ZNF143 binds DNA and stimulates transcripstion initiation to activate and repress direct target genes

Transcription factors bind to sequence motifs and act as activators or repressors. Transcription factors interface with a constellation of accessory cofactors to regulate distinct mechanistic steps to regulate transcription. We rapidly degraded the essential and ubiquitously expressed transcription factor ZNF143 to determine its function in the transcription cycle. ZNF143 facilitates RNA Polymerase initiation and activates gene expression. ZNF143 binds the promoter of nearly all its activated target genes. ZNF143 also binds near the site of genic transcription initiation to directly repress a subset of genes. Although ZNF143 stimulates initiation at ZNF143-repressed genes (i.e. those that increase expression upon ZNF143 depletion), the molecular context of binding leads to cis repression. ZNF143 competes with other more efficient activators for promoter access, physically occludes transcription initiation sites and promoter-proximal sequence elements, and acts as a molecular roadblock to RNA Polymerases during early elongation. The term context specific is often invoked to describe transcription factors that have both activation and repression functions. We define the context and molecular mechanisms of ZNF143-mediated cis activation and repression.


Introduction
Transcription factors bind directly to DNA and recruit cofactors and RNA Polymerase to regulate gene expression throughout the lifetime of all organisms.Transcription factors specify tissue patterning throughout development and their function is necessary to maintain cellular and organismal homeostasis (Lewis 1978;Nüsslein-Volhard and Wieschaus 1980).Sequence-specific transcription factors are comprised of DNA binding domains that recognize degenerate sequence motifs and effector domains that interface with cofactors that specialize in regulatory roles such as chromatin remodeling, initiation, pause release, and elongation.Just as cofactors are specialists, transcription factors can specialize by predominantly regulating specific steps in transcription (Duarte et al. 2016;Scholes et al. 2017;Scott et al. 2024).Factors achieve this specificity by preferentially recruiting certain cofactors through their effector domains.We propose that it is critical to classify transcription factors by their molecular function, as opposed to broad activator and repressor classes, in order to understand the context specificity of gene regulation.For example, recruiting a transcription factor that specializes in RNA Polymerase pause release to a gene may have little effect on transcription if redundant pause-release factors are already present; at this gene a fac-tor that recruits components of the pre-initiation complex may cause potent activation.This work provides a framework for systematically determining the molecular functions of transcription factors by stimulating their rapid depletion and quantifying changes in RNA polymerase distribution at direct target genes in the minutes following transcription factor depletion.
Despite advances in understanding the mechanisms of transcription and developments in the systems biology of gene regulation field, accurately predicting direct target genes of transcription factors is a challenge.This struggle to predict target genes is in part because of the lack of reliable input data into predictive models.Researchers cannot experimentally identify a comprehensive set of primary response genes for the vast majority of transcription factors because they cannot rapidly induce or rapidly inhibit their activity.We know the most about transcription factors that are rapidly activated by acute environmental changes such as heat stress or nuclear receptor factors that are induced by addition of steroids (Duarte et al. 2016;Guertin et al. 2010; Kumar and Chambon 1988;Kumar et al. 1987;Westwood et al. 1991).Many models rely on input data where a transcription factor is depleted chronically for days or the lifetime of a cell, so it is impossible to discriminate target genes from the complex cascade of transcription dysregulation that follows.However, recent development of rapidly inducible degron systems democratizes the study of transcription factors that are not easily activated or inhibited (Nabet et al. 2020(Nabet et al. , 2018;;Nishimura et al. 2009).Here, we C-terminally tagged all endogenous copies of the ubiquitous and essential transcription factor ZNF143 with FKBP F36V in HEK-293T cells.We rapidly depleted ZNF143 by adding dTAG V -1 to study its molecular function and define ZNF143's target genes.
The Xenopus homolog of ZNF143 was cloned and described three decades ago (Schuster et al. 1995).This original report characterized the binding site and activator function of ZNF143 (contemporaneously termed Staf) using reporter assays (Schuster et al. 1995).Subsequent transcriptional profiling upon chronic ZNF143 depletion identified many up regulated and down regulated genes (Ngondo-Mbongo et al. 2013); more genes were up regulated and the authors concluded that ZNF143 is primarily an activator while noting that ZNF143 may be involved in repression.Our results confirm that ZNF143 binds DNA to activate local genes and defines the mechanism of ZNF143 mediated activation.Furthermore, we define several molecular mechanisms through which a canonical activator can paradoxi- cally retain its activator function while directly repressing target genes in cis.Although cis repression may only account for up to 30% of direct ZNF143-repressed targets, alternative mechanisms of immediate indirect repression, such as relieving competition for cofactors (Guertin et al. 2014;Schmidt et al. 2016Schmidt et al. , 2015)), may explain why additional genes are repressed indirectly.
Many publications erroneously report that ZNF143 has a prominent role in chromatin looping (Bailey et al. 2015a;Heidari et al. 2014;Liu et al. 2022;Yang et al. 2017;Zhang et al. 2016Zhang et al. , 2024)).However, two recent studies found that ZNF143 has no role in chromatin looping (Magnitov et al. 2024;Narducci and Hansen 2024).These groups convincingly determine that the reason ZNF143 was ascribed a looping role is because a commonly used ZNF143 antibody crossreacts with the looping transcription factor CTCF (Magnitov et al. 2024;Narducci and Hansen 2024).
ZNF143's importance is highlighted by the rarity of characterized human disease with ZNF143 mutations because a lack of ZNF143 function is potentially incompatible with life.A case report of a child with missense and premature stop codon ZNF143 alleles exhibited deficiencies related to the lack of viable vitamin B12 cofactors and presented with global developmental delay (Pupavac et al. 2016).The authors linked the dysregulation of the ZNF143-regulated mitochondrial protein Methylmalonyl-CoA mutase to the observed signs and symptoms of vitamin B12-deficiency (Pupavac et al. 2016).This is consistent with the role of ZNF143 regulating nuclear encoded mitochondrial genes (Magnitov et al. 2024;Narducci and Hansen 2024).The patient also exhibited a range of symptoms that were not directly explainable by the effect of this mutation on vitamin B12 metabolism.Recent work shows that ZNF143 depletion in mouse embryo models limits the ability of stem cells to differentiate into their final lineages (Magnitov et al. 2024), which demonstrates ZNF143's important role in early development and is a plausible explanation for the constellation of symptoms experienced by this patient.The only other reported case of ZNF143 in human disease is a family with a history of inherited endothelial corneal dystrophy, which results in premature thickening of the cornea and requires surgery at a young age to preserve vision (Kim et al. 2019).This family's ZNF143 missense mutation is distinct from those of the previous case and appears to be only associated with the corneal symptoms, although vitamin metabolism abnormalities or symptoms affecting other organ systems were not reported.The varied manifestations and rarity of ZNF143 mutations in human populations make it difficult to study, but also demonstrates its importance to a wide variety of pathways that contribute to health and cell viability.For an essential protein such as ZNF143, rapidly inducible perturbation systems are generally more powerful when studying molecular mechanisms because molecular phenotypes can be quantified in the minutes and hours after losing the proteins function before downstream effects or compensatory mechanisms come into play.

