The dystonia gene THAP1 exhibits a co-essential relationship with the proteasome subunits PSMB5 and PSMB6

Leveraging co-essentially data from the DepMap project⁸, we set out to characterize novel roles for genes involved in the UPS. Focusing on a manually curated set of ~1000 genes implicated in UPS function, we examined co-essential gene relationships derived from genome-wide CRISPR-Cas9 screens across ~1100 different cancer cell lines. Supporting the utility of this approach to identify genetic relationships that are functionally relevant, many of the most significant positive co-essential relationships clustered genes whose products are known to act in multi-protein complexes to facilitate protein degradation (Fig. S1A). For instance, the RNF126 E3 ubiquitin ligase cooperates with BAG6 to target hydrophobic sequences mislocalised to the cytosol for proteasomal degradation²⁴ (Fig. 1A); Cul2, ElonginB/C and the von Hippel-Lindau (VHL) substrate adaptor comprise a Cullin-RING E3 ubiquitin ligase complex responsible for the degradation of hypoxia-inducible factor (HIF)-1α in normoxia (Fig. 1B)²⁵, and the CTLH complex is a multi-subunit E3 ligase orthologous to the yeast GID complex which degrades gluconeogenic enzymes²⁶ (Fig. 1C). Furthermore, several of the most significant negative co-essential relationships define E3 ligase-substrate pairs: for example, MDM2 mediates the degradation of p53²⁷ (Fig. S1B) and the Cul4 substrate adaptor AMBRA1 targets cyclin D^28–30 (Fig. S1C).

Open in new tabFigure 1: Transcriptional regulation of PSMB5 by THAP1 explains their co-essential relationship.

(A-C) Co-essential relationships involving UPS genes predict biological relationships, as exemplified by three E3 ligase complexes: the BAG6 complex (A), Cul2^VHL (B) and the CTLH complex (C). Network diagrams were produced using NetworkX; numbers annotating the edges indicate pairwise correlation coefficients as calculated in⁹.

(D-F) THAP1 exhibits a strong positive co-essential relationship with both PSMB5 and PSMB6 across DepMap data.

(G-H) THAP1 disruption is toxic in HEK-293T cells. Cells were transduced with a lentiviral vector expressing Cas9 and the indicated sgRNAs, followed by puromycin selection to eliminate untransduced cells commencing 48 hours later. A further 48 hours later, cells were counted, plated in equal numbers, and their viability assessed by counting (G) and brightfield microscopy (H). (Scale bar = 100 μm)

(I-J) Ablation of THAP1 decreases PSMB5 expression. HEK-293T expressing Cas9 and the indicated sgRNAs were analyzed by qRT-PCR (I) and immunoblot (J). (*P < 0.05, t-test; ns, not significant)

Our follow-up work focused on the most statistically significant uncharacterized co-essential relationship in the dataset: THAP1 exhibits a highly significant positive association with both PSMB5 and PSMB6 (Fig. 1D-F). THAP1 is a transcription factor which binds DNA in a sequence-specific manner using a THAP-type zinc-finger domain, while PSMB5 and PSMB6 encode catalytic subunits of the proteasome core particle. Thus, we set out to test the hypothesis that the co-essential relationship between THAP1 and PSMB5/6 could be explained by an essential role for THAP1 in regulating the expression of catalytic proteasome subunits.

Loss of THAP1 abrogates PSMB5 transcription

Lentiviral expression of Cas9 and CRISPR sgRNAs targeting THAP1 in HEK-293T cells was extremely toxic (Fig. 1G-H), consistent with DepMap data which demonstrates that knockout of THAP1 is broadly deleterious across cancer cell lines⁸. However, at day 5 post-transduction, before the onset of significant cell death, we found substantially reduced levels of PSMB5 transcripts by quantitative reverse transcription PCR (qRT-PCR) (Fig. 1I). In contrast, we observed no reduction in the expression of either PSMB6 or PSMB7, the other catalytic subunits of the proteasome (Fig. 1I). Concordantly, we also observed reduced abundance of PSMB5 protein as assessed by immunoblot (Fig. 1J). Thus, these data suggest that THAP1 is required to maintain basal levels of PSMB5 transcription.

