Poly zinc finger protein ZFP14 suppresses lymphomagenesis and abnormal inflammatory response via the HOXA gene cluster

Abstract

Poly zinc finger proteins (ZFP) that contain a KRAB (Krüppel-associated box) domain represent the largest class of transcription factors in higher organisms, but their roles in development and pathogenesis are largely undefined. ZFP14 (also known as ZNF531) contains thirteen zinc fingers and is highly conserved across species. Notably, we found that ZFP14 is frequently down-regulated in a multitude of human cancers, which correlates with poor prognosis of patients. Since ZFP14 has never been characterized, we generated a Zfp14-deficient mouse model to investigate the role of ZFP14 in development and pathogenesis. We showed that the mice deficient in Zfp14 had a short lifespan and were prone to diffuse large B-cell lymphoma (DLBCL), hyperplasia in multiple organs, systemic chronic inflammation, liver steatosis, and pancreatitis. Additionally, several pro-inflammatory cytokines, including IL-1β, IL18, and TNFα, were highly expressed in inflamed Zfp14^−/− mice spleens, livers, kidneys and lungs. To determine the underlying mechanism, RNA-seq analysis was performed and showed that the loss of ZFP14 led to increased expression of inflammatory and tumor-promoting genes. Out of the various tumor-promoting genes upregulated by ZFP14 loss, the HOXA gene cluster, which is known to promote lymphomagenesis and conserved between mouse and human, is consistently induced by loss of ZFP14. Moreover, we showed that the HOXA gene expression was inversely correlated with that of ZFP14 in human cancer patients and higher HOXA1 expression was correlated with poor patient prognosis. Together, we postulate that ZFP14 suppresses lymphomagenesis and abnormal inflammatory response by maintaining proper expression of the HOXA gene cluster.

1. Introduction

KRAB-ZFP proteins consist of a KRAB (Krüppel-associated box) domain followed by multiple zinc fingers and form the largest group of transcription factors in mammals [1,2]. KRAB-ZFPs are largely considered as DNA-binding transcriptional repressors and can activate transcription as well as bind to RNA and proteins to execute their key functions [2–4]. KRAB-ZFPs generally contain a KRAB domain at its N-terminus, followed by 10 or more tandem repeats of zinc fingers. The KRAB domain generally facilitates the recruitment of repressive complexes via its corepressor KRAB associated protein 1 (KAP1) [5] whereas the zinc finger domain aids in the recognition of specific nucleotides in DNA/RNA or specific amino-acids in proteins [3,6]. In the human genome, there are approximately 423 KRAB-ZFPs with only 103 KRAB-ZFP genes conserved across mammals [7], suggesting a functional diversity and lineage specificity of this special family.

Although several KRAB-ZFPs have been associated with the development and silencing of retro elements, the biological functions of most KRAB-ZFPs remain unknown and even less is known about their role in physiology and various diseases, including cancer. For the KRAB-ZFPs linked to various cancers, most of them are predicted to function as tumor suppressor genes, but some are known to be overexpressed in cancers and might act as oncogenes [8].

To explore the role of KRAB-ZFPs in cancer, we examined the expression profile of KRAB-ZFPs in various types of human cancers by analyzing the TCGA database with the focus on the conserved KRAB-ZFPs. We found that several KRAB-ZFPs with unknown functions, including ZFP14, were frequently altered in human cancers. ZFP14 (also known as ZNF531), which was originally identified from human adult and fetal brain cDNA libraries [9], is conserved across mammals and ubiquitously expressed in many tissues. Human ZFP14 contains 534 amino acids and shares 88 % sequence similarity with mouse Zfp14. Notably, the signature residues in all 13 zinc fingers are completely conserved between mouse and human ZFP14. Sequence alignment indicates that ZFP14 contains an N-terminal KRAB domain followed by 13 zinc fingers towards its C-terminus. Based on these functional domains, ZFP14 is predicted to regulate transcription by binding to sequence-specific DNA. However, the biological or physiological functions of ZFP14 have never been analyzed.

