Taf1 knockout is lethal in embryonic male mice and heterozygous females show weight and movement disorders

. Here, we generated and validated the first conditional cre-lox allele and ubiquitous


INTRODUCTION
The TATA-box binding protein-associated factor 1 (TAF1) is a ubiquitously expressed protein that plays a key role in initiation of RNA polymerase II-dependent transcription.TAF1 is best known for forming a protein complex with other TAFs, to form the basal transcription factor TFIID, and to regulate the transcription initiation complex assembly activation machinery (Bhuiyan, 2019;Malik, 2023).The TAF1 gene is located on the X Chromosome (Xq13.1);thus, in males there is only one copy, and in females there are two, one of which undergoes X-inactivation to maintain the same dosage in both sexes (Deng, 2014).
TAF1 missense variants in males (maternally inherited or de novo) can cause the neurodevelopmental disorder X-linked syndromic mental retardation-33 (MRXS33) also known as X-linked intellectual disability (XLID), which may present with craniofacial abnormalities and a variety of other clinical features including congenital heart disease (Gudmundsson, 2019;O'Rawe, 2015).TAF1 missense variants in females have been reported with similar phenotypes to males (Giovenino, 2023;Gudmundsson, 2019).
However, many female cases are asymptomatic and were identified due to their relationship with affected male family members (Gudmundsson, 2019).In both males and females, almost all TAF1 variants are missense.Only two unrelated families carry gene duplications including TAF1 (and surrounding genes at Xq13.1), resulting in a severe progressive neurodegenerative phenotype but few common clinical features to those with missense variants (O'Rawe, 2015).Finally, just one splice-site variant has been identified in an asymptomatic mother and her affected son (O'Rawe, 2015).Therefore, the genotype-Disease Models & Mechanisms • DMM • Accepted manuscript phenotype correlation between TAF1 mutations and clinical outcomes is not straightforward, and there is very little data regarding the relationship between mutation and mRNA/protein expression of TAF1.Currently, there are no successful treatments for these TAF1 disorders.
In most male patients, XDP is characterised initially by dystonia (93.4% of cases) that generalises within 5-10 years from the age of onset (~39.5 years in males), which eventually transitions into Parkinsonism (Lee, 2001;Lee, 2011).Typically affected males die around 16 years after diagnosis (Rosales, 2010).A few female XDP cases have been reported at a ratio of 1:100 (females:males), with a mean age of onset of 46 years old; these individuals present with variable dystonia-Parkinsonism features (Lee, 2011).However, as with TAF1 mutations, many mild or asymptomatic female cases may go undetected without genetic testing outside of families with affected members (Evidente, 2002;Evidente, 2004) and thus the true proportion of affected versus healthy females with XDP is difficult to estimate.
Currently, there are no interventions that have achieved success in the treatment of XDP (Bragg, 2019).
In both females carrying TAF1 mutations (Cheng, 2019;Hurst, 2018;O'Rawe, 2015) and those with XDP (Domingo, 2014), skewed X-inactivation (a process whereby one X chromosome inactivation is selected over the other) towards the non-mutant chromosome has been suggested to explain the lack of, or minimal, phenotypes in some individuals (Evidente, 2004).However, in some females, carrying a TAF1 mutation (Cheng, 2019;Giovenino, 2023) or the XDP allele (Domingo, 2014) results in disease features despite reports of skewed X-inactivation; this indicates that the phenotype is not masked and it has even been suggested that the mutant allele is more highly expressed in such cases (Domingo, 2014).

