THAP11F80L cobalamin disorder-associated mutation reveals normal and pathogenic THAP11 functions in gene expression and cell proliferation

Summary

Twelve human THAP proteins share the THAP domain, an evolutionary conserved zinc-finger DNA-binding domain. Studies of different THAP proteins have indicated roles in gene transcription, cell proliferation and development. We have analyzed this protein family, focusing on THAP7 and THAP11. We show that human THAP proteins possess differing homo -⁠ and heterodimer formation properties and interaction abilities with the transcriptional co-regulator HCF-1. HEK-293 cells lacking THAP7 were viable but proliferated more slowly. In contrast, HEK-293 cells were very sensitive to THAP11 alteration. Nevertheless, HEK-293 cells bearing a THAP11 mutation identified in a patient suffering from cobalamin disorder (THAP11_F80L) were viable although proliferated more slowly. Cobalamin disorder is an inborn vitamin deficiency characterized by neurodevelopmental abnormalities, most often owing to biallelic mutations in the MMACHC gene, whose gene product MMACHC is a key enzyme in the cobalamin (vitamin B₁₂) metabolic pathway. We show that THAP11_F80L selectively affected promoter binding by THAP11, having more deleterious effects on a subset of THAP11 targets, and resulting in altered patterns of gene expression. In particular, THAP11_F80L exhibited a strong effect on association with the MMACHC promoter and led to a decrease in MMACHC gene transcription, suggesting that the THAP11_F80L mutation is directly responsible for the observed cobalamin disorder.

Keywords:

Gene expression – DNA-binding proteins – Mutation – Sequence motif analysis – Immunoprecipitation – Cell proliferation – Cobalamins – Transcriptional control

Introduction

The THAP family of gene paralogs encodes 12 proteins in human, named THAP0 to THAP11. These are defined by their N-terminal THAP (for Thanatos—referring to the Greek God of Death—Associated Proteins) domain, an atypical zinc-finger DNA-binding domain [1,2]. THAP domains display similar three-dimensional structures, while recognizing different DNA target sequences [3–7]. Studies of different THAP proteins have indicated roles in gene transcription, cell proliferation and development. Here, we focus on two members: THAP7 and THAP11.

THAP7 represses gene transcription, both by promoting histone deacetylation and by masking histone tails from histone acetyltransferase complexes [8,9]. It is also implicated in the control of cell proliferation by abrogating Histone Nuclear Factor P (HiNF-P)-mediated activation of histone H4 gene transcription [10].

THAP11 (known as Ronin in mice, [11]) also regulates mammalian cell proliferation, with reported examples for both activation [11,12] and repression [13,14], and binds numerous promoters of genes involved in cell growth, metabolism, and cell cycle [12–17]. It also has essential transcriptional roles in pluripotency [11,18,19], hematopoiesis [20], and early development of retina [21], heart [17], and brain [22]. Furthermore, THAP11 is important for mitochondrial function by regulating nuclear mitochondrial-related genes, including components of the electron transport chain [21]. Consistent with the frequent involvement of THAP proteins in human diseases (e.g., dystonia 6 (THAP1, [23–25]), heart disease (THAP5, [26]), or cancers [7,12–14,27–32]), a THAP11 missense mutation is associated with cobalamin disorder, an inborn vitamin deficiency characterized by neurodevelopmental abnormalities [22].

Different human mutations can result in disorders of intracellular cobalamin (vitamin B₁₂) metabolism, defining several complementation groups. The cblC group, accounting for 80% of cobalamin-disorder patients, is owing to biallelic mutations in the MMACHC gene, whose gene product MMACHC is a key enzyme in the cobalamin metabolic pathway [33]. Interestingly, THAP11, together with its co-regulator HCF-1 (for herpes simplex virus Host Cell Factor 1), regulates the transcription of the MMACHC gene [5,22]. Indeed, certain mutations in the X-linked HCFC1 gene encoding HCF-1 are responsible for the so-called cblX complementation group and result in both decreased MMACHC mRNA and MMACHC protein levels [33]. A cobalamin-deficiency patient has been recently described with a clinically cblX-like disorder but without either any MMACHC or HCFC-1 mutation. Instead, this patient had a homozygous missense mutation in the THAP11 gene resulting in a phenylalanine to leucine change at position 80 (referred to here as THAP11_F80L) at the C-terminus of the THAP domain [22,33]—the causal link between this mutation and the cobalamin disorder, however, remains to be established.

Here, we analyze THAP7 and THAP11 interactions and probe their roles in cell proliferation, including through the study of the THAP11 F80L mutation.

