tissues and UPF1-depleted keratinocyte cells, while UPF1 overexpression conversely decreased AREG mRNA levels in keratinocytes (Fig. 4B and C). Consistent with the well-established NMD substrates noted above, the dysregulation of AREG mRNA under UPF1 depletion could be rescued by UPF1 WT, but not by UPF1 insA or UPF1 del (Fig. 4D). This effect was also observed at the protein level, as shown by western blotting (Fig. 4E).

Given that transcript destabilization is the hallmark of direct NMD targets (31), RNA half-life analysis was performed on DRB-treated 293T cells with an inhibited transcriptional activity (32); it was found that AREG mRNA was stabilized ~1.23-fold in response to UPF1 depletion (Fig. 5A). To verify the NMD-triggering effect of AREG mRNA, AREG vector with or without its 3′UTR sequence was constructed (Fig. 5B), and the overexpression and knockdown of AREG was confirmed (Fig. S5B and Data S1). After knocking down endogenous UPF1 and AREG in 293T cells, no significant change was identified in the expression of exogenous AREG following transfection with AREG without the 3′UTR (Fig. 5C; lanes 5 and 6); however, the expression of AREG following transfection with AREG with 3′UTR was increased significantly (Fig. 5C; lanes 7 and 8), which indicated that the 3′UTR region was indispensable for NMD targeting. The current study further examined whether the 3′UTR of AREG was targeted by NMD via the dual-luciferase reporter system pEZX-FR02 (Fig. 5D). The results showed that the insertion of the AREG 3′UTR to pEZX-FR02 increased the FLuc activity following UPF1 depletion, indicating that this sequence was adequate for the triggering NMD (Fig. 5E).

UPF1 knockdown has been reported to result in cell apoptosis (33,34), while AREG positively regulates the proliferation of keratinocytes (11). Successful target gene expression regulation by UPF1 and AREG vectors was confirmed in keratinocyte cells (Fig. S6 and Data S1), and the cell morphology and area of cells did not change significantly (Fig. S7 and Data S1). The viability of UPF1-depleted HaCaT and HEKα cells was decreased, and the inhibition of AREG further decreased cell viability in these cells (Fig. 6A). These results were confirmed by determining the mRNA expression levels of proliferation marker protein Ki67, and BCL2 associated agonist of cell death (BAD) (Fig. 6B). Epidermal acanthosis, hyperkeratosis and angiogenesis have been previously detected in the skin of AREG transgenic animals (35), which indicates an ability of AREG to affect cell differentiation. UPF1 depletion markedly inhibited the expression of the differentiation gene keratin-5 (K5), the terminal differentiation marker filaggrin (FLG) and keratin-10 (K10). The expression levels of the differentiation markers were reversed in HaCaT and HEKα cells when AREG inhibition was coupled with UPF1 depletion (Fig. 6C).

Keratinocytes in which terminal differentiation is halted undergo changes in cell migration, and keratinocytes can repair wounds partly by migration (36). A wound healing assay was therefore performed to estimate the extent to which UPF1 depletion promoted cutaneous wound healing. UPF1 depletion increased the ability of both HaCaT and HEKα cells to migrate and re-epithelialize as demonstrated by the changes in cell-free areas. Following double knockdown of AREG and UPF1 in keratinocytes, the cell re-epithelialization was impaired compared with the shUPF1 group (Fig. 6D and E). The perturbed NMD in the shUPF1 group induced the expression of motility markers, cyclooxygenase-2 (COX-2), matrix metalloproteinase 1 (MMP1) and snail family transcriptional repressor 1 (SNAI1) compared with control group, while double knockdown of AREG and UPF1 lost this capacity (Fig. 6F). Keratinocytes also produce proinflammatory mediators that react against damage (37). In the current study, an increased expression of chemokine (C-C motif) ligand 20 (CCL20) and human recombinant GRO-α (CXCL1) was observed following NMD disruption in HaCaT and HEKα cells. The increased expression of these proinflammatory chemokines following NMD disruption was restored by AREG knockdown in these cells (Fig. 6F).