A shortened ZNF334 gene mutation may cause inflammation triggered by cold

Whole-exome sequencing of the patient's genome was performed to elucidate the genetic mechanisms underlying this unusual case of autoinflammation and progressive SNHL, induced by mild cold exposure. Although no known pathogenic variant was detected in either the germline or the mosaic analysis pipelines, two candidate variants with rare population frequencies and high predicted protein impacts were identified (Appendix Table S1). Of these two variants, the ZNF334 frameshift mutation, which has a low frequency in the Taiwanese population (0.067%), was detected in the patient's blood cells, not in her healthy siblings', with this confirmed through Sanger sequencing (Fig. 1H). This heterozygous frameshift mutation (c.1197_1198del, encoding p.Thr399fs) was predicted to affect multiple C2H2-type ZnF domains of ZNF334. Two nucleotides were discovered to be deleted in ZNF334, that correspond to the mRNA sequence encoding threonine (399 T) and glycine (400 G). This deletion resulted in the conversion of glutamine (401E) to lysine, and led to the early termination of the protein, which in turn resulted in the loss of subsequent C2H2-type ZnF domains containing several highly conserved amino acids (Fig. 1G). As presented in Fig. 1G, the amino acids 401E, 406C, and 409C immediately following the frameshift mutation were conserved, whereas the flanking amino acids demonstrated variability across species. In addition to the patient's healthy siblings, one of her daughters was also tested as wild-type in ZNF334, with this confirmed using Sanger sequencing. However, no data were available on the patient's other child, who was working abroad, or on her parents, who were deceased (Fig. 1H). Although the sequencing depth and ratio of the ZNF334 frameshift variant detected in the patient's blood cells were 42/104 and 40.38%, respectively, suggesting a heterozygous germline mutation pattern, the likelihood of this variant originating from acquired post-zygotic mosaicism could not be ruled out because of the late-onset nature of the disease. Therefore, ZNF334 was further sequenced in monocytes (CD14 + ) and T cells (CD3 + ) isolated from the patient's peripheral blood mononuclear cells (PBMCs), and a heterozygous p.T399fs mutation was detected in both cell subsets (Fig. 1H). Notably, because no buccal swab sample was available, ZNF334 Sanger sequencing was performed on two formalin-fixed, paraffin-embedded (FFPE) blocks of skin biopsy lesion samples. These samples were obtained from the patient's back (at the age of 47 years), and chest (at the age of 51 years), and they primarily comprised epithelial cells and fibroblasts with a small number of infiltrating leukocytes. In contrast to expectations, a heterozygous ZNF334 frameshift mutation was detected only in the skin biopsy sample obtained from the patient's back (Fig. 1H). Collectively, these findings suggest that the ZNF334 frameshift mutation occurs during a postzygotic event, and that both lymphoid and myeloid blood cells carry the heterozygous ZNF334 mutation, potentially contributing to disease pathogenesis.

To analyze the inflammatory cell signature, CITE-seq was performed using PBMCs collected from the healthy sister and the patient under two conditions: cold + ++ is+ (S2) and cold+ is+ (S3). A total of 11 cell types were clustered in the uniform manifold approximation and projection plot (UMAP) (Fig. 2B). The top ten differentially expressed genes and cell surface markers (detected by 11 oligonucleotide-conjugated antibodies against cell surface markers, Fig. EV1A) per cluster are shown in the heatmap (Fig. 2E). Compared with her sister, the patient had increased proportions of cluster 2 (TCF7+ activated T cells), cluster 4 (CD4 + T cells) and cluster 10 (CD16+ proinflammatory monocytes) in PBMCs collected under both conditions (Fig. 2C,D). Of note, when the patient's PBMCs were exposed to long periods of cold (S2), the percentage of cluster 5 (composed of S100A4 + S100A6+ monocytes and IL32 + ANXA1+ activated T cells) markedly increased compared with in the control and S3 conditions (Fig. 2D). In addition, the patient's PBMCs exhibited reduced proportions of myeloid subsets expressing genes involved in the regulation of NF-ƘB signaling, inflammatory responses, and protein folding (Figs. 2D and EV1B,C). Further gating on CD45⁺ CD14⁺ monocyte populations revealed increased expression levels of the monocyte activation markers S100A4, S100A6 and CD44 and reduced expression levels of the NF-ƘB signaling regulators NFKB2, NFKBIA, and BCL3 in the patient's samples (Fig. 2F). Gating on CD45 + CD3 + CD4 + T cells also revealed decreased expression levels of NF-ƘB signaling regulators, together with the downregulation of PIM2 and the upregulation of ANAX1 and IL32, particularly in S2 (Fig. 2G). These data suggest not only enhanced inflammation and dysregulated stress responses involving NF-ƘB and protein folding signaling in the patient's monocytes, but also enhanced Th1 activation.

To explore the role of ZNF334 in regulating cold stress responses in monocytes, including inflammation mediated by TNF-α/NF-ƘB and IL-6/STAT3 signaling, cell proliferation, and cell death, we generated heterozygous and homozygous ZNF334-edited human THP-1 monocyte cell lines by using the CRISPR/Cas9 technology. These cells expressed a truncated protein (85 amino acids) for all ZNF334 isoforms that lacked all C2H2-type ZnF domains (Fig. EV2). Subsequently, we performed real-time PCR to analyze inflammation-related gene expression; determined IL-1β levels in the supernatants; and evaluated cell cycle regulation, apoptosis, and necroptosis in ZNF334-knockout and wild-type THP-1 monocytes at various time points during cold stimulation at 32 °C.