Results

Structure of the human THAP family of proteins

To initiate the study of human THAP proteins, we examined their structural and evolutionary relationships. Fig 1 shows an evolutionary tree of the 12 human THAP proteins sorted by THAP-domain sequence similarity shown in S1 Fig. Although sharing the structurally-similar THAP domain, the distinct THAP proteins display considerable variability with sizes extending from 213 to 903 amino acids. Two other features shared by most of them are a so-called ^D/_EHxY HCF-1-Binding Motif (HBM; [34,35]) sequence for HCF-1 interaction (Fig 1, orange and dashed-orange lines) and a coiled-coil domain involved in protein homo -⁠ and heterodimer formation [36]. It has been argued that all [4] or all but one (THAP10) [37] of the 12 THAP proteins contain a coiled-coil domain. Nevertheless, in our analysis with two independent prediction tools (COILS [38] and PairCoil2 [39]), we detected a coiled-coil domain in only nine of the 12 THAP proteins; we did not detect a coiled-coil domain in THAP0, 9 and 10 (Fig 1, blue boxes). These two motifs, HBM and coiled-coil, appear together in seven of the 12 THAP proteins with the HBM invariably within 1 to 13 amino acids N-terminal of the coiled-coil domain, suggesting a functional evolutionary relationship.

In an extended study of THAP proteins [40], we analyzed the propensity for homodimer and heterodimer formation among THAP4, THAP5, THAP7, THAP8 and THAP11 proteins as well as their abilities to associate with HCF-1 via the HBM. In the context of this study, we decided to focus on two THAP proteins broadly expressed (see [40]) and having reported roles in transcriptional regulation [8,9,11]: THAP7 and THAP11.

THAP7 and THAP11 form homodimers but not heterodimers

We first probe the formation of THAP7 and THAP11 homodimers and THAP7/THAP11 heterodimers. For this purpose, HEK-293 cells were co-transfected with Flag -⁠ and/or HA-epitope-tagged THAP7 and THAP11 expression constructs, and whole-cell lysates (wcl) subjected to anti-HA-epitope immunoprecipitation and analyzed by immunoblot. In Fig 2A, we test the ability of THAP7 to form homodimers. Interestingly, the ectopically synthesized THAP7 protein appears as a series of bands upon immunoblotting (lane 3, lower panel), probably owing to phosphorylation (S2A Fig). As evidenced by the recovery of Flag-tagged THAP7 protein after immunoprecipitation of co-synthesized HA-tagged THAP7 protein (compare lanes 8 and 4, lower panel), THAP7 forms homodimers effectively. Interestingly, the slower migrating form —⁠ and probably most heavily phosphorylated—is less effectively recovered than the faster migrating forms, suggesting that phosphorylation modulates THAP7 homodimer formation.

As shown in Fig 2B, THAP11 protein also forms homodimers in this assay (compare lanes 4 and 8, lower panel), consistent with previous reports [11,41]. Although each forms homodimers, THAP7 and THAP11 do not form heterodimers as evidenced by the failed recovery of Flag-tagged THAP7 after immunoprecipitation of co-synthesized HA-tagged THAP11 (compare lanes 4 and 8 of Fig 2A and 2C). The segregated abilities of THAP7 and THAP11 to form homodimers but not heterodimers suggest that, although involved in similar transcriptional and cell proliferation processes, they do not directly influence each other's transcriptional networks via dimer formation.

THAP7 and THAP11 HBM sequences enable them to associate with HCF-1

Using an analogous epitope-tagged protein association assay as in Fig 2, we probed THAP7 and THAP11 association with HCF-1 as shown in Fig 3. HCF-1 is proteolytically processed into associated N -⁠ (HCF-1_N) and C -⁠ (HCF-1_C) terminal subunits [42,43]. As HCF-1 associates with HBM-containing proteins via its N-terminal Kelch domain [44], we tested THAP7 and THAP11 association with HCF-1_N. As shown in Fig 3, both wild-type THAP7 and THAP11 associate with wild-type HCF-1_N as indicated by the recovery of HA-tagged HCF-1_N after immunoprecipitation of co-synthesized Flag-tagged THAP7 (Fig 3A) and THAP11 (Fig 3B) (compare lanes 4 with lanes 2; recovery of endogenous THAP11 in such an experiment is shown in S2B Fig).

To characterize the HBM–Kelch domain contribution to these THAP-protein–HCF-1 interactions, we (i) took advantage of the well-characterized HCF-1 Kelch-domain point mutation called P134S, which prevents its association with HBM-containing target proteins [44], and (ii) created HBM-disruptive point mutants in THAP7 and THAP11. Each mutation on its own impairs the THAP-protein–HCF-1_N association (Fig 3A and 3B, compare lanes 4 with lanes 6 and 10) and when in combination further disrupt the interaction (lane 12). Consequently, these two THAP proteins both bind HCF-1 via the well-characterized interaction between the HCF-1 Kelch domain and HBM.