ZNF143 is rapidly degraded from DNA within 30 minutes
To determine the molecular function of ZNF143, we engineered HEK-293T cells to express all alleles of ZNF143 tagged with a C-terminal inducible dTAG degron system, facilitating rapid and complete protein degradation (Na- Quantitative western blots indicate that less that 10% of ZNF143 remains after 15 minutes of dTAG V -1 treatment and ZNF143 is not detected at 30-90 minutes after induced degradation.We profiled ZNF143 binding genome-wide by ChIP-seq before and after 30 minutes of degradation to quantify depletion on chromatin.On the same scale, ZNF143 is not detectable after 30 minutes of degradation, but digital over-exposure of the heatmaps indicates that less than 5% of ZNF143 remains bound to DNA (Figure S1A).This cell line represents a powerful resource for investigating the molecular functions of ZNF143 and directly identifying ZNF143 target genes.

ZNF143 binds a degenerate 29 base motif at each binding site in the genome
Next, we conducted iterative and exhaustive motif analysis of ZNF143 ChIP-seq peaks to systematically identify and characterize the diverse sequence motifs that ZNF143 binds in the genome.We performed de novo motif analysis within the 4682 ZNF143 ChIP-seq peaks and identified a canonical ZNF143 sequence motif within 58% peaks (Figure 1C).ZNF143's seven zinc fingers are known to bind a wider region (Zhang et al. 2024), so we performed the same analysis on remaining peaks without the first identified motif.We found a second ZNF143 motif variant and two more iterations of this process identified other ZNF143 motifs (Figure 1C).Ninety-five percent of the peaks had a de novo-identified motif that was clearly a ZNF143 motif variant (Figure 1C).These findings demonstrate ZNF143's flexibility in binding; it can accommodate a wide range of sequences beyond the core TGGGA sequence that is recognized by zinc fingers 5 and 6 (Zhang et al. 2024).By anchoring the analysis on this core sequence, we calculated the average frequency of each nucleotide (A, C, G, T) within a 100-base range, highlighting how ZNF143's binding preferences extend into the surrounding genomic landscape.The nucleotide frequencies stabilize to the genomic background outside a 29-mer core ZNF143 motif.This 29mer core is similar to the biochemically determined 27-mer ZNF143 core motif found previously (Zhang et al. 2024).We generated a composite motif using the frequencies in this window (Figure 1D).We did not identify a motif de novo in 5% of binding sites; however, these peaks remain sensitive to ZNF143 depletion and retain potential functionality.Low-affinity binding sites of transcription factors, which do not strictly conform to a consensus motif, are increasingly recognized as having a critical role in gene regulation and development (Jindal et al. 2023;Lim et al. 2024).We determined the precise ZNF143 binding positions within each ChIP-seq peak by assigning the motif with the lowest pvalue in the peak as the inferred 29 base binding site (Figure S1B).All subsequent references and analyses concerning ZNF143 binding focus on the 29-mer ZNF143 recognition sequences within the ChIP-seq peaks.These analyses confirm the DNA-binding function of ZNF143 and indicate that ZNF143 binds a sequence-degenerate 29 base wide motif within chromatin.

ZNF143 maintains chromatin accessibility at a minority of binding sites
To investigate the impact of ZNF143 on chromatin accessibility, we conducted ATAC-seq analysis before and after 30 minutes of ZNF143 depletion.Fewer than 500 ATACseq peaks significantly (FDR < 0.1) change accessibility after 30 minutes of ZNF143 depletion (Figure 2A).Over 99% of the significantly changed ATAC peaks decrease accessibility and 94% of the decreased accessibility peaks overlap ZNF143 binding sites (Figure 2B&C).Although the vast majority of decreased ATAC peaks are directly regulated by ZNF143 binding, the inverse is not true.Only a minority (10%; 479/4684) of ZNF143 peaks significantly reduce chromatin accessibility upon ZNF143 ablation from chromatin.These results underscore that while ZNF143 maintains open chromatin at certain loci, its influence is not uniform across all binding sites, suggesting that regulating chromatin structure is not its primary function.