In cell cycle analysis, we observed higher proportions of ZNF334^+/− and ZNF334^−/− cells in the S and G2 phases and lower proportions of these cells in the G0/G1 phase at baseline (all P = 0.0002; Fig. 4I–K). Upon cold stimulation at 32 °C, the proportions of cells in the S and G2 phases increased significantly, and those of cells in the G0/G1 phase decreased in the ZNF334 wild-type, ZNF334^+/−, and ZNF334^−/− THP-1 cells over time. The most prominent change occurred in ZNF334^−/− cells after 7 h of cold exposure. The median proportions of cells in ZNF334 wild-type, ZNF334^+/−, and ZNF334^−/− THP-1 cells were 60.81%, 39.33%, and 15.62% in the G0/G1 phase (P = 0.0002; Fig. 4I); 33.90%, 45.02%, and 54.22% in the S phase (P = 0.0002; Fig. 4J); and 3.89%, 12.45%, and 26.36 in the G2 phase (P = 0.0002; Fig. 4K), respectively.

We examined cell death by using live cell imaging (Fig. EV4). Baseline cell membrane staining revealed that membrane integrity was impaired in ZNF334-knockout cells, particularly in the homozygous knockout THP-1 cells (P = 0.0002; Fig. 4L). The percentages of early apoptotic cells (P = 0.0005) and necroptotic cells (P = 0.0012) were higher in the ZNF334^+/− and ZNF334^−/− groups than in the wild-type group before cold stimulation (Fig. 4M,N). After 7 h of exposure to 32 °C, the percentage of early apoptotic cells increased among wild-type THP-1 monocytes. However, the highest percentages of early apoptotic cells were still detected in the ZNF334^+/− and ZNF334^−/− groups (P = 0.015; Fig. 4M).

To further evaluate the impact of ZNF334 truncation mutation on the expression of Hsp90, p-p65, and p-STAT3 (Tyr705), we performed the immunofluorescence staining of these molecules in ZNF334 wild-type and ZNF334-knockout THP-1 monocytes exposed to or not exposed to 8 h of cold stimulation at 32 °C. The expression of Hsp90 increased in viable ZNF334 wild-type, ZNF334^+/−, and ZNF334^−/− THP-1 monocytes after 8 h of cold stimulation at 32 °C compared with their baseline levels (all P < 0.0001; Fig. 4O). The Hsp90 protein expression was lower in ZNF334^+/− and ZNF334^−/− cells than in wild-type cells both before (P < 0.0001) and after 8 h (P < 0.0001) of 32 °C treatment (Fig. 4O). Furthermore, after 8 h of cold stimulation at 32 °C, increased nuclear levels of p-p65 (P = 0.006; Fig. 4P) and p-STAT3 (P < 0.0001; Fig. 4Q) were detected in ZNF334 wild-type monocytes compared with their baseline levels. ZNF334-knockout THP-1 monocytes exhibited significantly higher nuclear levels of p-p65 (37 °C, P < 0.0001; 32 °C, P = 0.0003; Fig. 4P) and p-STAT3 (37 °C, P < 0.0001; 32 °C, P < 0.0001; Fig. 4Q) than did ZNF334 wild-type cells. At baseline, ZNF334^+/− THP-1 monocytes had the highest level of nuclear p-p65, whereas ZNF334^−/− cells had the highest level of nuclear p-STAT3 (Fig. 4P,Q).

The results conclude that THP-1 monocytes harboring truncated ZNF334 with the loss of all C2H2-type ZnF domains had reduced intracellular Hsp90 expression, enhanced baseline and cold-induced inflammation, an interrupted cell membrane, and a higher rate of cell death.

To further investigate the potential molecular mechanisms of ZNF334 truncation mutation on monocyte responses to cold stress, we performed RNA-seq analysis on ZNF334 wild-type, ZNF334^+/−, and ZNF334^−/− THP-1 monocytes before and after 8 h of cold stimulation at 32 °C. In addition, differential gene expression analyses were conducted to compare the average normalized count per gene between ZNF334^+/− and wild-type, and between ZNF334^−/− and wild-type samples under both conditions. These analyses were performed using multiple t tests with Benjamini–Hochberg (B-H) adjustments for corrected P values. Network and functional enrichment analyses of genes upregulated or downregulated in ZNF334-edited THP-1 monocytes (adjusted P value (FDR) < 0.05 and |log₂ fold change | ≥0.8 compared with ZNF334 wild-type cells) were performed using String v.12.0 (sting-db.org). We identified the top three enriched downregulated biological pathways (Gene Ontology) were chaperone-mediated protein folding, response to topologically incorrect protein, and response to unfolded protein; whereas the top three enriched upregulated biological pathways were defense response, immune response, and immune system process (Fig. 4R). Specifically, the gene expression levels of key endoplasmic reticulum (ER)-resident enzymes responsible for antioxidative activity and oxidative protein folding (Wang et al, 2014; Zito et al, 2010), such as peroxiredoxin 4 (PRDX4), glutathione peroxidase 7 (GPX7), and protein disulfide isomerase 4 (PDIA4), were decreased in ZNF334^+/− and ZNF334^−/− monocytes as compared with ZNF334 wild-type at 37 °C and after 32 °C cold stimulations (Fig. 4S). The expressions of genes related to inflammatory responses and monocyte activations, including CD74, lymphocyte antigen 86 (LY86), hematopoietic cell kinase (HCK), and HLA-DRB1 were already upregulated at baseline and further increased after cold stimulations in ZNF334^+/− and ZNF334^−/− THP-1 monocytes (Fig. 4S). Of note, genes involving NF-κB signaling pathway, like RELB and TNF, were markedly upregulated in cold-stimulated ZNF334^+/− and ZNF334^−/− THP-1 cells (Fig. 4S). Network analysis of the genes shown in the heatmap using String v.12.0 revealed that these molecules were all related to the pathway of response to stress, and potential interactions were identified among them (Fig. 4T). These data suggest that wild-type ZNF334 is required to maintain the ER protein folding response necessary to regulate redox homeostasis, and plays a role in suppressing cold-induced NF-κB activation in monocytes.