TCGA database showed that the low expression of ZFP14 correlates with poor prognosis in human cancer patients, suggesting a role of ZFP14 in tumor suppression. Thus, we took an initiative to explore the biology of ZFP14 with a focus on its role in tumorigenesis. To analyze the role of ZFP14 in vivo, we generated a Zfp14-deficient mouse model and found that loss of ZFP14 led to a shortened lifespan and increased incidence of spontaneous tumors, chronic inflammation, and liver steatosis. To understand the mechanism underlying multifaceted functions of ZFP14, RNA-seq was performed and showed that the loss of ZFP14 led to increased expression of inflammatory and tumor-promoting genes, in particular, the HOXA gene cluster, which is known to promote lymphomagenesis and conserved between mouse and human [10–12]. Moreover, we showed that the HOXA gene expression was inversely correlated with that of ZFP14 in human cancer patients. Together, we postulate that ZFP14 suppresses lymphomagenesis and abnormal inflammatory response by maintaining a proper expression level of the HOXA gene cluster.

2. Materials and methods

2.1. Cell culture, cell line generation and reagents

The human cancer cell line HCT116 was obtained from ATCC and used within 2 months of thawing or below passage 20. The human DLBCL cell lines, Pfeiffer and SUDHL6, were a gift from Dr. Hongwu Chen at University of California, Davis. HCT116 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Invitrogen) supplemented with 10 % fetal bovine serum (FBS) (Hyclone, Logan, UT). Pfeiffer and SUDHL6 cells were cultured in RPMI-1640 medium (Invitrogen) supplemented with 10 % FBS (Hyclone, Logan, UT). Generation of ZFP14-KO HCT116 cell lines was achieved by pSpCas9(BB)-2A-Puro vector expressing guide RNAs. The cells were selected with puromycin, and individual clones picked, genotyped, sequenced, and confirmed by PCR. The primers used for genotyping are listed in Supplementary Table S4.

2.2. RNA isolation and RT-PCR

Total RNA was harvested and isolated using TRIzol reagent (Invitrogen). 2 μg total RNA was used to synthesize cDNA using RevertAid First Strand cDNA Synthesis kit according to the manufacturer’s protocol (ThermoFisher Scientific^™). The levels of various transcripts were measured by semiquantitative and/or quantitative PCR with primers listed in Supplementary Table S5.

2.3. sgRNAs to generate knockout cell lines

To generate ZFP14-KO cells, two single-guide RNA (sgRNA) expression vectors pSpCas9(BB)-2A-Puro-sgZFP14–1 and pSpCas9(BB)-2A-Puro-sgZFP14–2 were used to remove exon 3 and create frame shift deletion. The generation of sgRNA expression vector was performed as described previously [13]. The oligonucleotides used for cloning sgZFP14–1 are sense, 5′-CAC CGG ATA GCA TGC AAA CAG CAT T-3′, and antisense, 5′-AAA CAA TGC TGT TTG CAT GCT ATC C-3′; for sgZfp14–2 are sense, 5′-CAC CGC TTT CTC TGC TGT CAA TTT C-3′, and antisense, 5′-AAA CGA AAT TGA CAG CAG AGA AAG C-3′.

2.4. Mice

The use of animals and the study protocols were approved by the University of California at Davis Institutional Animal Care and Use Committee. Zfp14^+/− mice were generated by the Mouse Biology Program at the University of California, Davis (Davis, CA, USA). The primers used to genotype the Zfp14-WT or Zfp14-KO allele are listed in Supplementary Table S4.

2.5. Mouse embryonic fibroblast (MEF) isolation

MEFs were isolated from 12.5 to 13.5 days post-coitum (pc) mouse embryos, as described previously [14]. To generate WT, Zfp14^+/− and Zfp14^−/− MEFs, Zfp14^+/− mice were interbred. The MEFs were cultured in DMEM supplemented with 10 % FBS (Hyclone Laboratories, Erie, PA, USA), 55 μM β-mercaptoethanol and 1× non-essential amino acids solution (Cellgro, Manassas, VA, USA).