ZNF143 is an activator that predominantly regulates transcription initiation
We performed nascent RNA profiling (PRO-seq) after 30 minutes of ZNF143 depletion to identify direct ZNF143 target genes (Kwak et al. 2013).Hundreds of genes increase (up) and decrease (down) transcription after 30 minutes of ZNF143 depletion (Figure 3A).Although the expression of primary response genes goes in both directions, this does not necessarily mean that ZNF143 functions as an activator and a repressor.The transcription factors that we know most about are rapidly inducible, such as heat shock factor and nuclear hormone receptors such as the estrogen, androgen, and glucocorticoid receptors.Although genes are activated and repressed in the minutes following transcription factor induction, only the activated genes are closer to the respective binding sites of the factor (Duarte et al. 2016;Dutta et al. 2023;Guertin and Lis 2010;Guertin et al. 2014;Hasterok et al. 2023;Reddy et al. 2009).The repressed genes are no closer to the factor binding sites than unchanged control genes, so only the activated genes are considered direct cis targets of these transcription factors.We performed these same gene class/peak proximity analyses for the ZNF143 regulated genes and unchanged genes that are matched for expression levels (Figure 3B).Nearly all the down genes have their transcription start site within 500 bases of a ZNF143 binding site (Figure 3B) and the vast majority of ZNF143 binding sites are in the upstream promoter region (Figure 3C).Unlike the heat shock and hormone response systems, the opposing up gene class is enriched for proximal ZNF143 binding (Figure 3B), although ZNF143 binding is not limited to the promoter region (Figure 3C).These results reveal that the predominant function of ZNF143 is to activate transcription proximally and the remainder of this section will focus on determining the molecular mechanism of ZNF143-mediated activation.
The transcription cycle includes many steps including chromatin decondensation, RNA polymerase recruitment/initiation, promoter-proximal pause termination or release, and elongation (Fuda et al. 2009;Wagner et al. 2023).We quantified the change in PRO-seq density within the 5 pause region and gene bodies of down genes and incorporated these values into a compartment model (Figure S2) to determine whether ZNF143 predominately regulates initiation or pause release.Our modeling analysis indicates that a decrease in initiation rate at every down gene with ZNF143 binding within the 500 base promoter best explains the redistribution of RNA polymerase in the pause and gene body regions upon ZNF143 depletion (Figure 3D).In contrast, A) Nascent transcriptomics identifies 182 genes that increase expression (up) and 365 genes that decrease expression (down) after 30 minutes of ZNF143 degradation.Genes that are matched for expression level and unchanged are in dark grey.B) Ninetysix percent of down genes are within 500 bases of a ZNF143 binding site.The up genes are also significantly closer to ZNF143 binding sites compared to genes that are matched for expression level.C) Down genes have ZNF143 binding in the promoter; up genes have no clear pattern of ZNF143 distribution and 66% have no local ZNF143 binding.D) Compartment modeling indicates that ZNF143 regulates initiation and not pause release at direct down target genes.Each point is a down gene and the y-axis values are the changes in rates that most likely explain the data.E) The transcription start sites of direct down target genes tend to change upon ZNF143 degradation.F) We assigned predominant transcription start sites pre/post ZNF143 degradation.The decrease in TSS usage at the control prominent TSS is accompanied by an increase in TSS usage of the prominent dTAG-treated TSS.

A Differential Expression
a comparable fraction of gene's pause release rates change in both directions (Figure 3D).Compartment modeling of changes in PRO-seq signal cannot distinguish between decrease initiation and increasing premature pause termination because these are directly opposing rate constants (Figure S2).
To determine whether ZNF143 regulates initiation/recruitment versus premature pause termination we sought to determine whether ZNF143 affects transcription start site (TSS) usage.Our rationale is that if ZNF143 regulates initiation, then the predominant TSS would change upon ZNF143 depletion.Previous work showed that rapid depletion of the initiation factor TATA-binding protein (TBP) causes changes in TSS usage (Santana et al. 2022).The 5 end of PRO-seq reads accumulate at a gene's TSS and we inferred the most prominent TSS before and after 30 minutes of ZNF143 degradation from 5 read pileups.We focused this analysis exclusively on down genes with ZNF143 binding within 500 bases of the TSS in the promoter region.Down genes with promoter bound ZNF143 shift their most prominent TSS at twice the rate compared to a control set of genes matched for TSS signal intensity (Figure 3E).For example, after ZNF143 degradation ZNF30's TSS shifts from 159 to 18 bases away from ZNF143's binding site (Figure 3E).This suggests that ZNF143's close proximity inhibits initiation at the alternative dTAG TSS.Transcription start sites within the human genome are not focused at a single position, but occur in windows and genes can contain multiple TSS windows (Luse et al. 2020).The observed changes in the most prominent TSS could occur because signal decreases at the control TSS and an alternative less prominent TSS does not change intensity.However, we observe a coupled decrease in signal at the control TSS and an increase in signal at the ZNF143-depleted TSS (Figure 3F).This pattern suggests that not only does ZNF143 regulate transcription initiation as opposed to premature termination, but the redistribution of TSS signal suggests that ZNF143 competes with other transcription factors for RNA polymerase and/or initiation machinery.The role of ZNF143 stimulating initiation is consistent with the chromatin accessibility data from Figure 2.Although chromatin structure is influenced directly by the recruitment of chromatin remodelers and histone-modifying enzymes, the recruitment of RNA Polymerase or initiation factors can also deplete nucleosomes, thereby altering the chromatin landscape.

ZNF143 directly represses genes by binding directly over transcription start sites
The canonical molecular functions of ZNF143 are succinctly described as DNA binding and stimulation of transcription initiation.However, it is not immediately clear how these functions lead to the direct cis repression of ZNF143 target genes observed in Figure 3B.ZNF143 binds both upstream and downstream of up genes at comparable frequencies (Figure 3C), but it directly binds over the TSS of five genes (Figure 4A&B).ZNF143 exhibits stronger binding over the TSS of four out of these five genes com-   ZNF143 binds strongly over the TSS of 4/5 genes that increase transcription upon ZNF143 depletion.C) FIS1 is the up gene from panel B with weak TSS binding of ZNF143 and no discernible changes in TSS usage occur upon depletion, as measured by PRO-seq 5 end pile ups.D) ZNF143 binds directly downstream from the TSS of a different isoform of FIS1.This FIS1 isoform becomes the most prominent FIS1 TSS upon ZNF143 degradation.E&F) ZNF143 degradation from strong binding sites directly downstream of transcription start sites is associated with up genes.G&H) Genic bidirectional transcription at ZNF143 binding sites downstream of up genes significantly decreases at 7/28 up genes.
pared to matched and down genes with TSS-bound ZNF143 (Figure 4B).These observations support a model in which ZNF143 binding directly over the TSS competes with RNA Polymerase for access to the initiator sequence.The distribution of 5 PRO-seq reads remains consistent with or without ZNF143 depletion at the one down gene, FIS1, that exhibits weak ZNF143 binding over the TSS (Figure 4C).However, a closer inspection of this gene reveals that the most prominent TSS changes upon ZNF143 depletion and 5 PRO-seq signal substantially increases at this new TSS (Figure 4D).A much stronger ZNF143 binding site is located directly downstream of this FIS1 TSS-isoform (Figure 4D).This observation suggests an independent repressive mechanism, where ZNF143 binding directly downstream of a TSS acts as a molecular roadblock, which will inhibit RNA Polymerase from progressing into the gene body and effectively inhibit initiation of new RNA Polymerases.