Lethality resulting from loss of THAP1 can be rescued by exogenous PSMB5

Next, we sought to test the hypothesis that the essentiality of THAP1 is due to its role in activating PSMB5 expression. Should this be the case, we reasoned that, irrespective of any reduction in the expression of endogenous PSMB5, an exogenous source of PSMB5 should rescue cell viability upon THAP1 ablation. Strikingly, unlike their wild-type counterparts, HEK-239T cells transduced with a lentiviral vector expressing PSMB5 did not display any significant growth defect following CRISPR/Cas9-mediated targeting of THAP1 (Fig. 2A and Fig. S2A). Exogenous expression of PSMB6, however, was incapable of rescuing viability following THAP1 ablation (Fig. 2A and Fig. S2B). Thus, the toxicity that results from loss of THAP1 is due to insufficient PSMB5 expression, explaining the molecular basis for their co-essential relationship. In contrast, we found no evidence to support a direct relationship between THAP1 and PSMB6, suggesting that their co-essential relationship arises indirectly through their shared relationship with PSMB5.

Open in new tabFigure 2: Lethality resulting from THAP1 loss can be rescued by exogenous expression of PSMB5.

(A) Exogenous expression of PSMB5 rescues cell viability upon THAP1 ablation. HEK-293T cells were first transduced with lentiviral vectors expressing either PSMB5 or PSMB6; then, following transduction with Cas9 and the indicated sgRNAs, cell numbers were monitored over time.

(B) Loss of THAP1 is broadly toxic across cell types. CRISPR/Cas9-mediated ablation of THAP1 adversely affects the viability of 946/1100 cancer cell lines (blue dots, representing effect scores < -0.25) examined by DepMap.

(C-D) The toxicity associated with THAP1 ablation is rescued by exogenous PSMB5 expression in HeLa cells (C) and A549 cells (D).

(E-H) Like PSMB5, expression of PSMB8 also protects against the toxic effects of THAP1 loss. THP-1 cells do not exhibit any substantial growth defect following THAP1 ablation (E). High levels of PSMB8 expression are observed in the cell lines whose growth is not significantly affected by THAP1 loss (orange dots, representing effect scores >-0.25) in DepMap data (F), and THP-1 cells strongly express PSMB8 as assessed by qRT-PCR (G). Exogenous expression of PSMB8 can rescue the viability of HEK-293T cells following THAP1 disruption (H).

DepMap data demonstrates that disruption of THAP1 is broadly lethal across cancer cell lines (Fig. 2B). Thus, to generalize our finding that the essential requirement for THAP1 is to facilitate PSMB5 expression, we ablated THAP1 in three additional human cell lines: HeLa, A549 and THP-1. Mirroring our findings in HEK-293T cells, in A549 and HeLa we found that the toxicity observed upon loss of THAP1 could be ameliorated upon exogenous expression of PSMB5 (Fig. 2C-D). THP-1 cells, however, did not exhibit reduced viability following THAP1 disruption (Fig. 2E). This prompted us to examine in more detail the nature of the cell lines in which THAP1 is not essential, which were strikingly enriched (P < 1x10^-9) for immune cells (‘myeloid’ or ‘lymphoid’ as defined by DepMap). Considering that the immunoproteasome is constitutively expressed by many immune cells^31,32, we reasoned that expression of PSMB8, the analogous counterpart of PSMB5 in the immunoproteasome, might relieve the essential requirement for THAP1. Indeed, there is a strong correlation between the essentiality of THAP1 and PSMB8 expression levels as measured by RNA-seq, wherein the cell lines in which THAP1 knockout has little or no impact on viability express the highest levels of PSMB8 (Fig. 2F). Indeed, we found that THP-1 cells expressed high levels of PSMB8 by qRT-PCR (Fig. 2G). We further validated these conclusions in HEK-293T cells, where, like PSMB5, exogenous expression of PSMB8 maintained the viability of THAP1 knockout cells (Fig. 2H and Fig. S2C). Thus, sustained expression of either PSMB5 or its immunoproteasome counterpart PSMB8 can rescue the toxicity associated with loss of THAP1.