Since TRPM8 is the main cold sensor and regulator of cellular ROS homeostasis (Bidaux et al, 2016; Iftinca and Altier, 2020), we then examined the expression levels of TRPM8 in ZNF334-knockout THP-1 monocytes through immunofluorescence confocal microscopy. TRPM8 expression was detected on the plasma membrane and ER in THP-1 monocytes (Fig. EV5). Overall TRPM8 levels were found to be significantly lower in ZNF334-knockout THP-1 monocytes than in ZNF334 wild-type cells before cold stimulation (P < 0.0001; Fig. 5B). Although the total and nuclear levels of TRPM8 increased after 8 h of 32 °C treatment in the three THP-1 cell lines (ZNF334 wild-type, ZNF334^+/−, and ZNF334^−/− monocytes), lower levels were still detected in ZNF334-knockout cells (P < 0.0001; Fig. 5B).

While examining the expression of TRPM8 in THP1 cells under confocal microscopy, we noted the presence of TRPM8+ extracellular vesicles (EVs)/microvesicles ranging in size from 1 to 5 μm. It has been reported that damaged mitochondria induced by oxidative stress could be segregated into large EVs called exophers in C. elegans (Melentijevic et al, 2017), and activated human monocytes could also release microvesicle-embedded mitochondria, which trigger inflammatory responses in endothelial cells (Puhm et al, 2019). Therefore, we further analyzed the mitochondrial component in TRPM8 + EV using MitoTracker staining under confocal microscopy. We found that the secretion of 1–5 μm TRPM8+ mitochondria+ EVs was markedly increased in ZNF334 knockout THP-1 monocytes upon 8 h of cold stimulation, as compared with ZNF334 wild-type monocytes (P = 0.0029; Fig. 5C,D).

To examine the impact of the exophers on the differentiation and activation of other monocytes, we isolated microvesicles secreted by cold-stimulated ZNF334 wild-type, ZNF334^+/− and ZNF334^−/− THP-1 cells, ranging in size from 0.45 to 1 μm via qEV Isolation Columns (Izon Science, Christchurch, New Zealand) and membrane filters, then applied the respective isolated microvesicles to ZNF334 wild-type THP-1 monocytes and collected the supernatant after 48 h of incubation at 37 °C for cytokine analysis. We found that as compared with cold-stimulated wild-type-derived exophers, microvesicles derived from 32 °C-treated ZNF334-knockout monocytes stimulated the ZNF334 wild-type THP-1 cells to produce more TNF-α and chemokines IP-10 (CXCL10), RANTES (CCL5), MCP-1, and MIP1-ß in this single experiment (Fig. 5E).

These data indicate that ZNF334 truncation in THP-1 monocytes reduces the cellular expression of TRPM8, which is further associated with excessive oxidative stress and cold-induced secretion of TRPM8+ mitochondria-containing EVs with proinflammatory potentials and chemotactic activities for monocytes and lymphocytes.

Next, since Hsp90 is known to be the major stress regulator and an abundant chaperone able to interact with other proteins to maintain cellular homeostasis (Bhattacharya et al, 2022), and we have detected dysregulations in Hsp90 expression, chaperone-mediated signaling and cold stress responses in monocytes harboring the truncated ZNF334, we further studied if ZNF334 could interact with Hsp90, and if ZNF334 p.Thr399fs truncation mutation would impair this interaction, contributing to the observed cellular immunophenotype. Prediction of the protein-ligand complex binding affinity between ZNF334-wild type and Hsp90α or Hsp90ß homodimer using AlphaFold-Multimer revealed potential binding of ZNF334 with Hsp90α homodimer. Yet, when forming a complex with Hsp90α homodimer, ZNF334 mutant was predicted to have smaller interface and less hydrogen bonds than ZNF334 wild-type, suggesting decreased binding affinity (Fig. 6C). Of note, the ZNF334 p.Thr399fs mutant was predicted to lose its interaction with the N-terminal ATPase domain of Hsp90 (Fig. 6C). We then evaluated the impact of ZNF334 p.Thr399fs on the interaction with Hsp90 via immunoprecipitation (IP)/Western analyses on monocytes derived from the patient and from healthy controls. Although the Immunoprecipitation using anti-Hsp90α/ß antibody detected interactions of Hsp90α/ß with ZNF334 in both patient's and healthy control's monocytes, immunoprecipitation using anti-ZNF334 antibody only detected Hsp90α/ß in healthy controls, not in patient's monocytes (Fig. 6D,E).