ZNF143 directly represses genes by acting as a molecular roadblock immediately downstream of transcription start sites
FIS1 was not initially classified as a ZNF143-downstream gene because we inferred a single isoform of each primary transcript using the control data (Anderson et al. 2020).We then extended our analysis to the twenty-eight up genes with ZNF143 binding within 500 bases downstream of their control TSSs (Figure 4E).These 28 genes have strong ZNF143 binding downstream of their start sites compared to matched and down genes with downstream ZNF143 peaks (Figure 4F).These results are consistent with a molecular roadblock model, whereby strong transcription factor binding can inhibit RNA polymerase during early elongation when it is accelerating and more vulnerable to pausing, backtracking, and disassociation.Beyond ZNF143's role in DNA binding, we have established that it also stimulates initiation.Importantly, RNA polymerase initiates transcription at bidirectional regions across the genome, not solely at genic TSSs (Core et al. 2014(Core et al. , 2008)).Bidirectional transcription (Azofeifa and Dowell 2017;Wang et al. 2019), especially an RNA polymerase colliding head on with a genic preinitiation complex or paused RNA polymerase, could also act to repress transcription.We find that 7 out of these 28 downstream ZNF143-binding genes have significantly reduced bidirectional transcription after ZNF143 depletion (Figure 4G&H).This evidence supports the hypothesis that ZNF143 not only activates but may also repress transcription by promoting RNA polymerase initiation, highlighting the importance of proposing specific mechanisms when claiming transcription factors have dual activation/repression functions.

ZNF143 occludes downstream sequence elements to repress transcription
ZNF143 binds to the promoters of three up genes that have an upstream shift in their TSS after ZNF143 depletion (Fig- ures 5B&S4).We hypothesized that ZNF143 binding occludes a downstream sequence motif that contributes to initiation.The downstream promoter element (DPE) located +28 to +32 downstream of TSSs was the first downstream element shown to contribute to initiation by interacting with TFIID (Burke and Kadonaga 1997).ZNF143 blocks a DPE that is located at positions +27 to +31 within the LYSMD1 gene.Additionally, YY1, an initiation factor (Athanikar et al. 2004;Seto et al. 1991), has a binding motif that is  commonly found directly downstream of TSSs and between +100 to +300 positions downstream (Benner et al. 2013;Dudnyk et al. 2024) (Figure 5A).ZNF143 binds upstream of LYSMD1, GMPR2, and ZNF583, and upon ZNF143 depletion, a YY1 motif becomes accessible at each of these genes (Figures 5B&S4).YY1 binding at the exposed site may facilitate efficient initiation at the upstream TSS.This YY1 occlusion mechanism may be acting simultaneously as a roadblock and/or directing genic bidirectional transcription for the upstream dTAG TSS (Figure S4).These three up genes are unique because ZNF143 is promoterbound and ZNF143 depletion causes an upstream TSS shift.This mechanism cannot explain repression at up genes with promoter-bound ZNF143 and no upstream TSS shift upon depletion.

SP1 redistributes to ZNF143 binding sites to stimulate initiation
ZNF143 bound the promoters of 25 up genes with distances and intensities comparable to those of down genes and no TSS shift upstream of ZNF143's binding site (Figure 3C).Given the limited number of up genes, we catalogued all bidirectionally transcribed putative regulatory elements across the genome and analyzed transcriptional changes in these regions after ZNF143 depletion (Love et al. 2014;Wang et al. 2019) (Figure 6A&B).We then carried out de novo motif discovery within the bidirectionally transcribed regions of both down and up classes.As anticipated, the ZNF143 motif was prevalent in the down class; however, its presence in the up class was unexpected (Figure 6B).Additionally, the SP transcription factor family DNA motif was identified de novo in the up class (Figure 6B).Ninetynine bidirectional up regions have both ZNF143 ChIP peaks and ZNF143 motifs, with nineteen of the ZNF143 motifs overlapping SP motifs (Figure 6C&D).We performed SP1 ChIP-seq before and after 30 minutes of ZNF143 depletion to address the possibility that SP1 is redistributing to these sites when ZNF143 vacates.Each of the these 19 regions overlapped an SP1 ChIP-seq peak and 13 of the peaks increased SP1 intensity upon ZNF143 depletion (Figure 6E).SP1 also regulates initiation (Dutta et al. 2023;Gill et al. 1994), so these results are consistent with a model whereby an canonical activator can repress a gene if it competes with a more potent activator for the same stretch of DNA.SP1 is clearly not replacing ZNF143 at all nineteen sites, but recall that ZNF143 binds a 29 base region and the ZNF143 motif may be overlapping the motifs for other initiation factors.
SP1 motifs were among the first promoter sequence elements to be identified as regulatory (McKnight and Kingsbury 1982) and SP1 has had a defined role in initiation for thirty years (Gill et al. 1994).Recent studies have further refined SP motifs as defining elements of promoter architecture in the genome (Benner et al. 2013;Dudnyk et al. 2024;Jones et al. 2024).Motifs recognized by paralogs of ZNF143, along with NRF1, ETS, NFY, CREB/ATF (bZIP), and SP/KLF factors, are considered critical in dictating promoter structure and initiating transcription (Benner et al. 2013;Dudnyk et al. 2024;Jones et al. 2024).The sequencespecific transcription factors that bind these motifs may be a class of sequence-specific transcription factors that recruit the initiation machinery and RNA polymerase.Our findings confirm the enrichment of these motifs within bidirectional promoters across the human genome (Figures 7A&S3).Sixteen up genes with ZNF143 in the promoter have an SP motif within an SP1 ChIP-seq peak that overlaps the ZNF143 element and 10 of the 16 have an increase in SP1 ChIP signal after ZNF143 depletion (Figure 7B).The promoters of ZNF688, UTP3, and PYCR2 exhibit a statistically significantly (FDR < 0.01) increase in SP1 signal upon ZNF143 depletion; however, the redistribution of SP1 at other genes is either more subtle or non-existent (Figure 7B).We searched for the motifs of these sequence-specific initiation transcription factors within the 29mer ZNF143 motif of the 28 genes with promoter-bound ZNF143.At least one of the bZIP, SP/KLF, ETS, and NRF1 motifs overlapped the ZNF143 motif (Figure 7C&D) in all genes except TSNAXIP1 (Figure 7D).We propose that other activating factors replace ZNF143 at these promoters when ZNF143 is depleted.For example, ZNF143 is ablated and SP1 intensity reduced two-fold at their overlapping binding site in the promoter of GSK3A, but we hypothesize that this reduction of binding enables an ETS factor to bind strongly and stimulate transcription (Figure 7D).We also observe the same overlap of ZNF143 and NRF1/ETS/SP motifs at bidirectionally transcribed regions from Figure 6D (Figure 7E).These results suggest that a canonical activator may repress genes by competing for DNA binding with a more potent activator.However, we appreciate that the molecular calculus for such a mechanism is complex.For instance, a very strong activator might have a brief residency time on DNA, making its overall contribution minimal, while a modest activator could have more stable DNA binding and effectively stim- Up genes with promoter ZNF143 binding ulate transcription.Given that each promoter has a unique sequence context, predicting transcriptional outcomes solely based on sequence is challenging.To fully understand the competitive interactions between two transcription factors targeting the same DNA region, one must consider: 1) their relative binding efficiencies and residency times within specific chromatin contexts; 2) the transcription cycle stages each factor influences; 3) the relative activation potency of each factor; and 4) which steps in the transcription cycle limit the output of target genes.