A fluorescent reporter at the endogenous PSMB5 locus monitors THAP1 activity in live cells

We extended these findings by knocking-in a fluorescent reporter to the endogenous PSMB5 locus, enabling us to monitor PSMB5 expression in live cells. Following transfection of HEK-293T cells with Cas9, an sgRNA targeting the transcriptional start site of PSMB5 and a homology donor vector encoding the green fluorescent protein (GFP) variant mClover3³³ followed by a 2A peptide (Fig. 3A), we were readily able to establish a population of cells (~10%) which were stably GFP-positive (Fig. 3B). Single cell clones isolated from the GFP-positive population (Fig. 3C) harbored GFP at the intended site as validated by PCR from genomic DNA (Fig. S2D). Furthermore, lentiviral expression of shRNAs targeting PSMB5 resulted in a reduction in GFP expression (Fig. S2E), validating that the reporter clones could be used to quantitatively assess PSMB5 expression.

Open in new tabFigure 3: A fluorescent reporter measures endogenous PSMB5 expression in live cells.

(A-C) CRISPR/Cas9-mediated knock-in of GFP into the endogenous PSMB5 locus. A schematic representation of the procedure is shown in (A). Transfection of HEK-293T cells with Cas9, an sgRNA targeting PSMB5 and a donor template resulted in ~10% GFP-positive cells (B), which were purified by FACS and single cell cloned (C).

(D-G) THAP1 ablation reduces PSMB5 expression. CRISPR/Cas9-mediated disruption of THAP1 in a P_PSMB5-GFP reporter clone (expressing exogenous PSMB5 to ensure viability) reduced P_PSMB5-GFP expression (D), permitting the derivation of GFP^dim THAP1 KO clones (E). These THAP1 KO clones exhibited greatly reduced expression of PSMB5 by qRT-PCR (using primers annealing to the 3’UTR to ensure selective amplification of the endogenous transcripts) (F), and an absence of THAP1 protein by immunoblot (a non-specific band is indicated by an asterisk) (G). (***P < 0.001, t-test; ns, not significant)

We exploited our findings above to generate viable THAP1 knockout (KO) reporter cells. Following CRISPR/Cas9-mediated disruption of THAP1 in P_PSMB5-GFP reporter cells (sustained by exogenous expression of PSMB5) (Fig. 3D), we isolated single cell clones from the GFP^dim population by fluorescence-activated cell sorting (FACS) (Fig. 3E). Disruption of the THAP1 locus was confirmed by PCR from genomic DNA followed by Sanger sequencing (Fig. S2F). Using primers specific to the 3’ untranslated region of PSMB5 to allow for selective detection of the endogenous transcript, we confirmed a substantial reduction in PSMB5 expression in two THAP1 KO clones by qRT-PCR (Fig. 3F). However, our attempts to validate efficient knockout of THAP1 by immunoblot were hampered by the paucity of effective commercial antibodies. In particular, we found that the Proteintech antibody (12584-1-AP) used in multiple previous studies detected a prominent band running around the expected molecular weight (~25 kDa), but whose abundance was not affected upon CRISPR/Cas9 targeting of THAP1 (Fig. 3G). However, this antibody could readily detect exogenous THAP1 as a separate band that migrated just slightly slower, and, upon prolonged exposure, was able to detect endogenous THAP1 in control cells but not in the THAP1 knockout clones (Fig. 3G).

THAP1 acts through cognate binding sites located within the PSMB5 promoter

Next, we sought to determine how THAP1 might regulate PSMB5 expression. In support of a direct effect, chromatin immunoprecipitation followed by sequencing (ChIP-seq) data from the ENCODE project³⁴ revealed THAP1 occupancy immediately upstream of the PSMB5 transcription start site (TSS) (Fig. 4A). Despite THAP1 binding to thousands of gene promoters (Table S1), this property is not shared among the genes encoding proteasome subunits: PSMD8 is the only other gene to exhibit THAP1 occupancy (Fig. 4A). THAP1 contains a THAP-type zinc finger domain which mediates sequence-specific DNA binding, and competitive EMSA experiments³⁵ have defined the consensus binding sequence (“THABS” motif) as TNNNGGCA (where N represents any nucleotide) (Fig. 4B). Strikingly, examination of the PSMB5 proximal promoter region revealed two perfect matches within 200 bp of the TSS, and a third near-perfect match a further ~500 bp upstream (Fig. 4C). Together, these data suggest that THAP1 binds cognate motifs in the PSMB5 promoter to activate its transcription.

Open in new tabFigure 4: THAP1 binds cognate motifs within the PSMB5 promoter to regulate its expression.

