Zfp92 tissue expression, mRNA forms and Zfp92 knockout mice

Using data from our prior gene correlation network analysis of purified cell populations representing main stages of β-cell development from embryonic day (E) 8.0 to postnatal day (P) 60 in mice [31], we observed that Zfp92 expression is increased at E15.5 in pro-endocrine pancreatic progenitor cells that express Neurog3, further upregulated in nascent β-cells and maintained in mature β-cells (Fig 1A). RT-qPCR of mouse adult tissues showed that Zfp92 is predominantly expressed in pancreatic islets, with higher levels also detected in brain, whole pancreas, and testis samples (Fig 1B). In addition, scRNA-Seq data from Tabula Muris [32] indicate that Zfp92 is predominantly expressed in β-cells, with lower levels in α- and δ-cells in mouse islets [31].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Zfp92 expression and CRISPR-derived knock-out allele.

A) Expression levels of Zfp92 mRNA profiled by RNA-seq in sorted pancreatic developmental lineages: gut tube endoderm (Sox17⁺ at E8.0/8.5), posterior foregut endoderm (Pdx1⁺ at E9.5), pancreatic multipotent progenitor cells (Ptf1a⁺ at E10.5), endocrine progenitor cells (Neurog3⁺ and Insm1⁺ at E15.5), nascent β-cells (Ins⁺ at E16.5), and adult β-cells (Ins⁺ at P60). Two profiled mutant conditions for endocrine progenitor cells (Neurog3^-/- and Insm1^-/- at E15.5) indicate that Zfp92 is reduced in the absence of Neurog3. N = 3.B) Tissue expression patterns of Zfp92 determined by the RT-qPCR analysis of RNA from tissues from adult mice (10 weeks old). All tissues, except ovaries, are from male mice. N = 3. Relative expression is calculated in comparison to the same gene expression in the intestine. Br, brain; Lu, lung; He, heart; Th, thymus; Mu, skeletal muscle; Sk, skin; Ad, adipose; Li, liver; Pa, whole pancreas; Is, pancreatic islets; Sp, spleen; In, small intestine; Ki, kidney; Ov, ovary; Te, testis. C) Schematic representation of Zfp92 gene and relative gene location and sequence of CRISPR/Cas9 generated 7 bp deletion. Deletion of 7 bp in Zfp92 causes a frameshift that interrupts ZFP92 protein translation resulting in a predicted 50 aa peptide containing 8 missense amino acids. D) Representative image of a PCR genotyping gel for Zfp92 wild type (WT), heterozygous (Het), and knockout (KO) mice.

https://doi.org/10.1371/journal.pgen.1010729.g001

Since the Zfp92 gene model predicts two mRNA forms, a long form that encodes a full length ZFP92 protein and a short form that encodes a C-terminally truncated ZFP92 (S1A Fig), we designed primer pairs specific to each form and performed RT-PCR and qPCR on islet samples to establish which is predominantly expressed in islets, (S1A–S1D Fig). The results show that the predominant RNA form expressed in pancreatic islets is the longer variant that encodes a 488 aa protein containing an N-terminal KRAB domain and an array of 8 C₂H₂ zinc fingers.

Since Zfp92 is highly enriched in both islets and pancreatic β-cells, we hypothesized that this gene may serve an important but undefined role in pancreatic islet development and/or function. For this reason, we performed CRISPR/Cas9 mutagenesis by injecting guide RNAs targeting Zfp92 into mouse zygotes that express CRISPR/Cas9. Analysis of the offspring revealed a founder animal bearing a 7 bp frameshift deletion downstream of the translation initiation codon in the gene (Fig 1C and 1D). The 7 bp deletion creates a frameshift in an open reading frame that leads to a premature STOP codon resulting in a truncated 50 aa protein thereby generating a null (KO) allele for Zfp92 (Fig 1C).

Zfp92 KO mice exhibit sex-specific alterations in body weight and fasting blood glucose

Zfp92 homozygous KO mice were viable, fertile and did not demonstrate any gross abnormalities. Weekly monitoring of body weight for 10 weeks of male and female mice maintained on normal diet revealed that Zfp92 KO males have lower body weight than their wild type (WT) counterparts beginning from 6 months of age (Fig 2A). We did not observe any significant differences in body composition measurements at 14 months indicating that weight differences are not due to disproportionate accumulation of fat (Fig 2B and 2C). We assessed glucose homeostasis in WT and KO mice by measuring fasting and fed blood glucose, plasma insulin, and by performing intraperitoneal glucose and insulin tolerance tests (Fig 2D–2I, S2A and S2B Fig). These measurements revealed a slight decrease in fasting blood glucose in male mice (Fig 2D), however, the mice showed similar fed blood glucose and insulin, insulin secretion in response to glucose injection (Fig 2H), and no differences in glucose (Fig 2G) and insulin tolerance (Fig 2I). Immunostaining and morphometric analyses of pancreatic tissue sections from KO and WT mice did not show any abnormalities in islet architecture, islet composition or β-cell area (Fig 2J and 2K).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Growth, body fat composition, glucose homeostasis, and islet structure of Zfp92 KO mice.

A) Body weight measurements from 4 to 14 weeks showing Zfp92 KO male mice have reduced growth in comparison to WT males (n = 7,8 for each sex and genotype. Lean body mass (B) and body fat (C) measurements show no difference between KO and WT mice body composition at (n = 5–7 for each sex and genotype). Fasted (D) and fed (E) blood glucose concentration measurements. KO mice have lower fasting blood glucose than WT mice. (n = 8–11 for each sex and genotype). F) Random fed plasma insulin levels (n = 4, male mice). G) Intraperitoneal glucose tolerance test (GTT) results are presented as the area under the curve (AUC) measurements for S2A Fig (n = 7 for each sex and genotype). H) Insulin secretion in response to glucose injection in GTT (n = 4, males). Both genotypes show an equal increase in plasma insulin concentration 15 min after injection. I) Intraperitoneal insulin tolerance test (ITT) results presented as the area under the curve (AUC) measurements for S2B Fig (n = 7–8 for each sex and genotype). J) Immunofluorescent staining for insulin (red) and glucagon (green) of pancreatic islet tissues from WT and Zfp92 KO mice. Nuclei are stained with DAPI (blue). Scale bar: 100 microns. K) Quantification of hormone-positive areas in KO and WT islets (n = 3, males at 14–15 weeks). B-K) All mice are 14–15 weeks old. Statistical significance between KO and WT samples is indicated by black brackets, and between sexes by gray brackets. Error bars: ± SEM. **p≤0.01; *p≤0.05. p-value determined by ANOVA or by t-test (panels F, K).

https://doi.org/10.1371/journal.pgen.1010729.g002

Zfp92 KO mice have altered body composition and elevated fasting blood glucose on a high fat diet

To investigate whether Zfp92 influences β-cell function in response to metabolic stress, we fed WT and Zfp92 KO mice a high fat diet (HFD) for 10 weeks and then assessed body composition and glucose homeostasis at 14–15 weeks of age. While WT and KO mice gained similar amounts of weight on HFD (Fig 3A), KO female mice had decreased lean body mass and increased body fat after 10 weeks on a HFD (Fig 3B and 3C). Glucose homeostasis measurements showed that KO male mice have slightly increased fasting blood glucose (Fig 3D), but no significant differences in fed blood glucose, insulin, or glucose tolerance were found (Fig 3E–3G, S2C and S2D Fig). Male KO mice had increased insulin sensitivity, suggesting that Zfp92 affects the function of one or more insulin target tissues (Fig 3H). Feeding of a HFD did not significantly affect β-cell area in mice lacking Zfp92 (Fig 3I and 3J). Combined, these data indicate that the absence of Zfp92 has only a mild effect on β-cell function, even during metabolic stress brought on by the feeding of a HFD. However, sex-specific differences in fat accumulation and insulin sensitivity were observed, suggesting a role for Zfp92 in insulin target tissues such as adipose, muscle and liver.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Effects of a high fat diet (HFD) on weight gain, body composition, glucose homeostasis and islet structure of Zfp92 KO mice.