To further characterize the physiological changes in the interactions of Hsp90 with ZNF334 and other cold stress regulators identified in our monocyte model of cold stimulation, we performed IP/Western analyses on monocytes derived from healthy controls with or without in vitro 8 h of 32 °C cold exposure. Hsp90 interacted with ZNF334, TRPM8 (83 kDa and 41 kDa isoforms) and IKKα at 37 °C and after 32 °C cold stimulations (Fig. 6F–H). The binding between ZNF334 and Hsp90 decreased (Fig. 6F), while the interaction between Hsp90 and IKKαincreased after cold exposure (Fig. 6G). These results identified ZNF334 and TRPM8 as novel proteins interacting with Hsp90, and suggest that the reduced interaction between ZNF334 and Hsp90 might contribute to the enhanced inflammatory responses.

In this study, we have examined the functional impact of monocyte ZNF334 truncation mutation in a patient with an AID induced by repeated cold exposure, which we referred to as "ZACAS". We searched gnomAD v4.1, a large population-based genomics database that contains the genotypic data of more than 1.6 million alleles, for other ZNF334 truncation mutations nearby. We noticed several ZNF334 variants predicted to cause frameshift or stop-gain effects around the 300th to 400th amino acids of ZNF334, with low total population frequencies (Appendix Table S2). However, these variants were not reported in ClinVar, and data of Sanger sequencing validations, monocyte functional assays, and environmental cold exposures were unavailable, making it difficult to interpret their clinical significance. Next, we examined whether the ZNF334 loss-of-function expression signature identified through RNA-seq analysis is present in monocytes derived from patients with other chronic inflammatory diseases, such as juvenile idiopathic arthritis (JIA). Analysis of the RNA-seq data of monocytes derived from patients with polyarticular JIA (E-MTAB-14035↗) (Hounkpe et al, 2024) revealed the downregulated expression of ZNF334 and the oxidative protein folding enzyme PRDX4, along with upregulated expression of genes related to monocyte activation and pro-inflammatory cytokines (Appendix Fig. S2).

In order to investigate whether other low-level acquired genetic variants related to autoinflammation contributed to the patient's late-onset phenotype, we further deep-sequenced DNAs extracted from PBMCs of the patient and her ZNF334-wild type sister for NLRP3, NLRP12, NLRC4, PLCG2, TNFRSF1A, and ZNF334. Similar to the results of whole exome sequencing, ZNF334 p.Thr399fs with a variant allele frequency of 53.1% was detected in the patient, but not in the sister. Moreover, NLRP3 p.Thr347Ile with variant allele frequency of 14% (sequencing depth 16544) was also detected only in the patient; while a highly prevalent NLRP3 synonymous single-nucleotide polymorphism was detected in both samples with variant allele frequency of 99.8%.

To further evaluate the potential impact of the low-level NLRP3 p.Thr347IIe on inflammatory cytokine production in monocytes, we reanalyzed the CITE-seq single-cell data by gating on CD14+monocytes derived from PBMCs of the healthy sister (HC) and from PBMCs of the patient collected at two conditions: after a period of frequent cold exposure (S2) and after a period of cold avoidance (S3), both were under the use of immunosuppressant. A total of 592 monocytes were analyzed in each sample: the number of TNF+ monocytes were 7, 25, 17 in HC, S2, and S3, respectively; the number of NLRP3+ monocytes were 2, 12, 5 in HC, S2, and S3, respectively; the number of IL-1B+ monocytes were 9, 8, 9 in HC, S2, and S3, respectively. Taken together, different levels of somatic mutations (53.1% ZNF334 p.Thr399fs and 14% NLRP3 p.Thr347Ile) were detected in the patient's blood cells. The numbers of TNF+ monocytes and NLRP3+ monocytes were highest in S2, while the proportions of IL-1B+ monocytes were similar among the three samples.

This study included the patient, her healthy sibling (who had wild-type ZNF334), and unrelated healthy controls without cold-induced disorders, other rheumatic diseases, or malignancies.

Venous whole blood samples were collected from the patient and healthy controls; from these samples, PBMCs were isolated through density gradient centrifugation performed using Ficoll-Paque PLUS (GE Healthcare Biosciences, IL, USA, Cat#17144002) and cryopreserved in a liquid nitrogen tank for subsequent experiments, including whole exome sequencing and CITE-seq. For RNA extraction and western blotting, monocytes were further isolated from PBMCs by using the overnight adhesion method (monocyte-enriched to 80–90%).