2.6. Histological analysis

Mouse tissues were fixed in 10 % neutral buffered formalin, processed, and embedded in paraffin blocks. Tissue blocks were sectioned (6 μm) and stained with H&E.

2.7. . RNA-seq analysis

Total RNA was harvested from two isogenic controls (#24 and #41) and two ZFP14-KO (#10 and #25) HCT116 cells using TRIzol reagent (Invitrogen). 1 % agarose gels were used to monitor RNA degradation. NanoPhotometer spectrophotometer (IMPLEN, CA, USA) was used to measure RNA purity, whereas the RNA Nano 6000 Assay Kit of the Bioanalyzer 2100 system (Agilent Technologies, CA, USA) was used to assess RNA integrity and quantitation. RNA sample preparation was performed by using 1 μg of RNA per sample. NEBNext UltraTM RNA Library Prep Kit from Illumina (NEB, USA) was used to generate sequencing libraries following manufacturer’s recommendations and index codes were added to attribute sequences to each sample. Briefly, poly-T oligo-attached magnetic beads were used to purify mRNA from the total RNA. The mRNA was fragmented in NEBNext First Strand Synthesis Reaction Buffer containing divalent cations under elevated temperature. Following this, random hexamer primers and M-MuLV Reverse Transcriptase (RNase H-) was used to synthesize first strand cDNA. Subsequently, DNA polymerase I and RNase H were used to perform second strand cDNA synthesis. Overhangs remaining after second strand synthesis were converted into blunt ends by exonuclease/polymerase activities. 3′ ends of DNA fragments were adenylated and ligated with NEBNext Adaptor with hairpin loop structure to prepare for hybridization. cDNA fragments of preferentially 150– 200 bp in length were selected by purifying the library fragments with AMPure XP system (Beckman Coulter, Beverly, USA). Then 3 μl USER Enzyme (NEB, USA) at 37 °C for 15 min followed by 5 min at 95 °C before PCR. Then PCR was performed with Phusion High-Fidelity DNA polymerase, Universal PCR primers and Index (X) Primer. Finally, PCR products were purified (AMPure XP system) and library quality was assessed on the Agilent Bioanalyzer 2100 system.

2.8. RNA-seq data analysis

2.8.1. Quality control

Raw data (raw reads) of FASTQ format were initially processed through fastp. In this step, clean data (clean reads) were obtained by removing reads containing adapter and poly-N sequences and reads with low quality from raw data. At the same time, Q20, Q30 and GC content of the clean data were calculated. All the downstream analyses were based on the clean data with high quality.

2.8.2. Mapping to reference genome

Reference genome hg19 and gene model annotation files were downloaded from genome website browser (NCBI/UCSC/Ensembl) directly. Paired-end clean reads were aligned to the hg19 reference genome using the Spliced Transcripts Alignment to a Reference (STAR) software, which is based on a previously undescribed RNA-seq alignment algorithm that uses sequential maximum mappable seed search in uncompressed suffix arrays followed by seed clustering and stitching procedure. STAR exhibits better alignment precision and sensitivity than other RNA-seq aligners for both experimental and simulated data.

2.8.3. Quantification

FeatureCounts was used to count the read numbers mapped of each gene. And then RPKM of each gene was calculated based on the length of the gene and reads count mapped to this gene. RPKM, Reads Per Kilo-base of exon model per Million mapped reads, considers the effect of sequencing depth and gene length for the reads count at the same time.

2.8.4. Differential expression analysis

Prior to differential gene expression analysis for each sequenced library, the read counts were adjusted by Trimmed Mean of M-values (TMM) through one scaling normalized factor. Differential expression analysis of two conditions was performed using the edgeR R package. The P values were adjusted using the Benjamini and Hochberg methods. Corrected p-value of 0.005 and $\left|{\log }_2^{(\text{Fold Change})}\right|$ of 1 were set as the threshold for significantly differential expression.