Discussion
When discussing the context-specific dual roles of transcription factors as both activators and repressors, it is crucial to propose an underlying mechanism (Figure 8).The most well characterized example of dual activator/repressor function in gene regulation is that of the λ repressor.The λ repressor is an activator that interacts with the bacterial RNA polymerase to stimulate initiation, but it was named repressor because it represses transcription of the lytic genes, which account for nearly all the phage genes.The λ repressor binds to a precise position in its own promoter to stimulate RNA polymerase initiation and activate the gene encoding itself (Ptashne 2004).This same binding event blocks RNA polymerase from accessing and initiating transcription of lytic genes (Johnson et al. 1979;Meyer et al. 1980;Ptashne 2004).The molecular functions of both λ repressor and ZNF143 are to bind DNA and stimulate initiation.Binding of ZNF143 to DNA is necessary to stimulate initiation; like λ repressor, ZNF143 typically binds upstream of a initiation site to activate transcription.Bind-ing in a promoter and stimulating initiation may result in repression if ZNF143 displaces a more potent stimulator of initiation.Binding over an initiator sequence or within a gene body can cause repression via multiple described mechanisms.While ZNF143 can both activate and repress genes, its molecular functions of DNA binding and promoting initiation remain constant.
Many transcription factors are described as having both activator and repressor activities and context specificity is often vaguely invoked as the explanation.There are conflicting reports on whether MYC acts as an activator at specific genes, an amplifier of all genes, or a direct repressor at some genes (Lin et al. 2012;Lorenzin et al. 2016;Nie et al. 2012;Sabo et al. 2014;Walz et al. 2014).The acute depletion of an auxin inducible degron tagged MYC within 30 minutes, coupled with nascent RNA sequencing, allowed for the identification of primary MYC-responsive genes (Muhar et al. 2018).98% of these genes were repressed, indicating that the direct effect of MYC regulation is transcriptional activation of only a fraction of the expressed genes in a cell (Muhar et al. 2018).OCT4 is one of the four pluripotency factors, the expression of which is sufficient to reprogram differentiated fibroblasts into induced pluripotent stem cells (Takahashi and Yamanaka 2006).However, the genes OCT4 regulates in pluripotent stem cells are difficult to identify, as the half-life of its protein and mRNA are much too long for traditional knockdown methods to isolate the primary effects of depletion (Bates et al. 2021).A recent study compared extended knockdown to rapid depletion with targeted protein degradation and found that only the latter was able to identify that the primary effect of OCT4 on transcription is the activation of pluripotency factors and that the delayed activation of trophoblast-associated genes is a secondary effect of OCT4 depletion (Bates et al. 2021).A key takeaway from these studies is that these factors directly activate transcription of their target genes.The growing list of transcription factors that can be acutely perturbed provides evidence that most factors do not activate some direct targets and repress others.Another theme is that, in contrast to extended knockdown, acute depletion of most sequencespecific factors affects transcription of a limited number of primary response genes (Sheppard et al. 2021;Stengel et al. 2021).
There is substantial data suggesting that transcription factors specialize and recruit either corepressors or coactivators.We can conceive of exceptions to this rule, but we would argue that a factor can become a fundamentally different protein with distinct functions based on ligand binding status or post-translational modifications.The most well-characterized example of this is the thyroid hormone receptor (THR).THR binds to DNA and recruits corepressors in the absence of thyroid hormone (Hörlein et al. 1995); THR recruits coactivators when bound by thyroid hormone (Chen et al. 1997;Fondell et al. 1996;Grøntved et al. 2015;Lin et al. 1997).Post-translational modifications, such as phosphorylation or acetylation, fundamentally change the identity of a protein and this may result in the ability to differentially interact with cofactors.Although we were unable to identify well-characterized examples of posttranslational modifications that switch interaction partners from coactivators to corepressors or vice versa, this mechanism has been suggested for STAT3, which recruits corepressors exclusively when acetylated (Gambi et al. 2019).Extending this general mechanism, a transcription factor may be bound to DNA, but only interact with coactivators when posttranslationally modified.In the absence of the modification, the factor would be a functional repressor by occluding the binding sites of other activating factors and failing to recruit coactivators.
Another key part of context for defining the role of a TF is promoter strength.Transcription factor activity modeled in E. coli demonstrates that a TF can appear to change from an activator and repressor or vice versa depending on promoter strength and which step of the transcription cycle is regulated by the TF (Ali et al. 2023).In their model, transcription factors were characterized by their ability to either regulate the stability of RNA polymerase or promote transcription initiation.They found that TFs with strong stabilizing functions but weaker initiating abilities moved from activation to repression as promoter strength increased.We can envision how the same RNA polymerase stabilization mechanism might appear to repress transcription in the presence of a strong promoter if the stabilization interferes with the ability of RNA polymerase to begin transcription.On the other end of the spectrum, TFs that were strong initiators but weaker stabilizers moved from repression to activation with increasing promoter strength.Once again, we see that a TF appears to change activity when in reality the strong promoter may make it easier for transcription to initiate when perhaps other factors are able to improve RNA polymerase stability.The importance of promoter strength in eukaryotes is highlighted by an effector domain mutagenesis and mapping study that identified effector domains with the ability to seemingly both activate and repress transcription (DelRosso et al. 2023).Here the definition of "activator" and "repressor" becomes important, as this study defined activators as effector domains able to increase expression as compared to basal levels from a minimally active promoter and repressors as those able to decrease expression as compared to basal levels from a constitutively active promoter.According to these definitions, it is possible that a "bifunctional" effector domain is actually a weak activator if it promotes transcription, but at a lower level than the constitutively active promoter.Although the details of transcription regulation in prokaryotes differs from eukaryotes, the overall principle appears to be similar in that promoter strength, combined with the particular role of a TF, both contribute to whether a particular TF appears to be activating or repressing transcription.Understanding the relationship between promoter strength and the specific functions of TFs can explain their seemingly dual role in activating and repressing gene expression.
Although this study convincingly demonstrates that ZNF143 depletion can lead to the repression of local genes via our proposed plausible mechanisms (Figure 8), we have yet to distinguish between incidental regulation and conserved evolutionary mechanisms that specifically govern these genes.The identity of the ZNF143-repressed genes suggest some coordinated regulatory control.FIS1 is a mitochondrial fission gene and ZNF143 positively regulates nuclearly encoded mitochondrial genes (Magnitov et al. 2024).We speculate that activating nuclearly encoded mitochondrial genes that are involved in mitochondrial biogenesis and function while simultaneously repressing fission genes can be a strategic response to enhance mitochondrial function and biogenesis.This coordinated response may build a robust mitochondrial network to meet increased metabolic demands or recover from damage more effectively.Another gene that is repressed in cis by ZNF143 is THAP11.THAP11 binds to a nearly identical sequence motif as ZNF143 (Ngondo-Mbongo et al. 2013;Vinckevicius et al. 2015), although there is no clear evolutionary conservation of their respective DNA binding domains.The regulation of THAP11 by ZNF143 is not clear from our data, although we propose that ZNF143 displaces an ETS factor (Figure 7).Regulation of THAP11 is likely more complicated, as close inspection of the locus reveals an unannotated TSS that is flanked by two ZNF143 ChIP peaks, which are also likely THAP11 binding sites (Figure S5).Moreover, usage of this unannotated TSS increases substantially upon ZNF143 degradation.Given the interplay between ZNF143 and the regulation of genes like FIS1 and THAP11, it is likely that at least some aspects of regulatory repression are crucial for maintaining homeostasis and orchestrating cellular responses to metabolic changes.This work underscores the significance of using rapidly inducible systems to study molecular mechanisms.Additionally, we highlight that while genomic experiments and analyses indicated ZNF143 was repressing a subset of genes in cis, a more detailed examination of these repressed genes was essential to suggest mechanisms of repression.We recognize that a thorough understanding of biological systems necessitates an in-depth grasp of the individual components and mechanisms.We therefore anticipate that the pendulum of scientific inquiry will swing back from broad molecular genomics studies, high throughput reporter assays, and large scale screens towards more focused, mechanistic investigations.