WES was performed on genomic DNA extracted from whole blood from the patient and healthy siblings. Briefly, genomic DNA was extracted from the PBMCs by using the DNeasy Blood & Tissue Kit (Qiagen, Manchester, UK). DNA libraries were prepared using the Agilent SureSelect Human All Exon Kit (Agilent Technologies, CA, USA) and sequenced on the Illumina HiSeq platform by using 2× 100-bp paired reads. Bcl2fastq Conversion Software was used to demultiplex data and convert BCL files into the FASTQ file format. Alignment and variant calling were based on the hg19 reference using Burrows–Wheeler Aligner (Li and Durbin, 2009). Data were analyzed as described previously (Yang et al, 2018). The following criteria were used to select potential candidates: a mutant allele frequency of ≥30%, a global minor allele frequency of <1%, and pathogenicity predicted by at least one of three software programs (SIFT, PolyPhen, and CADD_PHRED).

To validate candidate variants and detect the ZNF334 genotype in the PBMCs derived from the family members of the patient, Sanger sequencing was performed as described previously (Chang et al, 2020). The ZNF334 reference sequence used for primer design and nucleotide numbering was NM_ 001270497. PCR products amplified by specific primers (whose sequences are listed in Appendix Table S3) were sequenced on an ABI 3100 Genetic Analyzer (Applied Biosystems) and analyzed using the DNAStat Lasergene software. The patient's PBMCs were further sorted by BD FACSMelody Cell Sorter using antibodies against CD3 (SK7, BD Bioscience Cat#345765) and CD14 (M5E2, BD Bioscience Cat#557742) for DNA extraction and Sanger sequencing. DNAs of the FFPE blocks of skin biopsy samples derived from the patient were also extracted using the DNA Sample Preparation Kit (Roche Diagnostics) for Sanger sequencing. Sequencing of T cells, monocytes, and skin biopsy samples were performed on SeqStudio Genetic Analyzer (Applied Biosystems).

Sera collected from the patient and her sister at different time points were stored at −80 °C, and the levels of TNF-α (R&D Systems, Cat#HSTA00E), IL-1β (R&D Systems, Cat#HSLB00D), IL-6 (R&D Systems, Cat#HS600L), and eHsp90 (Elabscience, Cat#E-EL-H1864) were measured using ELISA kits in accordance with the manufacturer's instructions.

Oligonucleotide-conjugated antibodies (Abseq antibodies, all purchased from BD Biosciences; Appendix Table S4) and BD Rhapsody Single-Cell Analysis System (BD Biosciences) were used to integrate cell surface protein immunophenotyping and transcriptome measurements into a single-cell readout, as reported previously (Geng et al, 2021). In brief, PBMCs from the patient (and her healthy sister) were collected during an episode of cold-induced urticaria and arthralgia and during a period without cold-induced inflammatory symptoms. Each sample was first stained with Calcein AM (Thermo Fisher Scientific, Cat#C1430) and Draq7 (BD Bioscience, Cat#564904) to determine cell viability. Viable cells in each sample were labeled with different sample tags by using the Human Single-Cell Multiplexing Kit (BD Bioscience, Cat#633781), stained with Abseq antibodies, and pooled together to load into a primed BD Rhapsody Cartridge following the manufacturer's protocol. Then, the cells were lysed in the microwell cartridge to hybridize mRNA molecules to barcoded capture oligos on beads, which were subsequently retrieved by using the BD Rhapsody Cartridge Reagent Kit (BD Biosciences, Cat#633733). Microbead-captured single-cell transcriptome and sample tag information were reverse-transcribed into cDNA by using the BD Rhapsody cDNA Kit. Libraries were created using the BD Rhapsody Whole Transcriptome Analysis Amplification Kit and sequenced on the NovaSeq 6000 platform (Illumina; 150-bp paired-end run).

The BD Rhapsody Whole Transcriptome Analysis Pipeline (version 1.9.1) based on GRCh38 mRNA annotation was used to process and demultiplex sequencing data (fastq file). The resulting output files were analyzed and visualized using SeqGeq v1.7 software (BD Biosciences). The quality control criteria were set at cells containing between 200 and 4000 expressed genes. The built-in R Plugin Seurat Clustering Pipeline was used for dimensionality reduction, graph-based KNN clustering, and differential gene expression analysis. Uniform manifold approximation and projection (UMAP) layouts and heatmaps were generated to visualize differentially expressed cell populations among the PBMCs derived from the three samples as described previously. In brief, principal component analysis was performed on the basis of highly variable genes, and a two-dimensional projection was generated in the UMAP space by using the first 50 principal components. Cells labeled as "undetermined" or "multiplet" were excluded from the subsequent analyses. The UMAP approach identified 11 clusters. The top 30 upregulated genes in adjacent cluster contrasts were identified on the basis of a false discovery rate of <0.01 and a log2 fold change of >0.25, sorted by log2 fold change. The presence of sequenced oligonucleotide-conjugated antibodies (cell surface markers), together with the top four enriched pathways of the top 30 genes in each cluster, was used to identify cell types. The top enriched pathways were detected using STRING v.11.5 network analysis against the following criteria: count in network >4, false discovery rate <0.01, and strength >1.0, sorted by strength.