A) Weight gain on HFD for 10 weeks was similar for both wild type (WT) and Zfp92 knockout (KO) animals (n = 7–9 for each sex and genotype). Lean body mass (B) and body fat (C) measurements show an increase in body fat content and a decrease in lean body mass in KO female mice on HFD (n = 7–8 for each sex and genotype). Fasted (D) and fed (E) blood glucose concentration measurements. KO male mice have higher fasting blood glucose than WT mice on HFD (n = 7,8 for each sex and genotype). F) Random fed plasma insulin levels (n = 4, male mice). G) Intraperitoneal glucose tolerance test (GTT) results are presented as the area under the curve (AUC) measurements for S2C Fig (n = 8 for each sex and genotype). H) Intraperitoneal insulin tolerance test (ITT) results are presented as the area under the curve (AUC) measurements for S2D Fig (n = 8 for each sex and genotype). KO male mice have higher insulin sensitivity than WT mice on HFD. I) Quantification of insulin-positive β-cell areas in KO and WT islets on HFD (n = 4, males at 14–15 weeks). J) Immunofluorescence staining for insulin (red) of pancreatic islet tissues from HFD-fed WT and Zfp92 KO mice. Nuclei are stained with DAPI (blue). Scale bar: 100 microns. B-J) All mice are 14–15 weeks old. Statistical significance between KO and WT samples is indicated by black brackets, and between sexes by gray brackets. Error bars: ± SEM. **p≤0.01; *p≤0.05. p-value determined by ANOVA or by unpaired t-test (panels F, K).

https://doi.org/10.1371/journal.pgen.1010729.g003

Transcriptional profiling of Zfp92 KO islets reveals an increase in Capn11 and a decrease in Mafb expression

To determine whether Zfp92 influences gene expression in mouse islets, we performed RNA-Seq of male WT and KO islet samples. Differential expression analysis showed that the expression of only 30 genes was significantly affected (p_adj<0.05), with 9 genes being upregulated and 21 genes downregulated in Zfp92 KOs (Fig 4A and 4B and S2 Table). Notably, there was a very pronounced increase in expression of Capn11 gene (Log₂FoldChange = 4.39), a gene that encodes a calcium-dependent protease calpain 11 that is usually expressed only specifically in testis [33]. Zfp92 was also increased in KO samples suggesting that ZFP92 may repress its own expression. The downregulated genes included the transcriptional regulators Mafb, Mamld1, Cited2, and Tref1, the former of which is known to be enriched in pancreatic α-cells in adult islets where it regulates the expression of pancreatic hormone glucagon (Gcg) [34]. To validate RNA-seq findings, we performed RT-qPCR on islet RNA and confirmed that the expression of Capn11 was increased more than 200 times, and that expression of Zfp92 is significantly upregulated and Mafb is downregulated in samples from KO mice (Fig 4C). While expression of Gcg was also decreased, and similar to the RNA-seq result, the changes were not significant. With the exception of Capn11, Mafb, and a few other genes, these results indicate that Zfp92 contributes in only a modest way to gene expression in adult islets.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. RNA-seq analysis of islets from Zfp92 KOs shows an increase in Capn11 and a decrease in Mafb gene expression.

A) Volcano plot showing differentially expressed genes (Log₂Fold Change (FC) over p_adj-value) in Zfp92 KO vs WT islets from 15 week-old male mice (N = 4). The top 10 differentially expressed genes (based on p_adj-value) are indicated by names and total numbers of downregulated (green) and upregulated (red) genes are provided (p_adj<0.05). B) Heat map showing clustering of the top variant genes (based on p-value) dysregulated in Zfp92 KO vs WT islets. C. RT-qPCR analysis of RNA expression for select genes in islet RNA samples collected from 15 week-old male mice. Zfp92 wild type (WT), and knockout (KO) samples. N = 4, error bars: ± SEM. ***p≤0.001; *p≤0.05. p-value is determined by an unpaired t-test.

https://doi.org/10.1371/journal.pgen.1010729.g004

Blood glucose and Mafb expression are decreased in Zfp92 KO mice at P1

In mice, the transcription factor Mafb is critical for the terminal differentiation of β cells during development through regulation of key β-cell genes. After birth, the function of Mafb in β-cells is largely replaced by Mafa, a closely related transcription factor [35]. Indeed, pancreas-specific Mafb KO mice display deficiencies in β-cell development and associated increased blood glucose levels at birth, but these defects are resolved by three weeks of age due to increased Mafa expression [36]. To determine whether the absence of Zfp92 and related decrease in Mafb expression affects early endocrine development, we analyzed blood glucose levels and gene expression in the pancreata of postnatal mice at P1 (postnatal day 1) (Fig 5A and 5B). We observed an increase in blood glucose concentration and a decrease in expression of Mafb and insulin (Ins1) at P1 in KO mouse samples (Fig 5A and 5B). These results indicate that pancreatic β-cell development is compromised in the absence of Zfp92 due to its effect on Mafb expression. As in adult islets, there was also a very strong increase in Capn11 expression in one day old Zfp92 KO neonates (Fig 5B). To confirm that ZFP92 regulates Mafb, we overexpressed Zfp92 in MIN6 cells, a mouse β-cell line, and observed a strong increase in Mafb expression. In addition, the expression of Ins1 and Slc2a2, two Mafb target genes [37], was also increased. These data all point to ZFP92 functioning as a positive regulator of Mafb expression.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Zfp92 KO mice have increased blood glucose and decreased Mafb and insulin expression at P1.