Thus, THAP7 and THAP11 form homodimers likely via their respective coiled-coil domains and associate with HCF-1 using their respective HBM sequences. Although these two interaction surfaces lie in proximity to one another in THAP proteins (e.g., by 9 amino acids in THAP11), at least one HBM in a double HA and Flag epitope-tagged THAP11 "heterodimer" can be disrupted without preventing THAP11 dimer formation as shown in Fig 2B (compare lanes 10 with 8); thus THAP-protein dimer formation may not be dependent on HCF-1 association.

THAP7 promotes cell proliferation

To study the impact of THAP7 on cell proliferation, we engineered multiple homozygous clones of each of three point-mutant HEK-293 cell lines using CRISPR/Cas9 mediated genome editing [45,46] (S3 Fig and Materials and Methods): (i) a likely THAP7 null mutant (THAP7_null), in which we created nonsense mutations at both codons 13 and 14 of the THAP7 coding sequence to disrupt THAP7 synthesis (three clones, S3A Fig); (ii) the aforementioned THAP7 HBM mutant (THAP7_HBM) (two clones, S3B Fig), and (iii) a THAP7 mutant C-terminally truncated after the HBM via two sequential nonsense mutations, thus deleting the coiled-coil domain (THAP7_ΔCC) (four clones, S3C Fig).

The three THAP7_null mutant cell lines were viable albeit cell numbers increased more slowly than for the parental cells (labeled THAP7_WT in Fig 4). As there was little, if any, difference in viability (S4A Fig), THAP7 is apparently important for HEK-293-cell proliferation (i.e., cell-cycle progression). The two THAP7_HBM mutant clones proliferated at different rates (S5A and S4B Figs): one as the parental cells and the other as the THAP7_null mutant cells preventing any conclusion to be drawn regarding the impact of the HBM mutation on THAP7 function. In contrast, the four THAP7_ΔCC mutant clones (S5B and S4C Figs) consistently exhibited a slowdown in cell proliferation similar to that of the THAP7_null cells (compare Fig 4 and S5B Fig), suggesting that the THAP7 coiled-coil domain, and thus likely its ability to form dimers, is critical for the THAP7 function observed in HEK-293 cell proliferation.

Translation of the THAP11_F80L cobalamin-disorder-associated mutation into a cellular model

We attempted to obtain THAP11 mutant cells corresponding to those described for THAP7, but no THAP11_null or THAP11_ΔCC cell clone was obtained. Furthermore, only one heterozygous THAP11_HBM mutant clone was obtained (S6A Fig) and these cells failed to survive. These results suggest that THAP11 is necessary for HEK-293-cell viability and that both the HBM and coiled-coil domain are important for THAP11 function.

During the course of this study, a human THAP11 mutation was identified in a patient with cobalamin disorder [22]. We thus took advantage of this finding to focus, as we do in the remainder of this study, on the clarification of the functional consequences of this cobalamin-disorder-associated THAP11_F80L mutation. The THAP11 F80 phenylalanine residue is highly conserved across vertebrate species [22] and also among the 12 different human THAP proteins (S1 Fig). As labeled in blue in S1 Fig, it is at the last position of the so-called “AVPTIF” box (green box in S1 Fig) lying at the C terminus of the THAP-domain; this box has been suggested to promote proper folding of the DNA-binding zinc finger [3,4]. We managed to obtain a single clone bearing the homozygous THAP11_F80L-encoding C240G mutation (S6B Fig); an attempt to obtain a second clone was unsuccessful. Although still proliferating, the mutant cells increased in cell number at less than one-half the rate of the parental cells (Fig 5). As for the THAP7 mutant cells, we observed little, if any, change in viability (S7 Fig), suggesting that THAP11 is important in sustaining HEK-293-cell proliferation (i.e. cell-cyle progression).

To determine the defect(s) caused by the THAP11_F80L mutation, we performed both THAP11-chromatin immunoprecipitation followed by high-throughput sequencing (ChIP-seq) and high-throughput RNA sequencing (RNA-seq) with both parental and THAP11_F80L mutant cells. Using this information, we first analyzed the effects of the THAP11_F80L mutation on THAP11 itself. As previously described [5], THAP11 binds to its own promoter at two near-by sites (overlapping THAP11-associated motifs; see below) of which only one was affected by the THAP11_F80L mutation (Fig 6A, track 2 and 4). Despite this difference in THAP11-promoter binding, consistent with what has been observed in fibroblasts from the THAP11_F80L patient [22], THAP11_F80L mRNA levels were unchanged relative to the THAP11_WT levels (Fig 6B). In contrast, the levels of THAP11_F80L protein were reduced relative to the THAP11_WT protein (Fig 6C). The difference in signal is unlikely owing to an altered recognition of the mutant protein by the antibody, because the antibody used recognizes the last 85 amino acids of THAP11 distal to the F80L mutation. This difference in protein level could be explained by protein destabilization as the mutation affects the “AVPTIF” box implicated in THAP-domain stability. Whichever the case, the levels of THAP_F80L protein in the mutant cells does not obviously affect THAP11 gene transcription.