Disease Models & Mechanisms • DMM • Accepted manuscript
Model organisms are essential to define the role of TAF1 in disease, and various knockout and knockdown mutations have been published.In yeast (Saccharomyces cerevisiae) temperature-sensitive deletion of different TAF1 domains indicated that ~90% of the yeast genome requires TAF1 for transcription (Irvin, 2006), others showed nearly all mRNA is strongly dependent on TFIID function (Warfield, 2017).In worms (Caenorhabditis elegans), RNAi-mediated knockdown showed that TAF1 in the TFIID complex is important for embryonic transcription (Walker, 2004).Very few vertebrate models of TAF1 dysfunction exist.In zebrafish (Danio rerio), homozygous knockout (KO) of taf1 was embryonically lethal showing neurodevelopmental defects (Gudmundsson, 2019).To deplete taf1 gene expression in zebrafish embryos, a transient in vivo splice-blocking morpholino (MO) approach was taken (O'Rawe, 2015); both approaches reduced the area of the optic tectum (midbrain) indicating microencephaly, comparable to human XLID patients (Gudmundsson, 2019;O'Rawe, 2015).Two rat models with postnatal knockdown of Taf1 have been published.In the first model, intracerebral injection into post-natal day 3 rats of lentivirus vectors carrying CRISPR/Cas9 reagents was used to target exon 1, resulting in behavioural and neurodevelopmental defects (Janakiraman, 2019).In the second model, intrastriatal infusions into 3-week-old rats of adeno-associated viral vectors carrying miRNAs was used to target Taf1 in a splice variant-selective manner (Cirnaru, 2021), resulting in similar motor deficits.These same targeting constructs were used with intracerebroventricular injection into the brains of newborn mice, and phenotypes were similar to, or more even pronounced than, the in the rat models (Cirnaru, 2021).
While these various models have provided important insights into TAF1 function, the field has suffered from the lack of a genetic mouse model of TAF1 disease to explore mechanism and treatments.Despite the recognition of limitations in all models, mice are well recognized as the premier genetic model (Eppig, 2017) and conditional cre-lox alleles are currently the most powerful (Kaloff, 2017).Such a mouse model will allow us to assess the role of TAF1 from the very earliest developmental stages and throughout life.It will allow molecular investigation of TAF1 transcription and protein expression coupled with gross brain anatomy and behaviour studies to reveal disease mechanisms.Finally, a well validated mouse model will be a critical platform for therapy development to address these serious developmental and neurodegenerative diseases.