A) Random fed blood glucose concentration measurements in wild type (WT) and Zfp92 knockout (KO) animals show increased blood glucose levels in KO pups at P1 (N = 11,12). B) RT-qPCR analysis of mRNA expression for select genes in pancreatic RNA samples collected from newborn pups (P1). Zfp92 wild type (WT), and knockout (KO) samples (N = 5). C) RT-qPCR analysis of mRNA expression for select genes in MIN6 mouse β-cell line that was infected with lentivirus to overexpress Zfp92 (Zfp92-OE) and control cells (WT) (N = 6). The samples where Zfp92 was increased >10 times are shaded light blue. Error bars: ± SEM. **p≤0.01; *p≤0.05. p-value is determined by an unpaired t-test.

https://doi.org/10.1371/journal.pgen.1010729.g005

De-repression of an IAPez transposable retroelement in intron 3 of Capn11 in the absence of Zfp92

The striking increase in Capn11 gene expression in Zfp92 KOs prompted us to examine the alignment of RNA-seq reads in relation to the Capn11 gene. We observed an increase in mRNA reads in Zfp92 KO samples that maps to a location more than 10 kb downstream from the Capn11 transcription start site, with exon alignments starting only from exon 4 (Fig 6). Notably, there is an increase in expressed reads present in intron 3 preceding exon 4. Examination of RepeatMasker annotation revealed that these reads align with an IAPez TE consisting of IAPLTR1-Mm1 and IAPez-int repeating elements (ERVK family; LTR class) and adjacent small SINE repetitive elements (B1_mus2 and B1Mus_1, B3A, PB1D10; B1/Alu and B2 families; SINE class). Intracisternal A particles (IAPs) are LTR-containing ERV sequences belonging to an active class of TEs that can compromise genomic integrity and affect genome regulation by providing cis-regulatory modules [38]. Considering that many KRAB-ZFPs are linked to repression of TEs, these data suggest that ZFP92 functions to repress TEs within the intron 3 of Capn11 gene on chromosome 17 in mice. In the absence of Zfp92, the IAPez retroelement is re-activated and this, in turn, results in the marked expression of downstream Capn11 gene sequences. Since our qPCR primers for Capn11 were designed to amplify exons 4 and 5, we were able to detect a strong increase in Capn11 expression in Zfp92 KO mice (Figs 4C and 5B). RT-qPCR with primers designed to amplify exons 1–2 of Capn11 gene did not yield any product in islet samples, confirming that Capn11 gene is activated only downstream of IAPez insertion. To investigate whether IAPez-driven Capn11 expression is de-repressed in other tissues besides pancreatic islets, we performed RT-qPCR on RNA from WT and KO samples from whole pancreas, brain, lung, muscle, intestine testis, and liver (S3 Fig). These results show that while IAPez-driven Capn11 expression is increased to varying degrees in multiple tissues, the largest increase is in pancreatic islets, indicating that Zfp92 represses Capn11 IAPez activity in a cell type-specific manner. Further, to assess changes in global IAPez expression, we designed primers along different regions of IAPez from Capn11 locus (S4A Fig). RT-qPCR analysis with these primers on islet RNA samples did not reveal any differences in expression between WT and KO samples (S4B Fig), suggesting that ZFP92 represses IAPez in a genomic region-specific manner.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. IAPez ERVK LTR retroelement in Capn11 gene intron 3 is de-repressed in Zfp92 KO islets.

UCSC Genome Browser view of Capn11 gene region on the mouse chromosome 17 (GRCm38/mm10 assembly). Alignments of RNA-seq sequenced reads for the representative WT (blue) and Zfp92 KO (red) islet samples are displayed by density graphs and show a strong increase in expressed reads in intron 3 of Capn11 gene (encoded on the opposite DNA strand) and in further downstream exons of Capn11. Full view of RepeatMasker tracks shows that this region is occupied by IAPez ERV consisting of IAPLTR1, IAPez-int repeat elements (ERVK family, LTR class) and adjacent B1_Mus1 Alu-repeat SINE element that appear to be expressed and therefore de-repressed in Zfp92 KO islets. Capn11 gene exons are shaded in light gray, SINE element—light orange, IAPez-int element—light green.

https://doi.org/10.1371/journal.pgen.1010729.g006

Zfp92 KO-activated IAPez LTR and B1/Alu SINE retroelements form a fusion gene transcript with Capn11 gene

To further define how re-activation of an IAPez TE in Zfp92 KO mice drives the expression of Capn11 mRNA, we performed rapid amplification of 5’ cDNA ends (5’-RACE) on RNA from WT and KO islet samples using reverse primers in exon 5 of Capn11 gene (Fig 7A). While the WT samples did not yield specific PCR products, the KO sample yielded a 1538 bp product that contained a portion of IAPez and immediately adjacent B1/Alu SINE transposable elements that are then spliced to the exon 4 and 5 of Capn11 gene (Fig 7A). These data indicate that ZFP92 represses these TEs in Capn11 locus and, in the absence of this KRAB-ZFP, the IAPez element is activated and acts as a novel promoter region that drives the expression of a fusion TE-Capn11 transcript. The full-length TE-Capn11 transcript contains an open reading frame for an N-terminally truncated 481 aa form of CAPN11 protein (Fig 7B), however, we were unable to detect any specific protein bands by Western blots with antibodies raised against human CAPN11 gene (S5 Fig). These data suggest that ZFP92 binds to the TEs in intron 3 of Capn11 gene, producing a repressive chromatin environment that blocks IAPez expression. In the absence of Zfp92, the IAPez is de-repressed and serves as an alternative promoter that drives transcription of a novel IAPez-Capn11 transcript.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. De-repressed IAPez LTR retroelement sequences drive the expression of a fusion TE-Capn11 transcript in Zfp92 KOs.

A) 5’-RACE analysis of mRNA from Zfp92 KO islet sample revealed the production of a TE-Capn11 fusion transcript when the IAPez ERVK LTR element is derepressed. The panel shows a schematic representation of Capn11 gene (blue) and a relative location of B1 SINE (light green) and IAPez ERVK LTR (tan) TEs within the Capn11 intron 3, location of 5’ RACE primers (blue arrowheads), schematic, and the sequence of obtained 5’RACE product that represents fusion transcript. The IAPez-derived sequence is highlighted in tan, B1 SINE-derived sequence–in light green and a Capn11 downstream exon sequence—in blue. ZFP92 binding sequence is in red font. B) Structure of the full-length CAPN11 and N-terminally truncated CAPN11 protein that is encoded by the open reading frame within the TE-Capn11 fusion transcript.

https://doi.org/10.1371/journal.pgen.1010729.g007

Dysregulation of LINEs and LTR elements in Zfp92 KO islets

To determine whether Zfp92 influences the expression of additional TE elements we re-analyzed islet RNA-seq data by using the TE database and randomly assigning multi-mapped reads. To quantify the findings, we performed differential expression (DE) analysis both on a global and individual chromosome basis. The global DE analysis suggests that expression of 11 TEs is altered (p_adj<0.1). Among them, RLTR44B showed the largest increase (S6A Fig and S3 Table). However, similar to our qPCR results, there is no change in expression of IAPez TEs at a global level. Considering the very few changes in TE expression and that TE dysregulation might be locus dependent (like the IAPez in Capn11 locus) we also performed DE analysis at the chromosomal level. These results revealed significant changes in expression of 25 TEs, including ERVs and LINE elements, with the top dysregulated TEs all belonging to ERVK (or ERV2) family of LTR retrotransposons (Fig 8A and S3 Table). Specifically, reads attributable to RLTR44 ERVK-related sequences (RLTR44C, RLTR44-int, RLTR44B) on chromosomes 9 and 2 showed high increases. At the same time, most downregulated TEs were also from the ERVK family including RLTR19C and IAPEy-int elements on chromosomes Y and 4, respectively. Manual inspection of the read alignments for selected individual RLTR44 TEs on chromosome 9 and RLTR19C TE on chromosome Y confirmed respective strong increase or decrease in mapped reads at unique locations (Fig 8B). However, many other retroelements of the same kind in other locations on the same or different chromosomes either did not have any expressed reads or did not show any visible change in expression (S6B Fig). These results indicate that in the absence of Zfp92, there is a genome-wide dysregulation of retroelements and that their expression varies both by the specific retroelement and by the genomic context of the retroelement.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. RNA-seq analysis against the TE database shows dysregulation of LINE and LTR retroelements in Zfp92 KO islets.