Incubation of monocytes at 32 °C has been used to evaluate the hyperinflammatory responses to mild cold exposure in patients of Familial cold autoinflammatory syndrome (FACS) (Rosengren et al, 2007). Therefore, for cold stimulation assays, two million PBMC-derived monocytes or THP-1 monocytes were incubated in complete Roswell Park Memorial Institute (RPMI) medium (RPMI 1640 medium supplemented with 10% fetal bovine serum and 1% penicillin–streptomycin) overnight at 37 °C under 5% CO₂; and subsequently, the temperature of the incubator was adjusted to 32 °C for 24 h to stimulate PBMC-derived monocytes. For the cold stimulation of THP-1 monocytes, different time points were used, ranging from 1 to 24 h.

The human monocytic THP-1 cell line harboring ZNF334 wild-type or truncation mutation were maintained in complete RPMI medium at 37 °C in a humidified atmosphere under 5% CO₂. For the evaluation of the effect of inhibitor against TNF-α(adalimumab, AbbVie Inc.), Hsp90/eHsp90 (geldanamycin, Enzo, Cat#BML-EI280), or TRPM8 (N-(3-Aminopropyl)−2-[(3-methylphenyl) methoxy]-N-(2-thienylmethyl) benzamide hydrochloride, AMTB, Sigma-Aldrich, Cat#SML0103) on the attenuation of cold-induced upregulation of inflammatory genes in THP-1 monocytes, and for the evaluation of the effects of inhibitors on the attenuation of cold-induced upregulation of inflammatory genes in THP-1 monocytes harboring wild-type ZNF334 or having heterozygous ZNF334 truncation mutation (ZNF334^+/−), THP-1 monocytes were cultured at 1 × 10⁶ cells per ml in complete RPMI medium at 37 °C with or without the presence of the respective inhibitor according to previous protocols (Hornsby et al, 2022b; Hsu et al, 2007; Lin et al, 2017; Wax et al, 2003), and then being cold-stimulated at 32 °C. Briefly, in TNF-α inhibitor assay, THP-1 monocytes were pretreated with 10 µg/ml adalimumab, 10 µg/ml heat-inactivated (95 °C, 10 min) adalimumab control (Ctrl), or medium only for 2 h, and then placed in the 37 °C or 32 °C incubator for another 2 h. Relative fold change of TNF, NLRP3, or IL-6 mRNA expression was calculated as compared with the mean expression level in cells cultured at 37 °C with medium only. In Hsp90 inhibitor assay, THP-1 monocytes were pretreated with 2 µM geldanamycin or DMSO only for 1 h, and then placed in the 37 °C or 32 °C incubator for another 24 h. Relative fold change of TNF, NLRP3 or IL-6 mRNA expression was calculated as compared with the mean expression level in cells cultured at 37 °C with DMSO only. In TRPM8 inhibitor assay, THP-1 monocytes were pretreated with 10 µM AMTB or DMSO only for 24 h, and then placed in the 37 °C or 32 °C incubator for another 24 h. Relative fold change of TNF, NLRP3 or IL-6 mRNA expression was calculated as compared with the mean expression level in cells cultured at 37 °C with DMSO only. In the ruxolitinib inhibitor assay, THP-1 monocytes were pretreated with 1 µM ruxolitinib or DMSO only for 1 h, and then placed in the 37 °C or 32 °C incubator for another 24 h (Prutsch et al, 2024). Relative fold change of TNF, NLRP3, or IL-6 mRNA expression was calculated as compared with the mean expression level in cells cultured at 37 °C with DMSO only.

CRISPR/Cas9 ribonucleoprotein–mediated genome editing was performed to establish heterozygous and homozygous ZNF334 loss-of-function THP-1 cell lines. THP-1 monocytes harboring the heterozygous ZNF334 truncation mutation were designated as ZNF334^+/− THP-1 cells, whereas THP-1 monocytes harboring the homozygous ZNF334 truncation mutation were designated as ZNF334^−/− THP-1 cells. In brief, an efficient gRNA was designed to remove an exon region present in all the coding transcripts of ZNF334, thus creating a truncation mutation (Fig. EV2A). The synthetic, chemically modified gRNA was obtained from Dharmacon (CO, USA). THP-1 cell lines were synchronized with 200 ng/mL nocodazole for 17 h; this was followed by Cas9/gRNA delivery through electroporation. On the day of electroporation, a gRNA complex solution (150 pmol) containing complementary RNA and transactivating CRISPR RNA (tracrRNA) was mixed with Cas9 protein (30 pmol) in Opti-MEM Reduced Serum Medium (Thermo Fisher, Cat#31985062) at room temperature for 15 min to form Cas9/gRNA ribonucleoproteins. Simultaneously, 1 × 10⁵ synchronized cells were collected through centrifugation, washed with Opti-MEM twice, and resuspended in optimal volumes of Opti-MEM; this was followed by the addition of Cas9/gRNA ribonucleoproteins. Then, the ribonucleoproteins were delivered into the cell by using the NEPA21 electroporator under optimized pulse voltage conditions. The electroporated cells were immediately transferred to 24-well plates containing 1 mL of complete RPMI medium and then incubated at 37 °C under 5% CO₂. After a 48-h recovery period, we performed a T7E1 digestion assay and high-resolution melting analysis to evaluate the efficiency of gene editing. THP-1 cells containing the gene edit were seeded, and single-cell clones were subjected to high-resolution melt-qPCR analysis to verify mutations in the target gene (Fig. EV2B).