The THAP11_F80L mutation selectively affects THAP11 DNA association

A global view (see Materials and Methods for analysis parameters) of the THAP11 ChIP-seq results in parental and THAP11_F80L HEK-293 cells revealed 2341 THAP11_WT and only 1473 THAP11_F80L peaks, with about half of the peaks in each case being located within +/ –⁠ 250 bp of an annotated transcription start site (TSS) (Fig 7A and S1 Table). S2 Table shows that the 1114 TSS-associated THAP11_WT peaks belong to expressed genes most prominently involved in transcription and its regulation, indicating the THAP11 controls programs of transcription, consistent with what has been previously reported [5].

Being particularly interested in THAP11 regulation of transcription and to focus the study, we present the analysis of the TSS-associated THAP11 peaks. To characterize these peaks, we classified them into three categories (Fig 7B): (i) peaks present in both the THAP11_WT parental and the THAP11_F80L mutant samples called “common”, (ii) peaks missing (i.e., falling below the peak-calling threshold) in the THAP11_F80L sample called “F80L-absent”, and (iii) peaks present exclusively in the THAP11_F80L sample called “F80L-only”. As shown in the Venn diagram in Fig 7B, only three TSS-associated peaks fell into the “F80L-only” category (these peaks had a low significance score and were attributed to noise in the data; they were subsequently ignored), whereas 380 TSS-associated peaks were in the F80L-absent category—fully one-third of the total 1114 THAP11_WT peaks. Thus, the THAP11_F80L mutant protein binds to only a subset of the THAP11_WT promoter targets and does no exhibit any evident de novo promoter binding compared to the THAP11_WT protein.

Fig 7C illustrates THAP11_WT vs. THAP11_F80L DNA-site specificity with a 3.3 Mb view of Chromosome 1 that includes the MMACHC gene. The view covers nine THAP11_WT TSS-associated peaks (1–9)—of which four (peaks 1, 5, 7 and 8) were selected for close-ups to illustrate the points illuminated below—and one non-TSS-associated peak (labeled *): two TSS-associated peaks (peaks 7 and 8) and the one non-TSS-associated THAP11 peak fell into the THAP11_F80L-absent category and the remainder in the common category (see S3 Table). Among those in the common category, we note that some peaks (e.g., peak 1 in Fig 7C) remain largely the same size between the THAP11_WT and THAP11_F80L samples, whereas others (e.g., peak 5) were smaller albeit still present in the THAP11_F80L sample. In the subsequent analyses described below, we do not distinguish between these two common peak subcategories. In conclusion, the THAP11_F80L mutation results in a selective disruption of THAP11 DNA binding at specific promoters in HEK-293 cells, and does not create de-novo THAP11 promoter-binding sites.

Restricted DNA-sequence recognition by the THAP11_F80L protein

To examine more broadly the nature of the effect of the THAP11_F80L mutation, we calculated size scores for each peak (see Materials and Methods for peak score determination) and plotted separately the distribution of the common and F80L-absent peak scores for both the THAP11_WT parental and THAP11_F80L mutant samples (Fig 8A). Consistent with their a priori categorization, the THAP11_F80L peaks scored higher in the common group than in the F80L-absent group (dashed lines). This tendency was also reflected at the level of the THAP11_WT scores as the common peaks (dark-red solid line) scored higher than the F80L-absent peaks (light-red solid line). The peak-score difference between the THAP11_WT and THAP11_F80L common peaks (dark-red lines) was less pronounced, however, than the difference between the THAP11_WT and THAP11_F80L F80L-absent peaks (light-red lines). Together, these observations suggest that: (i) the F80L-absent peaks represent lower affinity THAP11_WT binding sites and (ii) binding to these sites is hypersensitive to the THAP11_F80L mutation.

To understand the reason for these differences, we examined the DNA sequences underlying the common and F80L-absent peaks. The most prominent motif to emerge from this analysis of both sets of peaks was the THAP11-associated motif (TAM) also referred to as RBM [5] or SBS2 [47]. We conclude that the DNA sequences under both common and F80L-absent peaks are enriched for the TAM, and thus that the difference between the two sets of peaks may result from subtle differences among TAMs.