2.8.5. GO enrichment analysis

GO is the abbreviation of Gene Ontology (http://www.geneontology.org/), which is a major bioinformatics classification system to unify the presentation of gene properties across all species. GO enrichment analysis of differentially expressed genes was implemented by the clusterProfiler R package. GO terms with corrected P value <0.05 were considered significantly enriched by differential expressed genes.

2.9. Statistical analysis

Mice survival analysis was performed by Kaplan–Meier survival analysis using the log rank test. Data are presented as means ± standard error of the mean (SEM). The p values for qRT-PCRs were calculated using the two-tailed Student’s t-tests, and p < 0.05 was considered statistically significant. For each experimental data point, n = 3. Excel (Microsoft, Redmond, WA) was used for statistical analyses.

3. Results

3.1. Low levels of ZFP14 correlate with poor prognosis in human cancer patients

To identify the cancer relevant KRAB-ZFPs that could predict human patient prognosis, we examined the expression of several KRAB-ZFPs from the TCGA database using Xena browser. Many of the KRAB-ZFPs we examined were from recently reported publications shown to have aberrant expression in cancers [8,15]. Notably, out of the about 20 cancer-relevant KRAB-ZFPs examined, lower expression of ZFP14 significantly correlated with poor overall survival, poor disease-free survival and poor progression-free survival in human cancer patients from TCGA pan-cancer database (Figs. 1A–C). Moreover, lower ZFP14 expression significantly correlated with poor patient survival in specific cancer types (Figs. 1D–J). Thus, we focused on ZFP14 to determine its role in tumorigenesis.

Fig. 1.

Low levels of ZFP14 correlate with poor prognosis in human cancer patients.

(A-C) Kaplan-Meier survival curves showing low ZFP14 expression correlating with poor overall survival (A), poor disease-free survival (B) and poor progression-free survival (C) in human cancer patients from TCGA pan-cancer database. The survival curves were created using the UCSC Xena browser.

(D-J) Kaplan-Meier survival analysis reveals that low ZFP14 mRNA expression correlates with shorter patient survival in 8 types of tumors from TCGA database. The survival curves were created using the OncoLnc tool (http://www.oncolnc.org/).

(K) The KRAB domain and the C2H2 zinc fingers of human and mouse ZFP14 are depicted in the illustration. Also shown are the signature variable residues present at positions −1, +2, +3 and +6 in each zinc finger and are conserved between mouse and human ZFP14.

Human ZFP14 (ZNF531) contains an N-terminal KRAB domain followed by a linker region and 13 zinc fingers (Fig. 1K). Mouse Zfp14 is 88 % similar to human ZFP14 and consists of the same domains as human ZFP14 (Fig. 1K and Supplementary Fig. S1). Importantly, variable residues at positions −1, +2, +3 and +6 of each zinc finger that decide the substrate identity are identical between human and mouse ZFP14 (Fig. 1K), indicating conservation of their function across species.

3.2. Zfp14^−/− and Zfp14^+/− mice have a shorter lifespan and are prone to spontaneous tumors

As lower ZFP14 levels in human cancer patients correlate with poor patient survival, we reasoned that the Zfp14-KO mice should be susceptible to spontaneous tumor formation. Thus, a Zfp14-deficient mouse model was generated using traditional embryonic stem cell targeting strategy and confirmed by genotyping (Supplementary Fig. S2A and B).