(A) THAP1 binds the PSMB5 promoter. THAP1 ChIP-seq data in K562 cells³⁴ reveals an intense peak representing THAP1 occupancy at the PSMB5 transcription start site (TSS) (top). PSMD8 is the only other proteasome subunit at which concordant binding of THAP1 is observed (bottom).

(B) Consensus THAP1 binding site (THABS) motif (adapted from⁶⁶).

(C) Schematic representation of the three consensus THABS motifs located near the PSMB5

TSS.

(D-F) THAP1 targets cognate sites in the PSMB5 promoter to activate gene expression. (D) Schematic representation of a lentiviral reporter system in which ~1 kb of the PSMB5 promoter drives the expression of GFP. Removal of the three THABS motifs (“ΔTHABS”) reduced GFP expression (E); this effect was mediated through THAP1, as THAP1 ablation decreased GFP expression driven by the wild-type promoter but not the ΔTHABS promoter (F).

We examined this hypothesis by engineering a lentiviral reporter system in which ~1 kb of the PSMB5 proximal promoter was placed upstream of GFP (Fig. 4D). Single copy expression of this reporter construct in HEK-293T cells resulted in robust GFP expression (Fig. S3A). This appeared to be due in part to the activity of THAP1, as combined deletion of all three THABS motifs from the PSMB5 promoter (“ΔTHABS”) resulted in decreased GFP expression (Fig. 4E). Importantly, CRISPR-mediated ablation of THAP1 (performed following the introduction of exogenous PSMB5 to maintain cell viability) reduced GFP expression from the reporter construct driven by the wild-type PSMB5 promoter, but did not further reduce expression from the PSMB5 promoter lacking all THABS sites (Fig. 4F).

To investigate the relative importance of the three THABS motifs, we created an additional panel of mutant constructs in which each THABS motif was individually deleted. These data pointed to a critical role for site 2, as only the ΔTHABS2 construct exhibited decreased GFP expression relative to the level of the ΔTHABS construct (Fig. S3B). Moreover, the ΔTHABS2 construct was the only one unaffected following ablation of THAP1, whereas a marked reduction in GFP expression was observed in cells expressing ΔTHABS1 and ΔTHABS3 (Fig. S3C). Thus, THAP1 binding to a cognate motif (THABS2) immediately upstream of the PSMB5 TSS appears critical for PSMB5 expression.

Loss of THAP1 impairs proteasome function

As PSMB5 encodes one of the three catalytic subunits of the constitutive 20S proteasome core particle, we set out to test the hypothesis that the toxicity associated with loss of THAP1 was due to proteasome dysfunction. First, we examined whether the catalytic activity of PSMB5 was required to sustain cell viability upon THAP1 ablation. Whereas HEK-293T cells expressing wild-type PSMB5 did not exhibit any appreciable growth defect upon disruption of THAP1, cells expressing a catalytically-inactive PSMB5 mutant were not viable under these conditions (Fig. 5A). Second, as PSMB5 is critical to facilitate the integration of the other catalytic β subunits into the 20S core particle during proteasome assembly^36–38, we assessed whether loss of THAP1 resulted in defects in proteasome assembly. As a result of impaired autocatalytic cleavage of their N-terminal propeptide, an inability of the catalytic subunits to incorporate into the 20S core particle causes an accumulation of the immature proteins which can be detected by immunoblot³⁸. Supporting the idea of defective proteasome assembly in the absence of THAP1, we found decreased abundance of the mature forms of PSMB6 and PSMB7, concomitant with the accumulation of immature (uncleaved) species (indicated by asterisks) which were absent from control cells (Fig. 5B). Finally, we reasoned that proteasome dysfunction upon THAP1 loss should result in the stabilization of short-lived proteins. Exploiting the Global Protein Stability (GPS) two-color lentiviral reporter system³⁹ (Fig. 5C), we found that CRISPR-mediated ablation of THAP1 stabilized two exogenous GFP-degron fusion proteins to a similar degree as shRNA-mediated knockdown of PSMB5 (Fig. 5D). Similarly, THAP1 disruption also resulted in increased abundance of endogenous HIF-1α, which is constitutively degraded by the proteasome in normoxia^25,40 (Fig. 5E). Altogether, these data support a model whereby the death of cells lacking THAP1 is caused by defective proteasome function resulting from inadequate expression of PSMB5.