HEK-293T culture and ZNF143-dTAG clone generation
HEK293T cells were cultured at 37°C with 5% carbon dioxide in DMEM media supplemented with 10% fetal bovine serum (FBS), 100 U/mL penicillin-streptomycin, 2.2mM L-glutamine, and 1mM sodium pyruvate.We endogenously tagged ZNF143 at the C-terminus in HEK-293T cells as previously described (Sathyan et al. 2020;Scott et al. 2024).We targeted ZNF143 endogenously using CRISPR loaded with the sgRNA (GAGGATTAATCATC-CAACCCTGG).We cleaved the hSpCas9 plasmid PX458 (Addgene #48138) with the enzyme BbsI, then annealed oligonucleotides 5´-CACCGAGGATTAATCATCCAACCC-3´and 5´-AAACGGGTTGGATGATTAATCCTC-3´, and inserted the annealed product into the plasmid.We generated a linear homology-directed repair donor by amplifying the pCRIS-PITChv2-dTAG-Puro plasmid (Addgene #91796) with the primers in Table 1 (Nabet et al. 2018).Following transfection of the donor DNA and Cas9/sgRNA plasmid into HEK-293T cells, cells were selected and cloned as described in (Sathyan et al. 2020).We selected cells in media with 1 µg/ml puromycin and confirmed successful integration by Western blot as previously described (Scott et al. 2024).After obtaining a correctly-tagged clone (ZD29) we passaged cells thawed from frozen aliquots until the desired number of cells for each experiment was reached, and then treated with dTAG V -1 and collected cells for ATAC-seq, PRO-seq, and ChIP-seq.

ATAC-seq library preparation
We prepared ATAC-seq libraries as previously described (Grandi et al. 2022;Scott et al. 2024).We treated cells with either DMSO or dTAG V -1 in DMSO for thirty minutes.After treatment, we aspirated the media and scraped the cells on ice and centrifuged them at 500g for 5 min at 4°C.We resuspended cell pellets in lysis buffer (0.1% NP40, 0.1% Tween-20, 0.1% digitonin) made in cold resuspension buffer (10mM Tris-HCl ph 7.5, 10mM NaCl, 3mM MgCl 2 in water) and then incubated on ice for 3 minutes.We mixed lysates with wash buffer (0.1% Tween-20 in resuspension buffer) and centrifuged cells at 500g for 10 min at 4°C.We resuspended the pellets in a transposition mixture (1x TD buffer, 0.1% ultrapure distilled water, 0.01% digitonin, 0.1% Tween-20 in PBS) and added 2.5µl of TDE1 Tn5 transposase (Illumina Tagment DNA TDE1 Enzyme and Buffer Kit).We incubated the transposition reaction at 37°C for 30 min.We isolated DNA from the reaction with the DNA Clean and Concentrator-5 kit.We added adapters and amplified DNA over 8 cycles of PCR using the NEBNext Ultra II kit.We purified and size selected the library by incubating samples with AMPure XP beads (1.8x buffer to sample ratio) and eluting with nuclease-free water.

ATAC-seq analyses
We aligned raw sequence data to the hg38 genome assembly with bowtie2, converted to BAM format with samtools, and then to bigWig format with seqOutBias (Langmead and Salzberg 2012;Li et al. 2009;Martins et al. 2018).Peak calling with MACS3 (Zhang et al. 2008) employed the following arguments: -q 0.01 -keep-dup all -nomodel -shift -100 -extsize 200.We counted reads in peaks with the bigWig R package and called differentially accessible regions with DESeq2 (Love et al. 2014;Martins 2014).