To examine the biological functions of Zfp14 in vivo, a cohort of Zfp14^+/− and Zfp14^−/− mice was generated and monitored for long-term survival, predisposition to tumors and other pathological abnormalities. We would like to note that to reduce the numbers of mice used, WT mice were adapted from previous studies [16,17]. Nevertheless, these WT mice had the same C57BL/6 genetic background as Zfp14-deficient mice and were maintained in the same facility. We found that the median lifespan for Zfp14^+/− (n = 34; 86.9 weeks) and Zfp14^−/− (n = 19; 80 weeks) mice were significantly shorter than that for WT mice (n = 56; 117 weeks) (Fig. 2A; Supplementary Tables S1–S3). However, there was no significant difference in median lifespan between Zfp14^+/− and Zfp14^−/− mice (Fig. 2A). Moreover, we found that when compared to WT mice, both Zfp14^+/− and Zfp14^−/− mice were prone to spontaneous tumors (Fig. 2B; Supplementary Tables S1–S3). The tumor penetrance of 43.3 % in Zfp14^+/− mice was significantly higher than that in WT mice (21.6 %) (Fig. 2B, p-value = 0.047 by Fisher’s exact test). Similarly, the tumor penetrance in Zfp14^−/− (47.1 %) was higher than in WT mice, but it was not statistically significant (Fig. 2B, p-value = 0.061 by Fisher’s exact test). Notably, while both WT and Zfp14-deficient mice were prone to lymphomas, we found that the incidence of diffuse large B cell lymphoma (DBLCL) was markedly increased in both Zfp14^+/− and Zfp14^−/− mice as compared to that in WT mice (WT vs. Zfp14^+/− : p = 0.0113; WT vs. Zfp14^−/− : p = 0.0041 by Fisher’s exact test) (Fig. 2C, Supplementary Tables S1–S3). Indeed, 11 out of 30 Zfp14^+/− mice and 8 out of 17 Zfp14^−/− mice, whereas only 6 out of 51 WT mice, developed DLBCL, suggesting that Zfp14-deficiency promotes B-cell lymphomagenesis.

Fig. 2.

Zfp14^−/− and Zfp14^+/− mice have a shorter lifespan and are prone to spontaneous tumors.

(A) Kaplan-Meier survival curve for WT, Zfp14^+/−, and Zfp14^−/− mice.

(B) Tumor spectra and penetrance in WT, Zfp14^+/−, and Zfp14^−/− mice.

(C) The numbers and percentages of WT, Zfp14^+/−, and Zfp14^−/− mice with DLBCL.

3.3. Zfp14-deficient mice are susceptible to inflammation in multiple organs

In addition to increased tumor formation in Zfp14-deficient mice, we examined the major organs for other potential abnormalities by performing histological analysis. Histopathological analyses indicated that the mice deficient in Zfp14 developed chronic inflammation in multiple organs including liver, lung, and salivary gland (Fig. 3A). In fact, 90 % of Zfp14^+/− and 94 % Zfp14^−/− mice were found to develop chronic inflammation in >3 organs and 46.7 % of Zfp14^+/− and 82.3 % of Zfp14^−/− mice developed inflammation in >4 organs (Fig. 3B, Supplementary Tables S2–S3). By contrast, not a single WT mouse developed chronic inflammation in >3 or 4 organs (Fig. 3B, Supplementary Table S1). Additionally, Zfp14^−/− mice exhibited a marked increase in the percentage of mice with inflammation in >4 organs when compared to Zfp14^+/− (Fig. 3B).

Fig. 3.

Zfp14-deficient mice are susceptible to inflammation and pancreatitis.

(A, C) Representative images of hematoxylin and eosin (H&E)-stained WT, Zfp14^+/−, and Zfp14^−/− liver, lung and salivary gland showing inflammation (A) or pancreatitis (C).

(B, D) The numbers and percentages of WT, Zfp14^+/−, and Zfp14^−/− mice with inflammation in >3 or >4 organs (B) or pancreatitis (D).

(E-F) The levels of various inflammatory cytokine mRNAs were quantified by RT-PCR in spleen (E), liver (F), kidney (G) or lung (H) of gender-matched WT and Zfp14^−/− mice.