A) Volcano plot showing differentially expressed TEs in Zfp92 KO vs WT islets collected from 15 week-old male mice (N = 4). RNA-seq read alignment was done by using random assignment of multi-mapped reads to the TE database (RepeatMasker annotations) and differential expression analysis was done for each individual chromosome separately. The volcano plot only shows the TEs whose expression was significantly changed (FDR p_adj-value <0.1). TEs are labeled by name:family:class and chromosome (Ch) for which differential expression was identified. Top differentially expressed TEs with Log₂(Fold Change) higher than 2 are labeled in red. B) UCSC Genome Browser view of representative top upregulated (RLTR44C) and downregulated (RLTR19B) TEs on the mouse chromosomes 9 and Y, respectively (GRCm38/mm10 assembly). Alignments of RNA-seq sequenced reads for the representative WT (blue) and Zfp92 KO (red) islet samples are displayed by density graphs.

https://doi.org/10.1371/journal.pgen.1010729.g008

Genome-wide analysis of ZFP92 binding reveals that ZFP92 preferentially binds to B1/Alu SINE elements

To identify binding sites for ZFP92 we performed CUT&RUN (Cleavage Under Targets & Release Using Nuclease) experiments on MIN6 cells overexpressing HA-tagged ZFP92 protein. Next generation sequencing and data analysis of the resulting chromatin fragments revealed more than 500 ZFP92 binding peaks (q value<0.05). These peaks were further analyzed to determine whether they contain TEs, other genomic features, or ENCODE Candidate Cis-Regulatory Elements (cCREs) [39] (Fig 9A–9C and S4 Table). We found that 93% of the ZFP92-bound peaks contained one or more TEs with SINE elements being the most common (39%), followed by LINE elements (16%), LTR class (15%) of retroelements, and simple repeats (13%). Within the SINE element-bound peaks, the majority (53%) contained B1 type of Alu family of SINEs (Fig 9A). Enrichment analysis of specific TE classes and families in binding sites relative to their genomic distribution showed that SINE elements are the most enriched class (5.2 FoldChange (FC)), followed by DNA elements (2.2 FC) and LTR elements (1.2 FC), while LINE elements show negative enrichment (0.8 FC). Among the families of TEs, the most highly enriched families are B1/Alu (8.2 FC), B4 (4.6 FC) and MIR (3.1 FC) SINE elements (S4 Table). Notably, many of the ZFP92-bound peaks are located in gene-rich regions, with 40% of the peaks located within gene introns, 9% within exons, 6% within extended gene promoters, and 1% within 3’-UTRs (Fig 9B). Moreover, almost 13% of ZFP92-bound peaks contain distal enhancers, 5.5% proximal enhancers and 1% CTCF sites (Fig 9C). Importantly, one of the most enriched ZFP92 binding peaks is in the Capn11 gene intron 3, with the summit of ZFP92 binding peak being located over the B1/Alu SINE element adjacent to the IAPez element (Fig 9D). The peak extends further into IAPez element that becomes re-activated in the absence of ZFP92. qPCR of chromatin from independent ZFP92 CUT&RUN pulldowns with primer pairs specific to different regions under the ZFP92 peak showed the strongest enrichment for the Capn11 intron and B1 element, with less enrichment for IAPLTR, and no enrichment for the IAPez body (IAP1), confirming our chromatin sequencing data (Fig 9D and 9E).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9. Genome-wide analysis reveals ZFP92 binds preferentially to B1/Alu SINE elements.

A) Pie chart showing the percentage distribution of different TE classes within the ZFP92 binding peaks obtained by CUT&RUN analysis. 93% of peaks have TEs with SINE elements being the largest group. The inset pie chart shows the family distribution of ZFP92-bound SINEs, with Alu/B1 family being most represented. B) Pie chart showing the distribution of ZFP92 binding peaks within different genomic features. C) Pie chart showing the distribution of different ENCODE Candidate Cis-Regulatory Elements (cCREs) within the ZFP92 binding peaks. enhD, distal enhancer signature; enhP, proximal enhancer-like signature; prom, promoter-like signature; CTCF, CTCF-only; K4m3, DNase-H3K4me3. D) UCSC Genome Browser view of Capn11 locus showing ZFP92 binding peaks and RNA-seq reads in WT and Zfp92 KO samples (GRCm38/mm10 assembly). The main peak containing the B1 SINE element is highlighted in gray, with red lines denoting amplicons and primer pairs that were used for qPCR shown in (E). E) qPCR on CUT&RUN chromatin samples with the primers shown in (D) to different regions of Capn11 locus and TE elements found there. (N = 3) Error bars: ± SEM. **p≤0.01; *p≤0.05. p-value is determined by an unpaired t-test. F) ZFP92 binding motif predicted by MEME analysis of ZFP92 binding peaks. Under the ZFP92 motif are the two similar motifs (ZNF460 and ZNF135) that were identified by TomTom analysis from the MEME suite. G) Pie chart showing the distribution of ZFP92 binding motif within TE elements in ZFP92 peaks. The inset pie chart shows the family distribution of SINEs with FP92 motif, with Alu/B1 family being most represented.

https://doi.org/10.1371/journal.pgen.1010729.g009

We next used the MEME motif-discovery tool to analyze the ZFP92 peaks that revealed a 28 bp consensus ZFP92 binding sequence (Fig 9F). Comparison of the consensus ZFP92 binding motif with those predicted based on the protein sequence of ZFP92 [40] shows only minimal similarities within repeated ‘GAG’ sequences (S7A Fig) highlighting limitations of the latter approach. BLAST alignment search of ZFP92 motif against Dfam TE database [41] retrieved matches to a part of B1 Alu elements, further suggesting that these elements are the main target of ZFP92. Comparison of the ZFP92 binding motif against a database of known motifs revealed similarities to binding motifs for ZNP460 and ZNF135 (Fig 9F), two human KRAB-ZFPs that bind to SINE elements [24]. Moreover, 67% of ZFP92 binding motifs in ZFP92 peaks occur within SINEs, of which 76% are in B1/Alu elements, 17% in B4 elements and 7% in B2 elements (Fig 9G). The ZFP92 binding site is also found in 14% of LTR elements and 12% of LINE elements (Fig 9G and S7B Fig) indicating that SINE-derived sequences are present in some of these elements as well. Inspection of the RNA-seq tracks that span SINE element-specific ZFP92 peaks shows that expected de-repression of SINEs in Zfp92 KO samples does not lead to an increase in transcription of SINEs. Similarly, neighboring LINE and LTR elements are also often not de-repressed (S8A Fig). However, the absence of Zfp92 causes some LTR elements to be de-repressed in a context-dependent manner, such as in the Capn11 locus (Fig 9D) and a few other loci (S8B Fig). These observations confirm our RNA-seq analysis of TE expression highlighting the locus-dependent nature of TE dysregulation in the absence of Zfp92.