RNAs were extracted from wild-type, ZNF334^+/−, and ZNF334^−/− THP-1 cells at 37 °C or after 8 h of incubation at 32°C using QIAGEN RNeasy Micro kit (cat# 74004, QIAGEN GmbH, Hilden, Germany). Total RNA was subjected to high-throughput sequencing using the Illumina NextSeq 2000 at the Cancer and Immunology Research Center at National Yang Ming Chiao Tung University. Raw reads were processed using CLC Genomics Workbench v25.0.1. Filtered reads were aligned to the human reference genome (GRCh38), and rRNA reads were identified and filtered using the SILVA rRNA database (v138.1). Processed RNA sequencing count data (Gene expression matrices) were imported into RStudio running R (v4.5.0) packages. To reduce background noise, genes with low expression (fewer than 10 counts across all samples) were removed, except for ZNF334, which was retained due to prior biological relevance. Filtered counts were normalized using the median-of-ratios method implemented in the DESeq2 package (v1.48.1), with sample-specific size factors estimated from the distribution of counts. Multiple pairwise comparisons of gene expression values were made between ZNF334^+/− or ZNF334^−/− with ZNF-wild type at 37 °C or at 32 °C. Genes with an adjusted P value (FDR) < 0.05 and |log2 fold change | ≥0.8 were considered significantly differentially expressed. Adjusted P values were calculated by the Benjamini–Hochberg procedure. String v.12.0. (string-db.org) was used to predict potential interactions among these differentially expressed genes and to analyze enriched pathways in the network.

Briefly, mRNA synthesis was carried out using the HiScribe T7 mRNA Kit with CleanCap Reagent AG (New England Biolab, USA, Cat#E2080) according to the manufacturer's instructions. The transfection of IVT-mRNAs in primary monocytes was performed using Lipofectamine MessengerMAX (Thermo Scientific, Cat#LMRNA008). Monocytes from healthy controls were seeded in a 96-well plate, and the IVT-mRNAs in Opti-MEM medium were subsequently added to Lipofectamine MessengerMAX solution at a 1:1.5 ratio and mixed well. The Lipomessenger Max-mRNA complexes were then added to each well. Cells were incubated at 37 °C for 2 h, with or without subsequent cold stimulation at 32 °C for 6 h. RNAs were extracted from cells afterward for RT-PCR.

Total RNA was extracted from monocytes or THP-1 cells by using TRIzol (Cyrusbioscience, New Taipei City, Taiwan, Cat#2001) and reverse-transcribed by using the cDNA Reverse Transcriptase Kit (Takara Bio USA, San Jose, CA, USA, Cat#RR037B). Real-time PCR (RT-PCR) was performed using the SYBR FAST qPCR Master Mix (KAPA Biosystems, Cat#KK4600) or the IQ2 TagMan Probe qPCR Master Mix (Bio-Genesis, Cat#BB-DBU-010). The cycling conditions were as follows: one preincubation cycle at 95 °C for 30 s; 50 amplification cycles at 95 °C for 10 s, 60 °C for 30 s, and 72 °C for 10 s; and a final cooling step at 40 °C for 30 s. The peptidylprolyl isomerase B (PPIB) gene was used as the housekeeping gene. The comparative threshold cycle (Ct) method was used to calculate the expression levels of the target genes: ΔCt = Ct^{target gene} − Ct^{PPIB (housekeeping gene)}. PCR primers are listed in Appendix Table S3.

THP-1 monocytes were incubated with 20 μM DCFDA (Abcam, Cat#ab113851) in phenol red-free RPMI medium at 37 °C. For cold stimulation assays, cells were exposed to 32 °C for 2 h in the incubator, and were harvested and centrifuged to remove the DCFDA-containing medium. The cells were then analyzed in the FITC channel of FACSCanto II flow cytometer.

THP-1 cell lines (ZNF334-wild-type, ZNF334^+/−, and ZNF334^−/−) were seeded onto 96-well imaging plates (Greiner); and stained with Hoechst 33258 (Sigma, Cat# 23491-45-4), Annexin V (BD Biosciences, Cat# 51-65874X), propidium iodide (PI; BD Biosciences, Cat#51-66211E), and the cell membrane stain CellVue (Merck, Cat#CB_000730). Then, the plates were placed in an environmentally controlled live cell imaging system; the obtained images were analyzed using MetaXpress (Molecular Devices). The application module using multiwavelength cell scoring settings was used to analyze membrane integrity per cell and the rate of apoptotic and necroptotic cells per image field. Because a higher PI intensity can differentiate necroptotic cells from apoptotic cells (Pietkiewicz et al, 2015), we defined the necroptotic gating threshold as triple-positivity for Hoechst, Annexin V, and higher PI intensity (>10,000 gray levels of intensity above local background; Fig. EV4) and the early apoptotic gating threshold as double-positivity for Hoechst and Annexin V. Cell cycle analysis was performed using the Cell Cycle Module of MetaXpress to determine the cold stimulation–induced changes in the proportions of cells in G1, S, and G2 phases.