Open in new tabFigure 5: PSMB5 insufficiency resulting from THAP1 loss impairs proteasome function.

(A) The catalytic activity of PSMB5 is required to rescue viability upon THAP1 loss. Exogenous expression of wild-type PSMB5, but not a catalytically-inactive mutant, restored the viability of HEK-293T cells following THAP1 ablation.

(B) Loss of THAP1 impairs proteasome assembly. THAP1 ablation decreases the abundance of mature, processed PSMB6 and PSMB7 as assessed by immunoblot, but leads to the accumulation of the uncleaved proproteins (indicated with asterisks).

(C-E) THAP1 impairs the proteasomal degradation of short-lived proteins. (C) Schematic representation of the lentiviral Global Protein Stability (GPS) two-color fluorescent reporter system to monitor protein stability. (D) Stabilization of two model GFP-degron fusion proteins upon ablation of THAP1, as assessed by flow cytometry; the N-terminal peptide derived from PTGS1 harbors an N-terminal degron targeted by UBR-family E3 ligases⁶⁷, while the C-terminal peptide derived from TNNC2 harbors a C-terminal degron targeted by Cul4^{DCAF12 68}.

(E) Increased abundance of endogenous HIF-1α upon THAP1 disruption, as assayed by immunoblot.

Defining the transcriptional targets of THAP1

The key transcriptional targets of THAP1 remain poorly defined, hampering our ability to understand the functional consequences of THAP1 mutations in disease. Armed with the knowledge that exogenous PSMB5 expression ameliorates the toxicity associated with THAP1 loss, we reasoned that we could leverage our findings to assess the impact of THAP1 deletion on the transcriptome. To avoid potential artefacts resulting from the analysis of single cell clones, we purified a population of THAP1 knockout cells. Following the CRISPR-mediated ablation of THAP1 in PSMB5-GFP knock-in reporter cells (overexpressing PSMB5 to ensure viability), we isolated GFP^dim cells by FACS and performed RNA-seq analysis (Fig. 6A). After discounting genes exhibiting differential expression between untransduced cells and cells expressing control sgRNAs, we identified 277 genes (220 downregulated, 57 upregulated) whose expression was significantly altered (FDR < 0.001, fold-change > 2) upon THAP1 knockout (Fig. 6B and Table S2). Supporting the veracity of the dataset, the most significantly downregulated gene was Shieldin complex subunit 1 (SHLD1, previously known as C20orf196), consistent with recent findings describing a role for THAP1 in DNA double strand break repair choice¹⁶. Furthermore, although the requirement for exogenous expression of PSMB5 precluded its identification as a differentially expressed gene, the abundance of intronic reads mapping to the endogenous PSMB5 locus was greatly reduced in the THAP1 knockout cells (Fig. S4A-B).

Open in new tabFigure 6: Defining the transcriptional targets of THAP1.

(A) Schematic representation of the RNA-seq experiment.

(B-C) Identifying the transcriptional targets of THAP1. A summary of the RNA-seq dataset is shown in (B): genes exhibiting differential expression between sgControl and sgTHAP1 cells (n = 277) are highlighted in yellow, with the subset of those genes that display THAP1 occupancy as assessed by ChIP-seq (n = 42) colored red. Expression changes amongst all differentially expressed genes are summarized in (C), with circles representing genes bound by THAP1.

(D) Consensus binding sites for YY1 (purple) and THAP1 (red) are enriched amongst the promoters of the 42 direct THAP1 target genes. Bubble size is proportional to the number of the gene promoters containing the binding site. See also Fig. S4C.

(E-I) Validation of THAP1 target genes. Five target genes directly activated by THAP1 binding are shown: ChIP-seq data indicating THAP1 occupancy is shown on the left; the RNA-seq expression data is summarized in the center, and qRT-PCR validation is shown on the right. (***P < 0.001, t-test)