Combined, our chromatin binding data indicate that ZFP92 binds to multiple TE elements with preferential binding to B1/Alu SINE elements. In the absence of Zfp92, chromatin changes may lead to de-repression of select retroelements in a locus-dependent manner. The ZFP92 binding peaks and corresponding TEs are often found within the gene bodies and may contain regulatory elements, indicating that ZFP92 might function not only as a repressor of TEs but as a potential regulator of gene transcription.

ZFP92 binds within and affects the expression of Sox17 and Acacb genes

Strong ZFP92 binding peaks were identified within the Sox17 and Acacb genes (S4 Table). To determine the potential significance of these bindings we further investigated both genes. Sox17 encodes a transcription factor that is important for early endoderm, vascular and hematopoietic cell development [42]. While Sox17 is expressed in islets at low levels, RNA-seq did not show any difference in total mRNA expression between WT or KO islets (S2 Table). The ZFP92 binding peak in Sox17 is located within the exon 4 and an alternative second promoter region that was shown to drive the expression of the short form of Sox17 mRNA containing only the extended exon 4 and exon 5 [43] (Fig 10A and 10B). To determine whether binding of ZFP92 to this region may affect expression of different Sox17 mRNA forms, we performed RT-qPCR on mRNA from WT and Zfp92 KO lung and testes, tissues where Sox17 is expressed at high levels in adult animals. This experiment revealed an increase in expression of the short Sox17 form in Zfp92 KO tissues with no change in expression of the long Sox17 form that contains exons 1 through 5 (Fig 10C), suggesting that ZFP92 represses expression of the short form of Sox17. Although there are no TEs recognized by the Repeat Masker within the Sox17 ZFP92 binding peak, there is a stretch of sequence (GAGGCAGG) within the extended exon 4 that is part of the identified putative ZFP92 binding sequence. In addition, a second strong ZFP92 binding peak is located 128 kb upstream of the Sox17 gene (Fig 10D). This second peak is located within conserved sequences that may be involved in formation of regulatory chromatin loop with the ZFP92-bound alternative Sox17 promoter, further suggesting a yet-to-be-defined regulatory function for ZFP92.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 10. ZFP92 binds within the Sox17 locus and may regulate expression of the short form of Sox17 mRNA.

A) UCSC Genome Browser view of Sox17 locus showing ZFP92 CUT&RUN binding peak and RNA-seq reads for mRNA expression in WT and Zfp92 KO samples in Sox17 locus. The peak is highlighted in gray and covers the extended Sox17 exon 4, the boundary of which is indicated by the dashed blue line defined by the transcription start site for the short Sox17 isoform (TSS Sox17-1), and an alternative Sox17 promoter preceding it. TSS for Sox17 short (Sox17-1) and long (Sox17-2) isoforms are from the Promoters from the EPD track in the UCSC browser (GRCm38/mm10 assembly). B) qPCR on CUT&RUN chromatin samples with the primers for the Sox17 peak (extended Sox17 exon 4) shows strong enrichment at the locus. (N = 3) Error bars: ± SEM. **p≤0.01; *p≤0.05. p-value is determined by an unpaired t-test. C) RT-qPCR analysis of different Sox17 isoforms expression in testes and lung tissues of WT and Zfp92 KO mice shows an increased expression of the short Sox17 isoform. Primers specific to extended exon 4 amplify the short Sox17 isoform (short), exons 1,2 –long Sox17 isoform (long), and exons 4,5 –both isoforms (both). (N = 3) Error bars: ± SEM. **p≤0.01; *p≤0.05. p-value is determined by an unpaired t-test. D) UCSC Genome Browser view of Sox17 locus showing two distant ZFP92 binding peaks in Sox17 locus. The peaks are highlighted in gray, the distal peak is located -128kb upstream of the main peak within extended Sox17 exon 4 and lies within a conserved DNA sequence that contains a distal enhancer signature (cCRE track) (GRCm38/mm10 assembly).

https://doi.org/10.1371/journal.pgen.1010729.g010

Another strong ZFP92 binding peak was identified in an intron of Acacb (Fig 11A) that coincides with a PB1D10 B1/Alu SINE element, an overlapping MT2B2 LTR element, and an L1 LINE element (S7A Fig). While we did not observe significant differences in Acacb expression in islets (S2 Table), where it expressed at very low levels, there was an observable decrease in RNA-seq aligned reads for exons surrounding the ZFP92-bound region in the Zfp92 KO sample (Fig 11A). Acacb encodes acetyl coenzyme A carboxylase β, the rate-limiting enzyme in fatty acid β-oxidation, and a decrease in Acacb expression is associated with increased lipolysis and insulin sensitivity [44]. Since our physiological studies revealed sex-based differences in fat accumulation and insulin sensitivity in mice fed a HFD, we explored whether they could be due to changes in Acacb expression by analyzing mRNA from WT and KO male and female mice from three tissues associated with body fat metabolism and insulin sensitivity: white adipose, muscle, and liver. RT-qPCR with primers to the Acacb exons surrounding the ZFP92 peak showed a decrease in expression of Acacb mRNA in adipose and muscle tissues from male Zfp92 KO but not in female KO tissues (Fig 11B). These results indicate that in the absence Zfp92 there is a decrease in expression of Acacb in males that may have contributed to the observed increased insulin sensitivity in Zfp92 KO male mice and higher fat accumulation in KO female mice on a HFD.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 11. ZFP92 binds a B1/Alu SINE element within the Acacb locus and affects the expression of Acacb mRNA in adipose and muscle tissues of male mice.

A) UCSC Genome Browser view of Acacb locus showing ZFP92 CUT&RUN binding peak and RNA-seq reads for mRNA expression in WT and Zfp92 KO samples in Acacb locus. The peak is highlighted in gray and contains B1/Alu SINE, as well as LINE and LTR elements (see S7 Fig for a close up). The expression levels of exons surrounding the ZFP92 binding site (arrows) appear to be reduced. B) RT-qPCR analysis of Acacb mRNA expression in adipose, muscle and liver tissues from male and female WT and Zfp92 KO mice. Decreased expression of Acacb is observed in Zfp92 KO male adipose and muscle tissues. (N = 3,4) Error bars: ± SEM. **p≤0.01; *p≤0.05. p-value is determined by an unpaired t-test.

https://doi.org/10.1371/journal.pgen.1010729.g011

ZFP92 predominately binds to SINE elements and regulates TE activity

Our chromatin binding and RNA expression data indicate that the main function of mouse ZFP92 is to regulate the activity of specific TEs consistent with an emerging role for KRAB domain containing ZFPs. Most of the binding sites identified for ZFP92 coincide with TEs, with a large proportion of these being B1/Alu type of SINE elements. Mouse B1 SINE elements are ~140 bp in length and like human Alu elements are derived from the signal recognition particle 7SL and contain a promoter for RNA polymerase III-dependent transcription [47]. SINE elements require LINE-encoded proteins for their mobilization, although the SINE retrotransposition rate is estimated to be higher than that of LINEs with some de novo insertions being potentially contributing to disease [48]. There are approximately 500,000 B1/Alu SINEs in the mouse genome that comprise approximately 2.7% of mouse DNA. The majority of B1 SINEs and other TEs are silenced through chromatin repression and DNA methylation by DNA methyltransferases (DNMTs) [49, 50]. However, while KRAB-ZFPs play a vitally important role in repressing TEs, few have been identified that bind to SINE elements. Consistent with ZFP92 preferentially binding to B1/Alu SINE elements, the predicted ZFP92 binding motif is a 28 bp sequence that is highly similar to B1/Alu SINEs. While shorter in length, the predicted binding motifs for ZNF460 and ZNF135, two other KRAB-ZFPs that bind SINEs [24], are very similar to that for ZFP92. Our analysis also revealed that while ZFP92 binds predominately in SINE elements, it sometimes also binds to LINE and LTR TEs, indicating that SINE sequence fragments are present in other TEs and, possibly, other genomic features [51]. Recent studies indicate that SINE sequences within the genome can play important functions such as act as enhancers, promote chromatin compartmentalization, and regulate chromatin accessibility through SINE-KRAB-ZFP interactions [52–54].