Generation and sequence verification of Taf1 cKO allele
Fig. 1 A shows a schematic of the Taf1 conditional knock-out (cKO) allele, which follows the design recommendations of the International Knockout Mouse Consortium (Wefers, 2011).
The locus of X-over P1 (loxP) sites flank exon 8, which is present in all known Taf1 transcripts, and exons 8 and 8 in transcript 201 (Ensembl transcripts: ENSMUST00000101341.9,ENSMUST00000118878.8,ENSMUST00000149274.9).The 1,524bp deletion is designed to create a frameshift that will result in downstream stop codons.
The loxP sites were added in two steps.First, clustered regularly interspaced short palindromic repeats (CRISPR) strategies designed to simultaneously add both loxP sites, were tested in embryonic stem cells (ESCs).The optimal strategy (Table S1) was microinjected into C57BL/6J (B6) zygotes ex vivo.The resulting mice were genotyped with primers shown in Fig. 1B, and Table S2.This resulted in one male founder, which passed the new allele to offspring, but with only the 5 loxP site present (Fig. 1C).This 5-loxP-carrying strain was then expanded and used for a second round of CRISPR to introduce the 3 loxP site.For this second insertion, another CRISPR strategy was developed and optimized in ESCs.The optimal strategy (Table S1) was delivered using improved Genome editing via oviductal nucleic acids delivery (i-GONAD) electroporation into B6-Taf1 5-loxP late zygotes in vivo.This resulted in three founders: one female and two males.The female and one male each passed the new allele to offspring (Fig. 1D), which resulted in strains named respectively C57BL/6J-Taf1 em1Ems , and C57BL/6J-Taf1 em2Ems (for brevity called here Taf1 em1Ems , and Taf1 em2Ems ).Both strains were initially studied to ensure reproducibility.
A region of ~5.6 kb encompassing ~0.8 kb upstream of the 5 loxP site to ~1.5 bp downstream of the 3 loxP site was Sanger sequenced from a hemizygous male ear notch DNA sample from strains Taf1 em1Ems , and Taf1 em2Ems (Fig. 1B).This revealed in Taf1 em1Ems , two single base pair deletions within a 3 bp intronic region, whereas Taf1 em2Ems showed complete integrity of the expected Taf1 sequence plus the two loxP sites.Although we did not observe any phenotypic impact of this unplanned mutation, going forward, all studies focused on Taf1 em2Ems .DNA prepared from the brain of another Taf1 em2Ems 2loxP hemizygous male was sent for Samplix indirect sequence capture analysis (Table S3) (Blondal, 2021)  Ubiquitous KO of Taf1 resulted in 1loxP heterozygous females, but no 1loxP hemizygous males For brevity, the unrecombined conditional Taf1 allele will be referred to as 2loxP, since two loxP sites are present in the genome, and the recombined Taf1 allele will be referred to as 1loxP, since after deletion there is only 1loxP site remaining.
The introduction of loxP sites alone, without Cre deletion, can change the transcription of a gene causing an unplanned phenotype.Thus, we undertook characterization of unrecombined 2loxP Taf1 em1Ems and Taf1 em2Ems strains.The 2loxP heterozygous females and 2loxP hemizygous males bred with the typical fertility and fecundity of B6 wild-type (Wt) mice.Both maternal and paternal transmission to offspring was as expected.At least three mice of each genotype were aged to at least 1 year and no abnormal phenotype was observed.For the Taf1 em2Ems strain, 2loxP homozygous females were generated and bred to 2loxP hemizygous males.The 2loxP homozygous females also had typical fertility and fecundity and transmission to offspring was as expected.During all these breeding experiments, no cage-level phenotypes were observed in the 2loxP mice.
Thus, we conclude that the introduction of the two loxP sites into the introns of Taf1 was without detrimental consequence.
Although the strains were generated on the B6 background, we also backcrossed the mice to B6 to remove any unlinked mutations that might have been caused by the CRISPRbased methodology (Fig. 2A).For this we chose to breed the 2loxP allele through the female, since with an X-linked gene this strategy produced both 2loxP heterozygous female and 2loxP hemizygous male mice.Backcrossing continued to N5 with no impact on expected inheritance, nor fecundity or fertility.No other cage-level phenotypes were observed.Mice for molecular and histological characterization were from the N4 and N5 backcrosses.
To mimic the human disease situation, in which a pathogenic variant is typically inherited or occurs very early in development, we bred the Taf1 em2Ems 2loxP strain to the ubiquitously-expressing B6.FVB-Tg(EIIa-cre)C5379Lmgd/J (for brevity called here Ella-cre) (Lakso, 1996).EIIa-cre mice carry the cre transgene under the control of the adenovirus EIIa promoter that targets expression to the early mouse embryo, and is useful for wide-spread Disease Models & Mechanisms • DMM • Accepted manuscript recombination of loxP sites across all tissues, including in the germ line.We did the cross both ways, bringing the Ella-cre transgene in from the maternal (Fig. 2B) or paternal (Fig. 2D) parent, since others have shown that the parental source of cre can affect recombination efficiency (Luo, 2020).We found in both cases that the loxP sites were able to recombine, deleting 1,524 bp including exon 8 (or 8 and 8) of the Taf1 gene, but maternal transmission was more effective than paternal.Maternal (Fig. 2B) transmission of cre generated both partially recombined mosaic females (1&2loxP/+), and completely recombined 1loxP heterozygous females.However, because of the location of Taf1 on the X Chromosome, 1loxP hemizygous males were not produced.Fig. 2C shows representative examples of genomic-DNA amplification results for all mice, including partially recombined and completely recombined 1loxP heterozygous females.Paternal (Fig. 2D) transmission of cre should have generated both recombined 1loxP heterozygous female offspring, and 1loxP hemizygous male offspring, but neither were recovered, despite the presence of partially recombined females (1&2loxP/+).Thus, we bred the first-generation 1loxP heterozygous females from Fig. 2B to Wt males (Fig. 1E).Importantly, none passed a 2loxP chromosome to offspring, which could have happened if the females were mosaics.Despite scoring pups on the day of birth, no 1loxP heterozygous females nor 1loxP hemizygous males were recovered, which was highly significantly different from expected.The lack of 1loxP heterozygous females may be due to the mouse-specific phenomenon of paternal Xinactivation in extraembryonic tissues (in this cross the Wt X Chromosome was paternal), instead of the human random pattern (Huynh, 2005).Thus, we considered statistically the lack of 1loxP hemizygous males alone, which was also highly significantly different from expected (loxP:Wt, 0:40, Binomial Test, 1 tailed, p<0.0001).From this we concluded that in mice, dependence on the KO Taf1 allele, which is designed to be a protein null, is incompatible with postnatal life.