We probed this possibility by separately generating TAM consensus sequences (Fig 8B) for the closest TAM associated with either the common or the F80L-absent peaks. Of the 22 bp-selected TAM sequence shown in Fig 8B, the common-peak-associated one (top) displays two prominent areas of consensus: a 5' CTGGGA sequence and a 3' TGTAGT sequence. The 5' CTGGGA consensus was as prominent among the TAMs of the F80L-absent peaks, but the 3' TGTAGT consensus was significantly weaker (Fig 8B, bottom). Taken together with the peak score analysis in Fig 8A, this comparison suggests that the THAP11_F80L protein is more sensitive to changes in the 3' half of the consensus than the THAP11_WT protein.

In addition to a weaker TAM consensus sequence associated with the F80L-absent peaks, fully one-quarter of the F80L-absent peaks had no discernible nearby TAM (i.e., within 1000 bp on each side of the peak maximum), whereas essentially all common peaks were associated with one or more TAMs (Fig 8C). These analyses further emphasize the importance of a strong TAM consensus sequence for THAP11_F80L promoter recognition.

In addition to examining the TAMs of THAP11-peak-containing promoters, we asked if the nature of the genes associated with the common and F80L-absent peaks differ through gene-ontology (GO) analysis as summarized in Fig 8D (see S4 Table). Indeed, whereas the common peaks were prominently associated with transcription-regulatory genes, the association makeup of the F80L-absent-peak group was more heterogeneous and no longer evidently associated with transcription. Thus, a core function of THAP11—to regulate programs of transcription—is apparently sufficiently robust to resist the effects of the THAP11_F80L mutation; it may be disruption of such a core function of THAP11 that makes the aforementioned THAP11_null, THAP11_HBM and THAP11_ΔCC mutations lethal.

Altered patterns of gene expression in THAP11_F80L cells

To examine the consequences of the THAP11_F80L mutation at the gene-expression level we analyzed the RNA-seq results (S5 Table). Fig 9A plots the transcript levels for protein-encoding genes in THAP11_WT cells (X axis) against their fold change in THAP11_F80L cells (Y axis). More often transcript levels considered to be changing (see fold change and p-value cutoffs in Materials and Methods) had decreased in THAP11_F80L cells compared to THAP11_WT cells: 523 genes were downregulated (red) whereas 279 genes were upregulated (blue). Furthermore, the magnitudes of decrease were greater than those for increases. Thus, the THAP11_F80L mutation apparently affects activation of transcription—either directly or indirectly—more often than repression.

Among the 14 genes most downregulated in the THAP11_F80L cells, fully one-half were canonical S-phase histone-encoding genes (Fig 9A, right). In our THAP11_WT ChIP-seq data, the THAP11 protein was absent from canonical histone-encoding gene promoters (S1 Table) indicating that this downregulation is an indirect effect. Such downregulation may be linked to the slower proliferation rate of THAP11_F80L cells, either as a cause or a consequence.

Separate GO analyses of the upregulated and downregulated genes (Fig 9B and S6 Table) revealed enrichment for different gene functions. The downregulated set was more heterogeneous with weaker enrichments for each GO term than the upregulated set. Consistent with the aforementioned histone-gene expression change, it included nucleosome assembly and chromatin silencing terms. In contrast, the upregulated set was associated with only two GO terms that fell below the p-value cutoff—these being both transcription oriented and with highly significant p-values. Thus, whereas more genes were downregulated in THAP11_F80L cells, those that were upregulated were strongly enriched for a single biological process, transcription. Although the THAP11_F80L ChIP-seq results indicated that THAP11 association with the TSS of transcription-related genes was overall resistant to the F80L mutation, the THAP11_F80L mutation appears to nevertheless affect THAP11 regulation of gene-transcription programs.

Likely direct roles of THAP11 affected by the THAP11_F80L mutation

To probe which of these gene-transcription changes might be direct consequences of THAP11_F80L function, we identified those changing genes whose promoters were directly bound by THAP11_WT but not THAP11_F80L (i.e., F80L-absent ChIP-seq peaks). These genes included 11 upregulated and 44 downregulated genes as shown in the Fig 10 heat map (see also S3 Table for MMACHC and MUTYH). By GO analysis (S7 Table), we observed that a significant portion (one quarter) of these direct-effect genes are associated with regulation of transcription (Fig 10, blue arrows). This observation suggests that THAP11 plays a higher-order role in the regulation of gene transcription by directly regulating—most often activating—the expression of secondary transcriptional regulator-encoding genes.

Fig 10 also shows that the MMACHC gene ranks third among the most downregulated direct-target genes in THAP11_F80L cells (red arrow). This observation suggests that THAP11 activates MMACHC gene transcription and that the THAP11_F80L mutation impairs this activation, consistent with the cobalamin-disorder phenotype of the patient from whom the mutation was identified [22].