To identify direct transcriptional targets of THAP1, we cross-referenced the differentially expressed genes identified through RNA-seq with THAP1 binding sites as defined by ChIP-seq. Among the differentially expressed genes, 42 exhibited THAP1 occupancy in their proximal promoter (Fig. 6C); of these 42 direct targets, 19 were downregulated upon loss of THAP1 while 23 were upregulated, suggesting that THAP1 has the potential to act as either a repressor or activator of transcription depending on the genomic context. However, the primary role of THAP1 appeared to be as an activator, with several of its direct targets exhibiting marked downregulation in its absence (Fig. 6B-C). We found no significant functional enrichment among the differentially expressed genes through GO term analysis, but their promoter sequences were enriched for transcription factor binding motifs for both THAP1 and YY1, a known THAP1 co-factor^16,41,42 (Fig. 6D and Fig. S4C). None of these genes are currently associated with dystonia, but they include ECH1, an enzyme involved in fatty acid metabolism that has been previously identified as a THAP1 target⁴², and METTL3, the N⁶-methyladenosine methyltransferase, which is an attractive therapeutic target in cancer⁴³. Across five of the THAP1 target genes that exhibited the greatest degree of downregulation upon loss of THAP1, we further validated these findings by qRT-PCR (Fig. 6E-I). Altogether, these data define the genes directly targeted by the transcription factor activity of THAP1.

A deep mutagenic scan defines the landscape of THAP1 mutations in Dystonia

A wide range of autosomal dominant mutations distributed throughout the THAP1 coding sequence give rise to DYT-THAP1 dystonia^44–50, an early-onset neurological disorder characterized by involuntary muscle contractions and movements causing abnormal and painful posturing. Thus, we sought to exploit our P_PSMB5-GFP knock-in reporter clone to assess the functional impact of DYT-THAP1 mutations. Genetic complementation of THAP1 KO cells with wild-type THAP1 did result in a restoration in P_PSMB5-GFP expression, although this effect was partial and did not restore GFP fluorescence to the levels observed in the parental cells (Fig. S5A). However, this assay was sufficiently sensitive to report on THAP1 activity, as expression of an inactive THAP1 mutant unable to bind DNA (C5A, which abrogates zinc chelation by the zinc finger motif¹⁵) did not restore P_PSMB5-GFP expression (Fig. S5A).

With the goal of globally defining how mutations in THAP1 affect its function, we leveraged our phenotypic assay in P_PSMB5-GFP knock-in reporter cells to carry out a deep mutational scan. Through microarray oligonucleotide synthesis, we generated a library of mutant constructs in which each residue of THAP1 (with the exception of the initiator methionine, 212 amino acids total) was systematically replaced with all of the other 20 possible amino acids (Fig. 7A-B). The resulting site-saturation mutagenesis library was packaged into lentiviral particles and introduced into THAP1 KO P_PSMB5-GFP reporter cells at low multiplicity of infection, ensuring single-copy expression. We then used FACS to partition the population into GFP^dim cells, in which no restoration of P_PSMB5-GFP expression was observed, and GFP^bright cells, in which P_PSMB5-GFP expression was restored, and quantified the THAP1 variants present in each population by Illumina sequencing (Fig. 7B).

Open in new tabFigure 7: A deep mutational scan defines the functional landscape of THAP1 mutations.

(A) Schematic representation of the domain architecture of the THAP1 protein.

(B) Schematic representation of the deep mutational scan, designed to interrogate the ability of all possible single amino acid variants of THAP1 to activate expression of the endogenous P_PSMB5-GFP knock-in allele.

(C) Site-saturation mutagenesis reveals critical residues for THAP1 function. Each cell represents the performance of a single THAP1 mutant: the mean performance of all the wild-type proteins is centered at 1 (light gray), with red colors indicating mutants which abrogate activation of the P_PSMB5-GFP reporter and blue colors indicating mutants which may enhance the activation of the P_PSMB5-GFP reporter. Dark gray cells indicate mutants for which insufficient data was available for analysis.

After an initial filtering step to remove variants with low read counts, we recovered data for 4002 of the 4240 possible single amino acid variants (94.3%). Overall, we observed high concordance between mutant performance across two replicate experiments (Fig. S5B); however, we discarded 179 mutants which exhibited discordant behavior between the two replicate experiments, leaving a total of 3823 (90.2%) variants for analysis (Table S3). The results are summarized as a heatmap in Fig. 7C, with the data normalized such that the mean performance of all the control (wild-type) constructs is centered at 1; thus, the darker the red color the more deleterious the impact of the mutation on THAP1 function, whereas blue cells indicate mutations which may enhance the THAP1-mediated activation of P_PSMB5-GFP expression.