Interestingly, we did not observe a global increase in SINE expression in Zfp92 KO tissues, suggesting that the absence of ZFP92 alone is not enough to activate the transcription of silenced SINEs. While it is possible that the methods used to construct the library for RNAseq did not capture small SINE transcripts, we did observe the altered expression of several LINE, LTR elements, and genes in the Zfp92 KO mice, indicating significant genomic dysregulation in the absence of ZFP92. The most striking example of such dysregulation is the very strong activation of a novel TE-Capn11 fusion transcript in Zfp92 KO islets due to the activation of an IAPez retroelement located next to the ZFP92-bound B1 SINE element in the 3^rd intron of the Capn11 gene. In the absence of Zfp92, the IAPez ERV in this location is activated and serves as an alternative promoter that drives the robust expression of the TE-Capn11 transcript that contains IAPez and SINE sequences that are spliced to the downstream Capn11 exons. This surprising finding provides a very stark example of how de-repression of TEs can lead to creation and strong expression of a brand-new transcript and, potentially, impact cell function. It also indicates that ZFP92 is indeed involved in repressive complexes and its removal leads to chromatin opening and activation of nearby regulatory elements. Interestingly, similar activation of IAPez expression within Capn11 locus was observed in DNA methyltransferase Dnmt1 KO mice [55], further suggesting that ZFP92 is a part of TE repressive complex.

Although the novel TE-Capn11 fusion transcript contains an open reading frame that encodes an N-terminally truncated form of calpain 11, a calcium-dependent protease normally expressed only in adult testis in mice [56], we did not detect any protein product in islets using available antibodies. Despite this, we cannot rule out that the protein can potentially be made in islets or other tissues. While the predicted protein would lack only part of the N-terminal protease domain, so might not be catalytically active, it would still be expected to contain five C-terminal calcium-binding EF-hand domains (Fig 7B). EF-hand domain-containing proteins often act as calcium buffers which control the level of free Ca²⁺ ions in the cytoplasm [57]. It is possible, therefore, that the TE-Capn11 -transcript-derived CASPN11 protein, even if truncated, could influence function of different tissues. In Zfp92 KO mice, besides being overexpressed in islets, Capn11 fusion transcript is highly expressed in brain, intestine, testis, and liver, suggesting that Zfp92 may regulate the Capn11 IAPez in a tissue-specific manner. Even though the level of Zfp92 expression was low in some of these adult tissues, it may have been expressed in progenitor tissues at some point during development. In the absence of Zfp92, changes in the chromatin environment and de-repression of IAPez during development could lead to continuous overexpression of TE-Capn11 transcript at later stages in these tissues.

Global analysis of TE expression in Zfp92 KOs showed varied increases and decreases in several types of LINE and LTR elements. We also observed that the same types of elements were dysregulated in one position and unchanged in others. Similarly, while some changed elements were in proximity to ZFP92 binding sites, others were not. This suggests that regulation of TEs by ZFP92 is influenced by genomic context, the state of chromatin in the locus, distal chromatin interactions and by chromatin modifying complexes that may interact with this KRAB-ZFP. Indeed, a recent study reports that chromatin modifications affect TE expression in a complex way where not only repressive marks, but widespread marking of TEs by bivalent or activating marks, were shown to influence TE expression and chromatin accessibility. Furthermore, the effects of chromatin modifications are sequence- and context-specific, with different chromatin modifiers regulating the expression and chromatin accessibility of specific subsets of TEs [58–60]. For example, different LTR retrotransposons can also be modulated by different chromatin modifying complexes. RLTR44 is upregulated in the absence of Zfp92 and is repressed by the polycomb complex [61] suggesting that ZFP92 associates with polycomb or other repressive complexes to repress RTLR44. On the other hand, several IAPey retroelements are downregulated in the absence of Zfp92. IAPeys are enriched on the Y chromosome and are active in male germline [62,63]. While KRAB-ZFPs are generally considered to be transcriptional repressors, some KRAB-ZFPs activate gene transcription through interaction with activating complexes [64,65]. In any case, our studies of Zfp92 add to the growing evidence that IAPs and other LTR transposons act as cis-modifiers of gene expression and that KRAB-ZFPs regulate them in locus-, tissue-, species-, and strain-specific manners [25,66–69].

Role of Zfp92 in islets

The expression of mouse Zfp92 increases during development as endocrine progenitor cells are converted into nascent β-cells, and then is maintained in adult β-cells. Zfp92 KO mice fed a regular diet do not display major abnormalities in pancreatic islet structure, insulin secretion, or β-cell effects on glucose homeostasis. However, we did observe a decrease in fasting blood glucose levels in Zfp92 KO males which may be due to reduced levels of glucagon associated with a decrease expression of Mafb, a transcription factor that is important for islet maturation and function. Indeed, among the very few changes observed in the transcriptome of adult Zfp92 KO islets, there was a verified decrease in Mafb gene expression. The absence of an overt β-cell phenotype in adult Zfp92 KO mice is similar to pancreas-specific Mafb KO mice where only a decrease of glucagon secretion is observed in adult mice [36]. Mafb is highly expressed in progenitor α- and β-cells, is required for β-cell development, becomes largely replaced by Mafa in adult mouse β-cells, and continues to be expressed and regulate glucagon expression in adult α-cells [34,37,70]. Consistent with the role of Mafb during β-cell development, we showed that in the absence of Zfp92, newborn mice have decreased expression of Mafb and Ins1 genes that is accompanied by increased blood glucose levels. Additionally, Zfp92 KO mice exhibit increased fasting blood glucose levels when fed a HFD for 12 weeks, indicating a modest impairment of β-cell function under metabolic stress. Together, these results indicate that Zfp92 contributes to Mafb regulation, both during β-cell development and in adult islets.

The mechanism whereby ZFP92 protein promotes the expression of Mafb is not clear. We observed only weak ZFP92 binding peaks around Mafb promoter although there are several sites of stronger ZFP92 binding within the 20 kb of DNA surrounding Mafb gene that either bind to a TE element or to a partially conserved binding motif sequence (S10 Fig). It is also possible that there is a de-regulation of TEs that are more distal to Mafb gene that are important for the regulation of Mafb in islets. Indeed, several studies have shown that TEs can serve as enhancers, are associated with open chromatin regions, contain transcription factor binding sites, and are involved in cell-type specific chromatin domain organization and gene regulation [71–73]. Alternately, the dysregulation of other genes in endocrine cells that are important for Mafb expression could be affected.