Discussion

We have studied the THAP family of DNA-binding proteins extending from an evolutionary perspective of the 12 human members to a detailed analysis of a human disease associated THAP11 mutation.

The THAP family of proteins

The sequence comparison in Fig 1 accentuates how the THAP family is rapidly evolving—even just between human and mouse there are five additional THAP genes in human. Nevertheless, two non-THAP domain sequence elements, the coiled-coil domain and HBM, are shared among the large majority of the 12 human THAP proteins, illuminating their importance for this class of proteins.

The coiled-coil domain is implicated in association of THAP proteins [11,48,49]. As described previously [11,41] and reproduced here, THAP11 forms homodimers. We show that THAP7 also forms homodimers. In contrast, we did not detect human THAP7 and THAP11 heterodimer formation, and thus did not observe a previously reported THAP7 association with an N-terminal THAP11 protein fragment [11]. A wider study (S8 Table and [40]) has shown homodimer formation for THAP8 and heterodimer formation between THAP8 and THAP11 but not heterodimer formation between THAP4 and THAP11 proteins. Thus, THAP homo -⁠ and heterodimer formation possibilities are likely complex, permitting complicated transcriptional regulatory networks. This complexity is probably amplified by the rapid evolution of divergent DNA recognition properties of individual THAP domains [3–7].

The HBM permits many THAP proteins to extend their regulatory network to association with the transcription co-regulator HCF-1. We have shown that both THAP7 and THAP11 interact with HCF-1 via their respective HBM. Nevertheless, the presence or absence of an HBM in a THAP protein does not determine whether it will or will not associate with HCF-1: THAP5 contains an HBM but failed to associate with HCF-1, and THAP8 lacks an HBM and yet associated with HCF-1 (S8 Table and [40]). These results further underscore the complex regulatory potential of the THAP-protein family.

Roles of THAP7 and THAP11 in cell proliferation and viability

To probe regulatory roles of THAP proteins, in particular in cell proliferation, we designed targeted mutations for the THAP7 and THAP11 genes in HEK-293 cells to create null mutations and HBM and coiled-coil deficient proteins. For THAP7, we obtained all three categories of mutation, in contrast to THAP11 where we only transiently obtained a heterozygous HBM mutant. The sensitivity of HEK-293 cells to THAP11 null mutagenesis is consistent with the embryonic lethality observed upon mouse Thap11 (Ronin) deletion [11]. Indeed, THAP11 is even sensitive to individual HBM and coiled-coil disruption, indicating critical roles of THAP11 interaction with HCF-1 and dimer formation for cell viability.

Although THAP7 was not essential for HEK-293-cell viability, we observed consistent retarded cell proliferation for the THAP7_null and THAP7_ΔCC mutant cells. As we did not observe a decrease in viability, the decrease in cell accumulation probably reflects a defect in cell-cycle progression. Thus, these two THAP proteins appear to play important and in one case—THAP11 —⁠ an essential role in HEK-293-cell viability.

The THAP11_F80L cobalamin-disorder associated mutation

We observed that THAP11 is particularly sensitive to mutagenesis, as we did not obtain any viable THAP11_null, THAP11_HBM or THAP11_ΔCC mutants. This mutational sensitivity is also highlighted by the ExAC database of human exon sequence variants (http://exac.broadinstitute.org/gene/ENSG00000168286, [50]) where (i) the only six loss-of-function mutations reported are all located in the THAP11 poly-glutamine tract-encoding sequences and (ii) no missense mutations were reported in the THAP11 sequences encoding the THAP domain. This restricted number of genetic variants altering the THAP11 protein highlights the importance of THAP11 functions. And yet an individual homozygous for the THAP11_F80L mutation survived for 10 years [22], showing how a THAP11 mutation, located in the THAP domain, can still be tolerated for viability.

We took advantage of this finding to study the function of a natural human THAP11 mutant. We observed decreased cell proliferation in HEK-293 THAP11_F80L mutant cells. We also noted that, although THAP11_F80L and THAP11_WT mRNA levels were similar, there was less THAP11_F80L protein in the mutant cells compared to THAP11_WT protein in unmutated cells. The THAP11_F80L mutation affects a residue of the “AVPTIF” box, which is suggested to contribute to proper folding of the zinc finger [3,4]. The location of the mutation and the lower levels of THAP11_F80L protein—without concomitant decrease in THAP11_F80L mRNA—suggest that the mutant THAP11_F80L protein is destabilized and more rapidly degraded.