PRO-seq library preparation
We prepared PRO-seq libraries as previously described (Mahat et al. 2016;Sathyan et al. 2019).After thirty minutes of treatment with either 500nM dTAG V -1 or DMSO, we washed cells ice cold PBS.We collected cells by adding buffer W (10mM Tris-HCl pH 7.5, 10mM KCl, 250mM sucrose, 5mM MgCl 2 , 1mM EGTA, 10% glycerol, 0.5mM DTT, 0.004U/mL SUPERaseIN RNase inhibitor, and fresh protease inhibitors) and scraping the cells, followed by centrifugation at 500g for 5 minutes and resuspension in buffer W. We added buffer P (10mM Tris-HCl pH 7.5, 10mM KCl, 250mM sucrose, 5mM MgCl 2 , 1mM EGTA, 0.05% Tween-20, 0.1% NP40, 10% glycerol, 0.5mM DTT, 0.004U/mL SUPERaseIN RNase inhibitor, fresh protease inhibitors) for 5 minutes to permeabilize the cells.We centrifuged and resuspended the cells in buffer W twice before pelleting the cells again and resuspending in buffer F (50mM Tris-HCl pH 8, 5mM MgCl 2 , 1.1mM EDTA, 40% Glycerol and 0.5mM DTT).We snap froze aliquots in liquid nitrogen and kept them stored at -80°C.PRO-seq library prep was done in a method based on previously described protocols (Judd et al. 2020;Mahat et al. 2016).After the run-on reaction, we added adapters that included a random eight base unique molecular identifier (UMI) on the 5 end of adapter that is ligated to the 3 end of the nascent RNA.We eluted and reverse transcribed the RNA and performed 10 cycles of PCR.We purified the PCR reactions with a MinElute PCR purification kit and did not perform size selection in an effort to preserve short nascent RNAs in our libraries.

PRO-seq analyses
Quality control and read alignment were performed as described previously (Scott et al. 2022).We used cutadapt to remove adapters from our reads (Martin 2011), and fqdedup to deduplicate our libraries with the 3 UMIs (Martins and Guertin 2018).We removed 8-mer UMIs and reverse complemented the reads with FASTX-Toolkit (Gordon 2010).We aligned to hg38 with bowtie2 (Langmead and Salzberg 2012), sorted reads with samtools (Li et al. 2009), and used seqOutBias (Martins et al. 2018) to convert reads to big-Wig files.We used primaryTranscriptAnnotation and TSSinference to infer gene annotations from our PRO-seq data (Anderson et al. 2020;Dong and Guertin 2024).We quantified gene expression by querying the big-Wig files within the gene annotation coordinates with the bigWig R package and UCSC Genome Browser Utilities (Kent et al. 2010;Martins 2014).We found differentially expressed genes with DESeq2 (Love et al. 2014).We identified regions of bidirectional transcription with dREG (Wang et al. 2019).We identified overrepresented motifs de novo in dREG-defined regions with MEME (Bailey et al. 2015b).We modeled the rates of transcription initiation and pause release using a compartment model as previously described (Dutta et al. 2023).

ChIP and library preparation
We fixed cells with 1% formaldehyde for 10 minutes at 37°C and quenched them with 125mM glycine for 10 minutes at 37°C.We then moved plates to ice and washed and scraped the cells into ice-cold PBS containing fresh protease inhibitors.We centrifuged cells in aliquots of 2x10 7 cells at 1500g for 5 minutes, snap froze them in liquid nitrogen, and stored them at -80°C.After thawing the pellets, we lysed the cells in 1mL lysis buffer with protease inhibitors (0.5% SDS, 10mM EDTA, 50mM Tris-HCL pH 8.0) on ice for 10 minutes.Lysates were sonicated at 70% amplitude for 15 seconds on and 45 seconds off for 4 sets of 20-minute cycles.We moved sonicated lysates to 1.5ml tubes and clarified by centrifugation at 14,000rpm for 10 min in 4°C.We diluted 50µL of the supernatant into 760µL ChIP Dilution Buffer (0.01% SDS, 1.1% Fig.1.ZNF143 is degraded and off chromatin within 30 minutes of dTAG V -1-induced degradation.A) Quantitative western blots indicate that FKBPdegron F36V -tagged ZNF143 (lane 2) is expressed at comparable levels to untagged ZNF143 (lane 1).Lanes 2-7 are dilutions of the untreated and tagged cells line.Lanes 8-12 are a time series of degradation.Note that the arrow represents ZNF143 and there is a nonspecific band that is unchanged directly below the arrow.B) A heatmap of all ZNF143 ChIP-seq peaks indicates that ZNF143 is off chromatin after 30 minutes of degradation.C) Iterative de novo motif analysis identified four ZNF143 motifs that are anchored on a central TGGGA sequence.Ninety-five percent of ZNF143 peaks have at least one motif that conforms to a de novo-identified motif with a p-value of 0.0005 of lower.D) The 29 base seqLogo represents the average ZNF143 binding site, where substantial degeneracy is tolerated outside the core TGGGA postion.

Fig. 2 .
Fig. 2. ZNF143 degradation leads to a decrease in chromatin accessibility.A) Only four ATAC-seq peaks increase chromatin accessibility after 30 minutes of ZNF143 depletion and 479 peaks decrease accessibility.B) An example ATAC-seq peak that decreases accessibility overlaps a ZNF143 binding site.C) Ninety-four percent of ATAC peaks that decrease accessibility overlap ZNF143 peaks, indicating that ZNF143 is predominantly involved in maintaining open chromatin.Dong et al. | Mechanisms of ZNF143 activation and repression bioRχiv | 3 Fig. 3. ZNF143 binds in promoters to directly regulate transcription initiation.A) Nascent transcriptomics identifies 182 genes that increase expression (up) and 365 genes that decrease expression (down) after 30 minutes of ZNF143 degradation.Genes that are matched for expression level and unchanged are in dark grey.B) Ninetysix percent of down genes are within 500 bases of a ZNF143 binding site.The up genes are also significantly closer to ZNF143 binding sites compared to genes that are matched for expression level.C) Down genes have ZNF143 binding in the promoter; up genes have no clear pattern of ZNF143 distribution and 66% have no local ZNF143 binding.D) Compartment modeling indicates that ZNF143 regulates initiation and not pause release at direct down target genes.Each point is a down gene and the y-axis values are the changes in rates that most likely explain the data.E) The transcription start sites of direct down target genes tend to change upon ZNF143 degradation.F) We assigned predominant transcription start sites pre/post ZNF143 degradation.The decrease in TSS usage at the control prominent TSS is accompanied by an increase in TSS usage of the prominent dTAG-treated TSS.