When comparing inflammation in a single organ, Zfp14-deficiency enhanced the susceptibility of mice to pancreatitis, as well as liver, lung, kidney, and salivary gland inflammation (Figs. 3C–D; Supplementary Fig. S3A–D). Moreover, we observed substantially increased levels of various inflammatory cytokine mRNAs in spleen, liver, kidney and lung of Zfp14-deficient mice compared to WT mice (Fig. 3E–H). As DLBCLs associated with increased inflammation have been identified as a separate class of tumors by World Health Organization and are associated with more advanced stage and poor patient survival in humans [18,19], the above findings suggest that the Zfp14-deficient mice develop aggressive DLBCL.

3.4. Zfp14-deficient mice develop liver steatosis and are prone to hyperplasia in several organs

In addition to inflammation, we also looked for other abnormalities in Zfp14-deficient mice organs. Interestingly, we showed that both Zfp14^+/− and Zfp14^−/− mice developed liver steatosis at a significantly higher rate compared to WT mice (Fig. 4A–B; Supplementary Tables S1–S3), indicating aberrant lipid metabolism in Zfp14-deficient mice.

Fig. 4.

Zfp14-deficient mice are susceptible to liver steatosis and hyperplasia in multiple organs.

(A) Representative images of hematoxylin and eosin (H&E)-stained WT, Zfp14^+/−, and Zfp14^−/− livers showing steatosis in Zfp14-deficient mice.

(B) The numbers and percentages of WT, Zfp14^+/−, and Zfp14^−/− mice with liver steatosis.

(C) Representative images of hematoxylin and eosin (H&E)-stained WT, Zfp14^+/−, and Zfp14^−/− spleen, thymus and lymph nodes showing hyperplasia in Zfp14-deficient mice.

(D-G) The numbers and percentages of WT, Zfp14^+/−, and Zfp14^−/− mice with hyperplasia in spleen (D), thymus (E), skin (F) or lymph node (G).

Additionally, compared to WT, both Zfp14^+/− and Zfp14^−/− mice had significantly higher occurrence of hyperplasia in spleen, thymus, skin and lymph nodes (Figs. 4C–G; Supplementary Tables S1–S3), emphasizing their susceptibility to increased tumor formation. Taken together, Zfp14-deficient mice are susceptible to increased liver steatosis and hyperplasia in various organs in addition to their susceptibility to DLBCL formation and inflammation.

3.5. ZFP14-KO leads to increased expression of inflammatory and tumor-promoting genes

Although increased inflammation could result in increased tumorigenesis observed in the Zfp14-deficient mice [20,21], we generated ZFP14-KO HCT116 human colorectal cancer cells using CRISPR-Cas9 and performed RNA-seq analysis to determine the genes and pathways regulated by ZFP14 that might be involved in tumorigenesis (Fig. 5A–B). Notably, we found that ZFP14 differentially regulated several pathways linked to the abnormalities we observed in Zfp14-deficient mice, such as hyperplasia, pancreatitis, and dermatitis (Fig. 5B). Moreover, we found that the ZFP14-deficiency leads to up-regulation of multiple inflammatory cytokines as well as their receptors (Fig. 5B–C), corroborating with increased inflammation seen in the Zfp14-deficient mice.

Fig. 5.

ZFP14-KO leads to increased expression of inflammatory and tumor-promoting genes.

(A) Volcano plot showing genes differentially regulated by ZFP14. The X-axis indicates fold-change and the Y-axis p-value.

(B) Gene-set enrichment analysis was performed to determine pathways differentially regulated by ZFP14.

(C, D) Heat map of inflammation-related genes (C) and tumor-promoting genes (D), and their differential expression in isogenic control and ZFP14-KO HCT116 cells.

Furthermore, we observed increased expression of several tumor-promoting genes upon ZFP14 deletion. These genes included genes from the HOXA cluster, the L-type amino acid transporter SLC7A8, the G-protein coupled receptor ADGRF1, secretory leukocyte peptidase inhibitor SLPI and the transcription factor RUNX2 (Fig. 5D). Based on its predicted function as a transcriptional repressor, most of the genes altered upon ZFP14 deletion were upregulated (Fig. 5C–D). Interestingly, ZFP14 deletion led to the increased expression of all the HOXA cluster genes including the non-coding RNA HOTAIRM1 transcribed from the HOXA cluster (Fig. 5D), indicating involvement of ZFP14 in repressing the entire HOXA gene cluster.