Role of Zfp92 in other tissues

While our study began with a focus on pancreatic islets, our physiological studies also revealed a role for Zfp92 in adipose tissue of male mice. Male KO mice on a normal diet grew at a slower rate and exhibited increased insulin sensitivity and body fat content when on a HFD compared to female KO mice. An increase in total body fat is one of the two significant phenotypes reported by the International Mouse Phenotyping Consortium (IMPC) for an independently derived Zfp92 KO allele on different mouse strain (C57Bl/6NJ) [74]. The second significant phenotype described for IMPC Zfp92 KO mice was a decreased bone mineral content, something we did not study.

The sex-specific differences in body fat accumulation and insulin sensitivity in Zfp92 KO mice on a HFD are likely due to sex-specific changes in Acacb gene expression. ZFP92 binds to an array of TE elements that includes B1/Alu SINE element within the intron of Acacb gene and deletion of Zfp92 leads to a male-specific reduction in the expression of Acacb in adipose and muscle tissue. By catalyzing synthesis of malonyl-CoA, an inhibitor of rate-limiting step in fatty acid uptake and oxidation, ACACB is thought to be a key regulator of fatty acid oxidation [75,76]. Global deletion of Acacb in mice leads to decreased fat accumulation and increased insulin sensitivity due to increased fat metabolism in muscle and adipose tissues [77, 78]. The Zfp92-dependent reduction of Acacb only in males in our study may explain the greater fat accumulation in KO females and increased insulin sensitivity in male KO mice fed a HFD. The mechanisms responsible for the sex-specific regulation of Acacb are not clear, although Acacb was previously shown to be more highly expressed in female liver [79] and hypothalamus [80]. Interestingly, in humans a strong QTL influencing increased percent body fat specifically in women is located on chromosome 12q (12q24.3–12q24.32) near ACACB [81], further suggesting sex-specific differences in the expression or function of this gene.

In this study, we also showed that ZFP92 is involved in regulation of expression of essential transcription factor Sox17 [82]. We have previously shown that during early development Sox17 is expressed in two distinct progenitor cell populations, endodermal and hemogenic endothelial cells, that preferentially express short or long Sox17 RNA forms driven by two different promoters [43,44]. We discovered that ZFP92 binds to the promoter region for the short RNA form that, in the absence of Zfp92, is upregulated in adult lung and testis tissues. We also identified a distal site upstream of Sox17 that is bound by ZFP92. Additional studies are necessary to determine whether either of these two binding sites has effects on the development of endodermal, vascular, or other lineages that express Sox17.

Finally, while we only studied gene and TE expression differences for selected genes and tissues, it is possible that genome-wide chromatin dysregulation due to the lack of Zfp92 may impact other genes, tissues, and developmental processes. However, the observed mild phenotype of Zfp92 KO mice indicates that there’s a redundancy among KRAB-ZFPs targeting different TEs. Indeed, mice in which an entire cluster of KRAB-ZFPs is deleted are viable and relatively normal [23].

Ethics statement

Animal procedures were conducted in accordance with the ethical guidelines of the National Institute of Health (NIH) and approved by the Vanderbilt Institutional Animal Care and Use Committee. This study did not include research with human subjects.

Mouse lines and husbandry

Zfp92 KO mice were derived using CRISPR/Cas as previously described [31]. Briefly, guide RNAs (gRNAs) were microinjected into Cas9-expressing embryos heterozygous B6J/129(Cg)-Gt(ROSA)26Sor^{tm1.1(CAG-cas9}*^{, -EGFP)Fezh/J} mice (JAX 026179) by the Vanderbilt Genome Editing Resource. Founder animals were then backcrossed to C57Bl/6J mice for 7 generations, segregating the Cas9-expressing ROSA26 allele in the process. Pups were genotyped by PCR using primers surrounding the deletion site (S1 Table).

Mouse body weight and composition

Male and female mice were weighed weekly between 4 to 14 weeks of age. Mouse body composition was measured at 14–15 weeks by using a nuclear magnetic resonance spectroscopy imaging machine (Bruker Instruments, The Woodlands, TX) at Vanderbilt Mouse Metabolic Phenotyping Center.

Glucose homeostasis

Blood glucose concentrations were determined in tail blood samples by using a BD Logic glucometer. Intraperitoneal glucose tolerance tests (GTT) were performed following a 16-hour overnight fast. Mice were administered D-glucose (2 mg/g body mass) by intraperitoneal injection and blood glucose concentrations were measured at 0, 15, 30, 60, and 120 minutes. Insulin tolerance testing (ITT) was performed following a 4-hour morning fast by giving 0.1 U/mL of insulin (Humulin R (Eli Lilly) in DPBS (Thermo Fisher)) by intraperitoneal injection and blood glucose concentrations were measured at 0, 15, 30, 60, and 120 minutes. Plasma insulin concentrations were determined in triplicate by radioimmunoassay (Millipore, PI-13K) by the Vanderbilt Hormone Assay and Analytical Services Core.

Islet isolations

Mouse islets were isolated following an injection of 0.6 mg/mL Collagenase P (Roche) into the pancreatic bile duct. Partially dissociated tissue was fractionated using a Histopaque-1077 (Sigma) gradient followed by hand-picking of islets. Islets were isolated by injection of 0.6 mg/mL Collagenase P (Roche) into the pancreatic bile duct, and partially dissociated tissue was fractionated using a Histopaque-1077 (Sigma) gradient followed by hand-picking of islets. Islet isolations were performed by the Vanderbilt Islet Procurement and Analysis Core.

RT-qPCR

RNA was isolated from tissues and whole islets from 14–15-week-old mice, from pancreata removed from one day old animals (P1), and from MIN6 cells using Maxwell 16 LEV SimplyRNA Tissue Kits (Promega, TM351). After extraction, RNasin (40 U/uL, Promega) was added to the RNA samples before storage at -80°C. Reverse transcription was done using a High Capacity cDNA Reverse Transcription Kit (Thermo Fisher). 2 ng of cDNA were used in real-time qPCR with Power SYBR Green PCR master mix (Thermo Fisher) using CFX96 Real-Time PCR system (Bio-Rad). Relative expression was determined by the ΔCt method by normalizing to the expression of Actb or Hprt. Primers are listed in S1 Table.