We evaluated the quality of the dataset in several ways. First, we considered residues essential for zinc chelation and hence folding of the zinc finger motif¹⁵: C5, C10, C54 and H57. These critical residues were uniformly essential for THAP1 activity, as mutation to any other residue prevented activation of the P_PSMB5-GFP reporter (Fig. 8A). Moreover, mutations across all residues previously determined to be important for DNA binding through biochemical assays¹⁵ were extremely deleterious (Fig. 8B). Second, the global landscape of THAP1 activity correlated well with the predicted structure of THAP1 (Fig. 8C-D). In particular, mutation of most residues in the two structured regions, the THAP-type zinc finger (residues 2-81) and the predicted coiled-coil domain (residues 139-191) abrogated transcriptional activity, whereas most mutations targeting the unstructured central linker (residues 82-138) and C-terminus (residues 192-213) did not impair transcriptional activity (Fig. 7C and Fig. 8E). A notable exception, however, was the DHNY motif (residues 134-137) lying at the end of the central unstructured linker, which was absolutely critical for THAP1 function (Fig. 7C). Interestingly, this motif has been identified as the binding site for HCFC1⁵¹, an essential cofactor for the THAP1-mediated activation of SHLD¹⁶. Thus, the interaction with HCFC1 is also likely to be critical for the THAP1-mediated activation of PSMB5. Indeed, we confirmed that deletion of the DHNY motif and the coiled-coil domain, but not the disordered C-terminus, abolished the ability of THAP1 to activate the P_PSMB5-GFP reporter (Fig. 8F).

Open in new tabFigure 8: Determining the functional effects of THAP1 mutations found in Dystonia patients.

(A-B) The deep mutational scan successfully identifies THAP1 residues critical for DNA binding. (A) Performance of all THAP1 variants targeting zinc-chelating residues of the zinc finger domain. Bars indicate the mean of two replicate experiments. Constructs which encode the wild-type protein are indicated in bold; the mean activity exhibited across all the wild-type THAP1 constructs is set at 1 (dotted line). (B) Distribution of activity scores across all the THAP1 variants targeting residues previously determined¹⁵ to be important for DNA binding by THAP1.

(C-F) Defining structure-function relationships for THAP1. Mean activity scores from the genetic screen were mapped onto the predicted structure of THAP1 (C) or the experimental structure of the THAP1 zinc finger domain¹⁵ (D). Overall, residues predicted to lie in ordered regions of the protein (AlphaFold pLDDT > 60) were much less tolerant of mutations than residues predicted to lie in disordered regions (E). Individual validation of the screen results was performed using flow cytometry: deletion of the coiled-coil and HCFC1-binding motif abrogated THAP1 function, whereas deletion of the disordered C-terminus did not (F).

(G-H) Profiling the activity of THAP1 mutations found in Dystonia patients. (G) Performance of all missense variants identified in Dystonia patients, displayed as in (A). (H) Individual validation experiments measuring the activity of THAP1 mutants predicted to be inactive (top row) and THAP1 mutants predicted to be active (bottom row) by flow cytometry. See also Fig. S5C-E.

Many of the THAP1 mutations identified in dystonia patients remain of uncertain significance⁵⁰. Thus, we examined the utility of this dataset in classifying the functional effects of THAP1 variants identified clinically (Fig. 8G). The majority of these mutations strongly impaired the ability of THAP1 to activate expression of the P_PSMB5-GFP reporter, consistent with the prevailing view that disease-causing mutations represent loss-of-function alleles^19,41,50. However, some mutants exhibited activity at or approaching the level of the wild-type protein, suggesting that they might represent benign variants. To verify that the screen results could be faithfully recapitulated in individual experiments, we selected eight patient mutations predicted to abolish THAP1 activity (A7D, R13H, K24E, P26R, H57N, L72R, F81L and N136S) and compared their performance to five mutants predicted not to affect THAP1 activity (I80V, C83R, M143V, A166T and D192N). Validating the screen results, the eight inactive mutants exhibited little or no ability to activate P_PSMB5-GFP reporter expression (Fig. 8H and Fig. S5C-E), whereas the five active mutants exhibited similar performance to wild-type THAP1 (Fig. 8H). Altogether, these data illustrate structure-function relationships for THAP1 at high resolution, enabling the functional classification of clinical THAP1 mutations.