3.6. ZFP14 modulates tumorigenesis by regulating the expression of the HOXA gene cluster in humans and mice

Next, we confirmed the increased expression of the tumor-promoting genes obtained from the RNA-seq analysis, by qRT-PCR, in HCT116 ZFP14-KO cells (Fig. 6A). We were able to confirm the increased expression of most of the tumor promoting genes, except RUNX2 (Fig. 6A), showing the reproducibility of the RNA-seq data. Interestingly, all three HOXA cluster genes confirmed by qRT-PCR showed significant increase upon ZFP14-KO (Fig. 6A), reaffirming the regulation of the HOXA cluster by ZFP14.

Fig. 6.

ZFP14 modulates tumorigenesis by regulating the expression of HOXA gene cluster in humans and mice.

(A) qRT-PCR confirmation of some ZFP14 regulated genes from RNA-seq in isogenic control and ZFP14-KO HCT116 cells.

(B) qRT-PCR was performed in WT and Zfp14^−/− MEFs to confirm the regulation of some of the significantly regulated ZFP14 target genes from human cells.

(C-D) Pfeiffer and SUDHL6 cells were electroporated with scramble or ZFP14 siRNA and the levels of ZFP14 (C) and HOXA1 (D) mRNAs were measured by qRT-PCR.

(E-F) The levels of HoxA1 and HoxA9 mRNA were quantified by RT-PCR in spleen (E) or liver (F) of gender-matched WT and Zfp14^−/− mice.

(G) The inverse correlation of ZFP14 expression with HOXA1 gene expression in TCGA pan-cancer database using the UCSC Xena browser.

(H) Kaplan-Meier survival curve showing higher HOXA1 expression correlating with poor prognosis in human cancer patients.

As ZFP14 function in cancer appeared to be similar in mouse and humans, we next examined if the regulation of some of the tumor-promoting genes by ZFP14 in human cells is also conserved in mouse cells. For this, we performed qRT-PCR in WT and Zfp14^−/− MEFs and showed that the regulation of the entire HoxA gene cluster by ZFP14 appeared to be conserved in mice (Fig. 6B). However, the huge increase in expression of SLPI and SLC7A8 observed in ZFP14-KO human cells was not observed in Zfp14-KO MEFs (Fig. 6B). Previous studies showed that HOXA cluster is frequently up-regulated in human B-cell lymphomas and promotes leukemia in animal models [22–24]. Thus, we further corroborated the HOXA cluster regulation by ZFP14 in human DLBCL cell lines, Pfeiffer and SUDHL6. As these are suspension cell lines, we knocked down ZFP14 in these cell lines using electroporation. As seen in Fig. 6C, we observed a robust decrease of ZFP14 mRNA in ZFP14-KD samples. Importantly, we saw a significant increase of the HOXA1 mRNA levels in these cells (Fig. 6D), re-affirming the negative regulation of the HOXA genes by ZFP14. Next, we analyzed HoxA gene expression in the DLBCL present in the spleen of Zfp14^−/− mice as well as in the WT and Zfp14^−/− mice liver tissues. As seen in Fig. 6E, we observed increased expression of HoxA1 and HoxA9 mRNAs in spleen from Zfp14^−/− mice that had DLBCL, suggesting the increased expression of the HoxA genes could be responsible for promoting DLBCL in the Zfp14-KO mice. Moreover, we also showed increased HoxA gene expression in the liver tissues of Zfp14^−/− mice as compared to the WT mice (Fig. 6F).