To investigate the consequences of the THAP11_F80L mutation, we probed THAP11 DNA binding using THAP11 ChIP-seq. We observed no significant new THAP11 TSS-associated-binding sites, but did observe that the mutation causes the loss of THAP11 DNA binding at a specific subset of TSS-associated sites. A detailed analysis of the altered binding-site patterns revealed that some TSS-associated sites are particularly sensitive to the THAP11_F80L mutation. Overall, these are sites exhibiting a lower affinity for THAP11 binding, with a weaker THAP11 motif consensus sequence —⁠ particularly at the 3’ half of the consensus sequence. These observations suggest that THAP11-binding sites come in two classes that may respond to wild-type THAP11 activity differently, for example by being more or less sensitive to the activity of regulatory co-factors or dimer formation.

The analysis of how the THAP11_F80L mutation affects gene expression revealed alterations in the regulation of programs of transcription. The gene-expression changes can be direct or indirect consequences of the THAP11_F80L mutation and subsequent changes in DNA binding. For instance, the impaired expression of numerous canonical S-phase histone-encoding genes in THAP11_F80L cells is likely the result of an indirect effect of the mutation, as THAP11 is absent from the corresponding histone promoters. Indeed, the decrease of histone-gene expression could be the result, not the cause, of the decrease in cell proliferation observed in these mutant cells. Alternatively, the THAP11_F80L mutation and the alteration of THAP11 DNA-binding activity could have direct consequences for the expression of specific genes, apparently especially for genes implicated in the regulation of transcription.

Linking the THAP11_F80L mutation with human disease

Consistent with identification in a patient with cobalamin disorder [22], the THAP11_F80L mutation also has an apparent direct effect on the activity of the MMACHC promoter. Whereas THAP11_WT binds to the MMACHC promoter, the THAP11_F80L mutant binds poorly to the MMACHC promoter, consistent with the reduction in MMACHC mRNA levels observed in the fibroblasts of the patient in which the mutation was isolated [22]. These results suggest a direct link, missing so far, between the THAP11_F80L mutation and the cobalamin disorder observed in the associated patient.

Interestingly, the corresponding F-to-L missense mutation has also been identified in human THAP1, also named DYT6. This THAP1_F81L mutation is one of the THAP1 mutations associated with dystonia 6 [51]. Dystonia 6, a subtype of primary monogenic torsion dystonia, is a movement disorder involving dysfunctions of the central nervous system and characterized by involuntary muscle contractions. It is caused by a variety of THAP1 mutations (90 to date) spread over the entire THAP1 gene [23–25,52,53]. Studies are conflicting as to the effect of the THAP1_F81L mutation on THAP1 DNA binding—with either little, if any [54], partial [51], or apparently complete [55] loss of THAP1-DNA binding described, suggesting that the effect of the mutation on DNA binding is sensitive to the conditions for analysis. For example, as for THAP11_F80L, the THAP1_F81L protein is destabilized [54], which could influence DNA-binding affinity in vivo. Thus, the matching mutations in two distinct THAP proteins with different gene targets and cellular roles [3–7], have similar effects on the protein—consistent with the identical disruption of a highly conserved protein segment—and yet very different phenotypic and clinical effects.

The different results presented here suggest a general model for how the THAP11_F80L mutation is pathogenic: the mutation affects THAP11 protein levels and perhaps, by virtue of this position in the DNA-binding THAP domain, its DNA-binding activities. Either of these attributes could lead to the observed bimodal specificity of promoter occupancy by THAP11_F80L protein. We imagine that often the less-abundant mutated THAP11_F80L protein retains binding activity for DNA sites of strongest affinity—e.g., the ones with the strongest TAM consensus sequence.

This general model, however, is unlikely to explain all the effects of the THAP11_F80L mutation, because some promoters, in particular the MMACHC promoter, can be greatly affected by the THAP11_F80L mutation but possess TAMs with good matches to the TAM consensus sequence. And yet it is precisely the great sensitivity of the MMACHC promoter to the THAP11_F80L mutation that likely accounts in large part for the associated cobalamin-disorder phenotype, with other gene expression changes likely accounting for additional clinical manifestations that provide the specificity of this cobalamin-disorder subtype [33].

Importantly, the results described here indicate a direct role of the THAP11_F80L mutation in MMACHC gene expression and thus in the cobalamin disorder diagnosed in the patient.

Materials and methods

Detailed materials and methods are described in Supporting Information S1 File.

Cell proliferation assays

Cells were seeded at the same density (1.25 x 10⁴ cells per ml) on day 0, and for each cell line, 2 plates used for counting every 24 hours from day 1 to day 8 (except days 2 and 3). THAP7_HBM, THAP7_ΔCC, and parental cell lines were tested in parallel in a first experiment, whereas THAP11_F80L, THAP7_null and parental cell lines were tested in parallel in a second experiment.