RNA-seq

The quality of RNA samples was assessed using an Agilent 2100 Bioanalyzer. Only those samples with an RNA integrity number (RIN) 7 or above were used for sequencing. cDNA libraries were constructed using the Low Input Library Prep Kit pipeline. An Illumina NovaSeq6000 instrument at Novogene Corp was used to produce paired-end, 150 nucleotide reads for each RNA sample. Paired-end sequencing produced approximately 55 million reads per sample, which were processed utilizing Trim Galore 0.5.0 (which relies on Cutadapt 1.18) to remove adapter sequences and pairs that were either shorter than 20 bp or that had Phred scores less than 20. The Spliced Transcripts Alignment to a Reference (STAR v2.6.0c) application [83] was used to perform sequence alignments to the mm10 (GRCm38) mouse genome reference and GENCODE comprehensive gene annotations (Release M17). STAR’s two-pass mapping approach was used to increase the detection of reads mapping to novel junctions identified during the first mapping pass. Overall, approximately 85–87% of the raw sequencing reads were uniquely mapped to genomic sites. For additional sample-level quality control analysis and downstream pairwise comparisons, DESeq2 was employed [84]. For the analysis of TEs, the parameters for STAR were modified to account for the more complex TE content in mice by allowing a higher number of multi-mapped reads [85] (S1 Method). The alignments were then analyzed using the count feature of TE transcripts [86]. In addition, TE annotations were extracted from the UCSC Genome Browser’s RepeatMasker-generated RMSK track [87] which enabled the screening of DNA sequences enriched in interspersed repeats and low complexity DNA sequences. While alignments were run on the entire genome, the TE mapping and read counting were performed both on the whole genome and on a per-chromosome basis. DESEq2 was used for downstream pairwise comparisons. All data processing was performed at the Advanced Computing Center for Research and Education (ACCRE) at Vanderbilt University.

Zfp92 overexpression studies

MIN6 mouse insulinoma cells (C0018008, AddexBio) were cultured in advanced DMEM medium (C0003-04, AddexBio) supplemented with 15% FBS (Atlanta Biologicals), 1% penicillin /streptomycin and 5 μM of β-mercaptoethanol (both Thermo Fisher) at 37°C, with 5% CO₂ and 95% humidity. Lentiviral vectors submitted by Didier Trono’s lab were obtained from Addgene (Addgene # 12257, #12260, #12259). To overexpress Zfp92, a full-size mouse Zfp92 cDNA containing N-terminal HA-tag was generated by RT-PCR, cloned into the lentivirus expression vector, then co-transfected with packaging vectors into HEK293T cells using PolyFect reagent (Qiagen). Viral supernatants were harvested 24-48h post-transfection, filtered through a 0.45μm filter, and titered using Lenti-X qRT-PCR Titration Kit (Takara). For the experiment, MIN-6 cells were seeded in 24-well plates at 100,000 cells per well and infected with lentivirus at MOI ~2 then after 72 hours cells were lyzed and RNA was isolated. A similar lentivirus overexpression protocol was used for CUT&RUN experiments.

Identification of ZFP92 binding sites

HA-ZFP92 lentivirus-infected MIN6 cells were trypsinized and washed 2 times with 1%BSA in PBS solution (150 x 10³ cells were used per pull-down). The CUT&RUN (Cleavage Under Targets and Release Using Nuclease) assay was performed by using the CUT&RUN Assay KIT (#86652 Cell Signaling Technology) following the manufacturer’s protocol. For anti-HA-ZFP92 pulldowns from MIN6 cells, rabbit anti-HA antibodies (#3724) and normal rabbit control IgGs from the kit were used (all antibodies are from Cell Signaling). Briefly, buffer-washed cells were bound to Concanavalin A-coated magnetic beads, permeabilized with 0.05% digitonin buffer, incubated with antibodies at 4°C overnight, washed twice with digitonin buffer, and incubated with Protein A-MNase (pA-MN) for 1 hour at 4°C. After two washes, cell-bound beads were incubated for 30 min at 4°C in 2 mM CaCl₂ for pA-MN digestion, then incubated with RNAse-containing Stop Buffer (30 min. at 37°C) to release DNA. DNA was further purified using spin columns (14209S, Cell Signaling) and directly used for library prep or qPCR. For sequencing, DNA was processed through the ChIP-seq library prep and Illumina NovaSeq6000 PE150 next-generation sequencing pipeline at Novogene Corporation. The sequencing reads were processed with nf-core/cutandrun version 3.0 [88,89], which is an analysis pipeline for CUT&RUN experiments built using Nextflow for portability and reproducibility. FastQC was used for sequencing quality control reporting and Trim Galore for adapter trimming and to remove low quality reads. Bowtie2 [90] was used for DNA alignment against the UCSC mm10 mouse genome reference and gene annotations. Samtools [91] was then used to filter on quality, sort, and index alignments, followed by duplicate removal by Picard [92]. Peak calling was performed with MACS2 [93]. To quantify the frequency of classes and families of repeat elements across the genome, we calculated the total genomic length of each element from the coordinate data available from the RMSK track [87] and aggregated these either by class or family, then divided by the genome length to obtain a coverage ratio. Then, to determine enrichment of such elements within the CUT&RUN peaks, we matched repeat elements to the coordinate regions of the peaks and then aggregated these to determine enrichment frequency. The MEME,TomTom and FIMO tools within the MEME suit were used for the de novo prediction of the ZFP92 binding motif within ZFP92-bound peaks, comparison of the ZFP92 motif against a database of known motifs and finding occurrences of the ZFP92 binding motif within the peak TEs, respectively [94].

Immunofluorescent staining

Whole pancreata were fixed for 4 hours in 4% paraformaldehyde, incubated overnight at 4°C in 30% sucrose, embedded in OCT compound (Tissue Tek), frozen on dry ice, and sectioned at a depth of 8 μm. The staining was performed following standard procedure as previously described [95]. Two primary antibodies were used, a guinea pig anti-insulin (ThermoFisher, 1:1000) and a rabbit anti-glucagon (Linco, 1:1000). Two secondary antibodies used were a donkey anti-rabbit IgG Alexa-488 conjugated and donkey anti-guinea pig IgG Alexa-555 conjugated (ThermoFisher,1:1000). After antibody staining, slides were mounted with Invitrogen ProLong Gold antifade reagent with DAPI (ThermoFisher). Images were acquired using Aperio ScanScope CS imaging system. Quantifications of cell area were done on evenly spaced pancreatic sections at 150 μm apart. Relative β-cell (insulin⁺) and α-cell (glucagon⁺) areas were calculated by dividing the area of hormone⁺ cells by that of the whole pancreas (DAPI⁺). The images shown are representative of the phenotype observed in at least three different animals per genotype.

Western blot analysis

Tissues were lysed in RIPA buffer (10 mM Tris-HCl, pH 7.5; 140 mM NaCl; 1 mM EDTA; 1% Nonidet P-40; 0.1% sodium deoxycholate; 0.1% SDS) containing protease inhibitors (300 μg/ml phenylmethylsulfonyl fluoride [PMSF], 1X protease inhibitor cocktail (Sigma)). Samples (20 μg of total protein) in 1X Laemmli sample buffer were resolved on a gradient 12% SDS-PAGE and transferred to polyvinylidene difluoride (PVDF) membranes (Millipore). Membranes were blocked with PBST (PBS; 0.02% Tween 20, pH 7.5) containing 5% non-fat dry milk and incubated with primary antibodies overnight. Primary antibodies used were mouse anti-CAPN11 (1:200; MA5-21265, ThermoFisher), and rabbit anti-GAPDH (1:1000; Cell Signaling). After washing with PBST, the membranes were then incubated with a secondary anti-mouse or anti-rabbit HRP-conjugated antibodies (Cell Signaling, 1:2000), and detected with SuperSignal West Dura Extended Duration Substrate chemiluminescence kit (ThermoFisher).