Since ZFP14 is involved in repressing the HOXA cluster gene expression, we analyzed the TCGA pan cancer database and showed that HOXA1 mRNA expression inversely correlated with that of ZFP14 in cancer tissues (Fig. 6G). Moreover, we also showed that contrary to ZFP14, higher HOXA1 expression correlated with poor patient prognosis in the pan cancer TCGA database (Figs. 1A and 6H) as well in specific tumor types (Fig. 1D, H, I and Supplementary Fig. S4A–C), underscoring the inverse correlation of ZFP14 and HOXA gene expression in cancer. Taken together, ZFP14 appears to suppress tumorigenesis by repressing the expression of the HOXA gene cluster in human and mice.

4. Discussion

The KRAB-ZFPs consist of the largest family of transcriptional factors in higher organisms [1,2]. Although recent studies indicated that the expression of many KRAB-ZFPs is altered in several tumor types, the role of KRAB-ZFP in cancer remains ambiguous [8,15]. To expand our knowledge of KRAB-ZFPs in cancer, we carried out a study to determine physiological significance of ZFP14, a KRAB-ZFP with unknown function, by using multiple genetic and biomedical approaches. We showed that the lower expression of ZFP14 correlates with poor prognosis in human cancer patients as well as in mice. Specifically, we found that Zfp14-deficient mice were prone to increased inflammation and developed aggressive DLBCL, which is in part mediated by its regulation of the HOXA gene cluster.

DLBCL is the most common form of non-Hodgkin’s lymphoma in humans and accounts for about 49 % of the B-cell cancer worldwide [25,26]. Even with better treatment options available, approximately half of the patients with DLBCL show recurrence or have a poor outcome [25]. Recently, systemic immune-inflammation has emerged as a prognostic biomarker in DLBCL that can assist in separating high-risk patients [18,19]. Notably, patients with DLBCL that show higher systemic immune-inflammation have advanced disease and poor survival [18,19]. Not only DLBCL with increased inflammation are aggressive, but continuous chronic inflammation itself can act as a predisposition for tumor formation and metastasis [27]. As Zfp14-deficient mice are prone to multi-organ inflammation possibly due to elevated expression of pro-inflammatory cytokines, it is reasonable to assume that ZFP14 has an anti-inflammatory activity. Although acute inflammation is transient and beneficial to the organism, chronic inflammation elevates cancer risk in multiple organs, underscoring the function of the immune system in promoting tumorigenesis [21,28]. In line with this, we showed that several pro-inflammatory cytokines along with their receptors were up-regulated by loss of ZFP14 in mouse tissues and human cells. Indeed, like ZFP14, ZNF300 and ZNF281 have been shown to regulate inflammation in vitro [29,30]; however, their roles in inflammation in vivo have yet to be explored. Nonetheless, these data suggest that ZFP14 may have a profound role in modulating inflammatory response, which might be responsible for the development of aggressive DLBCL in Zfp14-deficient mice. While the regulation of inflammation may be a direct consequence of ZFP14 loss, increased HOXA gene expression resulting from Zfp14 deletion may also contribute to cancer-promoting inflammation [10]. These findings, although intriguing, warrant further work.

The clusters of the Hox family of genes are evolutionarily conserved and encode for DNA-binding transcription factors that are essential for anteroposterior body axis patterning during early development and are frequently dysregulated in cancers [10–12]. Of the four Hox gene clusters, the overexpression of genes from the HOXA cluster not only can cause leukemia in animal models, but are also overexpressed in human cancers [22–24]. Likewise, identifying factors that can regulate the HOX gene expression in normal as well as cancer cells is an important area of research in cancer biology. Although an underlying mechanism is not elucidated here, it is possible that ZFP14 functions as a transcriptional repressor to modulate expression of the HOXA gene cluster since they are located closely and often regulated simultaneously [31]. Thus, further studies are needed to elucidate how ZFP14 modulates the HOXA gene cluster and its role in the development of B-cell lymphoma. Taken together, we conclude that the ZFP14-HOXA pathway plays an indispensable role in inflammation and cancer formation in mouse and humans.

Data availability

All the data are available in the main text or the supplementary materials. The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.