Immunoblotting

The samples were separated by SDS-PAGE before being transferred onto a nitrocellulose membrane. After blocking, membranes were incubated overnight at 4°C with the primary antibody. After washing, they were further incubated with the appropriate secondary antibody during 1 hour at room temperature. Blots were finally visualized with the Odyssey R infra-red imaging system (LI-COR).

CRISPR/Cas9 mutagenesis

CRISPR/Cas9-mediated mutagenesis in HEK 293 cells was performed as recommended [56] with minor modifications. After co-transfection of the cells with a GFP, Cas9 and gRNA encoding plasmid together with repair template, cells were cell sorted and single-cell clones were tested by restriction digest. The mutation was further confirmed by sequencing.

RNA extraction and sequencing (RNA-seq)

Total RNA was used to prepare ribosomal RNA-depleted libraries for 50-nucleotide single-read high-throughput sequencing using an Illumina HiSeq 2100 device with 6 samples per lane (multiplexing).

Chromatin immunoprecipitation and sequencing (ChIP-seq)

For each ChIP, 5 IPs were performed in parallel and subsequently pooled, each time by incubating 9 μg of crosslinked and sonicated chromatin at 4°C with 2 μg of anti-THAP11 antibody and further with protein G agarose beads. Pre-IP samples (total chromatin—input sample) were prepared in parallel. 5 ng of each ChIP and input purified DNA were used to prepare paired-end sequencing libraries for 100-nucleotide paired-end high-throughput sequencing (Illumina, HiSeq 2100) with 3 samples per lane (multiplexing).

Bioinformatics

Analysis of THAP proteins

The evolutionary tree (Fig 1) was prepared by multiple sequence alignment (Muscle alignment) on the 90 N-terminal amino-acids of each human THAP protein sequence. The alignment was refined and visualized using the iTOL (interactive Tree Of Life) online tool [57,58].

RNA-seq analysis

Single reads were mapped onto the Hg19 human genome annotation using STAR (Spliced Transcripts Alignment to a Reference, [59]) and read counts and normalized RPKM (Reads Per Kilobase of transcript per Million mapped reads) were calculated using RSEM [60,61]. Only protein-coding genes were considered and genes with an RPKM value below 1.2 in all the samples were considered as not expressed and discarded from further analyses. Differential analyses were performed with DESeq2 [62]. Resulting gene sets were used for Gene Ontology (GO) enrichment analysis [63,64].

ChIP-seq analysis

Paired-end reads were mapped onto the Hg19 human genome annotation using STAR (Spliced Transcripts Alignment to a Reference) [59]. Peaks were detected using the Model-based Analysis of ChIP-Seq (MACS2) tool [65] and tested using the “Origami method” (V. Praz; see [66]), which identified enriched bins. Only MACS-identified peaks intersecting with Origami-enriched bins were kept. A peak was considered TSS associated if one or more underlying nucleotides was located within +/ -⁠ 250 bp of an RNA polymerase II transcription start site. Peak scores were defined as the difference of the IP and input log₂ counts, scaled by total tags for the sample and the peak width. Peaks were visualized with the UCSC genome browser [67].

Supporting information

S1 Raw Images [pdf]

S1 File [pdf]
Extended experimental methods.

S1 Fig [eps]
The THAP domain of human THAP proteins.

S2 Fig [a]
THAP7 is phosphorylated and endogenous THAP11 interacts with HCF-1 via the HCF-1 N-terminal subunit.

S3 Fig [left]
THAP7 CRISPR/Cas9 mutants.

S4 Fig [a]
Effect of the THAP7, THAP7 and THAP7 mutations on HEK-293-cell viability.

S5 Fig [a]
Effect of the THAP7 and THAP7 mutations on HEK-293-cell proliferation.

S6 Fig [left]
THAP11 CRISPR/Cas9 mutants.

S7 Fig [eps]
Effect of the THAP11 mutation on HEK-293-cell viability.

S1 Table [rpkm]
List of ChIP-seq peaks.

S2 Table [xlsx]
THAP11 controls programs of transcription.

S3 Table [pdf]
List of THAP11 peaks around the gene promoter.

S4 Table [xlsx]
Nature of the genes associated with the common and F80L-absent peaks.

S5 Table [rpkm]
RNA-seq data analysis of THAP11 and THAP11 HEK-293 cells.

S6 Table [xlsx]
Nature of the genes differently expressed between the THAP11 and THAP11 cells.

S7 Table [xlsx]
Direct gene expression changes caused by the THAP11 mutation.

S8 Table [pdf]
Summary of the results of the wider study investigating THAP dimerization and interaction with HCF-1.