Fig. 4 .
Fig. 4. ZNF143 occludes sites of initiation and acts as a molecular roadblock to directly repress genes in cis.A) ZNF143 binds a motif that spans PAF1's transcription start site.PAF1 transcription increases upon ZNF143 depletion.B) ZNF143 binds strongly over the TSS of 4/5 genes that increase transcription upon ZNF143 depletion.C) FIS1 is the up gene from panel B with weak TSS binding of ZNF143 and no discernible changes in TSS usage occur upon depletion, as measured by PRO-seq 5 end pile ups.D) ZNF143 binds directly downstream from the TSS of a different isoform of FIS1.This FIS1 isoform becomes the most prominent FIS1 TSS upon ZNF143 degradation.E&F) ZNF143 degradation from strong binding sites directly downstream of transcription start sites is associated with up genes.G&H) Genic bidirectional transcription at ZNF143 binding sites downstream of up genes significantly decreases at 7/28 up genes.

Fig. 5 .
Fig. 5. ZNF143 occludes downstream sequence elements to repress transcription.A) Transcription factor YY1's motif (MA1651.1) is enriched directly downstream of TSSs.B) ZNF143 depletion reveals an accessible YY1 and DPE that allows usage of a TSS that is 28 bases upstream of the ZNF143 binding site in the LYSMD1 promoter (cTSS: control max TSS position.dTSS: dTAG max TSS position).

Fig. 6 .
Fig.6.ZNF143 may compete with SP1 at overlapping binding sites to facilitate transcription initiation.A) We called sites of bidirectional transcription genome-wide and aligned on their center position(Wang et al. 2019).B) We identified the SP factor family of motifs de novo exclusively in bidirectionally transcribed that increase transcription.The ZNF143 motif was identified de novo in both up and down birdirectionally transcribed regions.C) SP1 binding intensity, as measured by ChIP-seq, increases at a bidirectionally transcribed region that has overlapping SP1 and ZNF143 motifs.D) Nineteen of the bidirectional transcription regions that increase and have ZNF143 binding have SP1 binding and the SP1 sequence motifs that overlap the ZNF143 sequence motifs.E) Thirteen of the nineteen regions from panel (D) increase SP1 binding upon ZNF143 degradation.Dong et al. | Mechanisms of ZNF143 activation and repression bioRχiv | 7 C C A C C T C C T GGG A G A T G T A G T T C C

Fig. 7 .
Fig.7.ZNF143 may compete with other promoter-bound sequence-specific transcription factors to regulate transcription.A) We identified promoter regions within sense and divergent TSSs called from pileups of the 5´ends of PRO-seq reads.SP motifs identified by FIMO(Grant et al. 2011) are enriched in regions between bidirectional start sites.B) After ZNF143 degradation, sixteen genes that increased transcription also had an SP1 ChIP-seq peak and motif overlapping the ZNF143 motif.The corresponding SP1 binding intensity (ChIP) increased for 10 of these genes.C) Other promoter-enriched motifs (Nrf1, bZIP, and ETS) also overlapped ZNF143 motifs in the promoter of ZSCAN5.D) Other motifs that overlap ZNF143 motifs in promoters of up genes have varying levels of motif conformity and overlap.The red asterisk marks an SP motif that overlaps ZNF143 but is outside of an SP1 ChIP peak.Circle size represents amount of overlap with the 29 base ZNF143 motif and circle color represents motif conformity as measured by MAST score quantiles that were calculated by querying the weight matrices against 1 million random sequences.E) The bidirectional regions identified in Figure6Ealso have other promoter-enriched motifs that overlap ZNF143.

Fig. 8 .
Fig.8.The molecular function of ZNF143 is to bind DNA and facilitate initiation.A) ZNF143 binds to the promoter and stimulates transcription at the majority of its target genes.B) ZNF143 binding can cause repression if binding occurs over a TSS, immediately downstream of a TSS, or in competition with other activators.C) ZNF143's repression mechanisms can be further specified depending on the local environment of the ZNF143 binding site.The number of genes in the left barchart from top to bottom is 28, 5, and 28.The right bar chart top to bottom has the following number of genes in each category: 7 (roadblock/bidirectional), 21 (roadblock), 1 (roadblock and TSS occlusion), 4 (TSS occlusion), 2 (complex promoters with no proposed mechanism), 13 (possible competition with sequence-specific promoter factors), 3 (occlude DPE and/or YY1), 7 (competition with SP1 and another factor), and 3 (competition with SP1).Dong et al. | Mechanisms of ZNF143 activation and repression bioRχiv | 9 * A * GAAGCCATCAGAATAGCGTCTAGAATCCAACAAGGAGAAACGCCAGGGCTTGACGACGGTGGCGGTGGCTCGGGC-3Ŕ everse 5´-G * A * TTAATCATCCAACCCTGGCGTTTCTCCTTGTTGGATTCTAGACGCTATTCTCAGGCACCGGGCTTGCGGGT-3T able 1. Homology directed repair primers.These primers amplify FKBP F36V -2xHA-P2A-Puro with 50 bases of ZNF143 homology.The asterisks mark phosphorothioate bond modifications to minimize degradation.Dong et al. | Mechanisms of ZNF143 activation and repression bioRχiv | 11 Fig. S1.ZNF143 is rapidly depleted from chromatin after 30 minutes of dTAG treatment and we precisely define the ZNF143 binding site with motif analysis.Motif analysis found that all ZNF143 ChIP-seq peaks have a ZNF143 sequence that conforms to the Figure1Dcomposite motif with a p-value of 0.13 or less, with 93% of peaks at a p-value less than 0.0005.
(Bailey and Gribskov 1998)2)adapt(Martin 2011) and aligned to the hg38 genome assembly with bowtie2(Langmead and Salzberg 2012).We sorted aligned reads with samtools and used seqOutBias to generate bigWig files(Li et al. 2009; Martins et al. 2018).We called peaks with macs3(Zhang et al. 2008).We quantified peak intensities by querying bigWig files in 400 base pair windows centered on peak summits and normalized the counts with DESeq2 size factors (SP1 ChIP) or read depthcalculated size factors (ZNF143 ChIP)(Love et al. 2014).We found de novo motifs with MEME(Bailey et al. 2015b) and later STREME (Bailey 2021) through rounds of iterative motif analysis whereby we used MAST(Bailey and Gribskov 1998)to identify and remove the most common motifs for SP1 or ZNF143 until we stopped finding significant de novo motifs.