A monolayer of THP-1 cells was attached to coverslips by using the Cytospin centrifugation method. The cells were fixed with 100% methanol for 15 min and blocked with 1% bovine serum albumin (Gibco, Cat#30063481) for 30 min on a shaker. Then, the cells were stained with primary antibodies against ZNF334 (Invitrogen, PA5-31843; 1:100 dilution), HSP90α/β (Santa Cruz, F-8, SC-13119; 1:100 dilution), TRPM8 (ECM, M571, TM5711, 1:100 dilution; Invitrogen, PA1-46239, 1:100 dilution), STAT3 (Invitrogen, 9D8, MA1-13042; 1:100 dilution), phospho-NF-κB p65 (Ser536) (Cell Signaling, 93H1, 3033, 1:100 dilution; Santa Cruz, 27.Ser 536, sc-136548, 1:100 dilution), phospho-STAT3 (Tyr705) (Cell Signaling, D3A7, 9145, 1:100 dilution), and Calreticulin (Invitrogen, 1G6A7, MA5-15382, 1:100 dilution) overnight at 4 °C. Next, the cells were washed and stained with fluorochrome-conjugated secondary antibodies, including anti-rabbit Alexa Fluor 488 (Invitrogen, A-11008; 1:200 dilution) and anti-mouse Alexa Fluor 594 (Invitrogen, A-11005; 1:200 dilution). For multicolor (>3 colors) staining, an Alexa Fluor 647 conjugation kit (Abcam, Cat#ab269823) was used for labeling. Coverslips were mounted with an antifade mounting medium containing DAPI (Abcam, Cat#ab104139). The slides were analyzed under a confocal laser scanning microscope (ZEISS, LSM 800 microscope). ImageJ (Fiji, v.1.53.t) was used to quantify the intensity of positive staining. For the detection of mitochondria-containing EVs, cells were labeled with 100 nM Mitotracker (Invitrogen, Cat#M22426) at 37 °C for 30 min before Cytospin centrifugation and further staining.

Cold-induced extracellular microvesicles ranging in size from 0.45 to 1 μm were isolated from cold-stimulated (32 °C 8 h) ZNF334 wild-type, ZNF334^+/−, and ZNF334^−/− THP-1 monocytes via qEV size-exclusion chromatography columns (Izon Science, cat#IC1-70) following the manufacturer's protocol. The isolated fraction containing microvesicles ranging in size from 70 nm to 1 µm were collected and further filtered through a 0.45-µm membrane filter (catalog number: PALL® MAPM45C68), to obtain microvesicles ranging in size from 0.45 to 1 μm. These microvesicles derived from ZNF334-wild-type or ZNF334 knockout THP-1 cells were transferred to ZNF334 wild-type THP-1 monocytes, and the supernatants were collected 48 h after incubation at 37 °C for cytokine analysis using Bio-Plex Pro Human Cytokine 27-plex Assay (BioRad, Cat#M500KCAF0Y).

For immunoblotting, monocytes derived from 2 million PBMCs or THP-1 cells were incubated with a lysis buffer (RIPA buffer containing a 1× phosphatase inhibitor cocktail and 1 mM phenylmethylsulfonyl fluoride) for 30 min on ice. Proteins in cell lysates were resolved through sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) and transferred to polyvinylidene difluoride membranes. Then, the membranes were probed with appropriate primary antibodies, washed, and incubated with relevant horseradish peroxidase–conjugated secondary antibodies (Thermo Scientific). After washing, the membranes were incubated in SuperSignal West Pico Plus substrate (Thermo Scientific, Cat#34580) to visualize the proteins of interest. The No-Stain Protein Labeling Reagent (Invitrogen, Cat#A44449) was used as a loading control. Quantification of the blotted band relative to the loading control was performed using iBright Analysis software (version 5.1.0).

For immunoprecipitation assays, cell lysates were first mixed with specific primary antibodies and Complete Co-IP/Wash Buffer (Universal Magnetic Co-IP kit, Active Motif, CA, USA, Cat#54002) at 4 °C on a rotator for 1.5 h, and then incubated with Protein G Magnetic Beads (Active Motif, Cat#54002) at 4 °C on a rotator overnight. Afterward, the precipitants were washed and eluted in accordance with the manufacturer's instructions (Active Motif, Cat#54002), and the immunoprecipitated proteins were separated through SDS–PAGE. Antibodies used for immunoblotting and immunoprecipitation were ZNF334 (Invitrogen, PA5-31843), HSP90α/β (Santa Cruz, sc-13119), and TRPM8 (Invitrogen, PA1-46239), and IKKα (Cell Signaling, 3G12, 11930) at 1:1000 dilution for immunoblotting and 1:20 dilution for immunoprecipitation.

Statistical analyses were performed using GraphPad Prism software, version 10.2.3 (Graphpad Software). The non-parametric Kruskal‒Wallis test was performed for multiple comparisons; and Wilcoxon signed-rank test was used for paired comparisons. Significant differences were reported if P < 0.05. No blinding was applied to our data analysis.

The study protocols were approved by the Institutional Research Ethics Committee of China Medical University Hospital (protocol numbers: CMUH109-REC3-165, CMUH110-REC3-015, and CMUH111-REC2-228). Written informed consent was obtained from each participant in accordance with the Helsinki Declaration. Written informed consent was received from the patient for the use of the photographs, and that the record of informed consent has been retained. In addition to the principles set out in the WMA Declaration of Helsinki the experiments with patient samples also conformed to the Department of Health and Human Services Belmont Report.