Taf1 1loxP hemizygous male embryos survived to blastocyst stage and implanted
To discover when the 1loxP heterozygous embryos died, and to eliminate the theoretic possibility that the 1loxP chromosome was not entering the germline of the first-generation 1loxP heterozygous females, we undertook timed pregnancy experiments.
First, we harvested embryonic day (E) 9.5 embryos from B6-1loxP dams crossed to B6 sires (Fig. 3A), and our initial observation was that a portion of the decidua did not Disease Models & Mechanisms • DMM • Accepted manuscript contain any embryonic tissue (embryo or yolk sac) (Fig. 3A,B).This "empty cup" phenotype suggested implantation occurred but the embryo failed to develop.Overall, we observed 15 empty decidua, and 15 decidua from which we could recover E9.5 embryos within yolk sacs.
The yolk sacs were genotyped for sex and Taf1 exon 8 status, and the results were 7 Wt female embryos, and 8 Wt male embryos.Neither 1loxP heterozygote females nor 1loxP hemizygote males were recovered at this stage of development.
Following this observation, E3.5 blastocysts were harvested, both from B6 mice as above, and hybrid B6129F1 mice using B6 dams crossed to 129 sires (Fig. 3C,D).Overall, we recovered a total of 55 blastocysts.Surprisingly, all four genotypes were recovered: 18 1loxP heterozygous female blastocysts, 15 Wt female blastocysts, 9 1loxP hemizygous male blastocysts, and 13 Wt male blastocysts.A similar breakdown of genotypes was observed on both backgrounds, demonstrating that recombinant 1loxP mice are able to survive to the blastocysts stage, and, as suggested by the E9.5 data above, are able to reach the implantation stage and induce uterine decidualization.
We then went on to produce 1loxP and 2loxP embryonic stem cells (ESCs) from B6129F1 E3.5 blastocysts (Fig. 3E,F).From the experimental 1loxP cross, we recovered a total of 19 ESC lines: 3 1loxP heterozygous female lines, 11 Wt female lines, zero 1loxP hemizygous male lines, and 5 Wt male lines (Fig. 3E).From our observations, it appeared that 1loxP hemizygous male embryos in culture were not able to hatch and form viable cell lines.For the control 2loxP cross, we recovered a total of 14 ESC lines: 5 2loxP heterozygous female lines, 3 Wt female lines, 4 2loxP hemizygous male lines, and 2 Wt male lines (Fig. 3F).

Overall expression of Taf1 mRNA brain transcripts is not changed between genetic groups
To determine whether insertion of the loxP sites in 2loxP heterozygous, 2loxP homozygous, or 1loxP heterozygous mice affected gene transcription, quantitative reverse transcription PCR (RT-qPCR) was performed.The mouse Taf1 gene encodes at least three known proteincoding transcripts including: Taf1-201, Taf1-202 (canonical transcript), and Taf1-206 (Ensembl: ENSMUSG00000031314.19)(Fig. 1A).Cre-loxP recombination targets exon 8 and downstream microexon 8' leading to induction of a stop codon in exon 9, for termination of further mRNA translation.Primers were designed against specific exons for detection of each of the Taf1 transcripts: exon 7 (upstream of the deleted region), exon 8 (deleted region), exon 34, exon 34' (targets the neuronal specific nTaf1 isoform; Taf1-206 variant),

Disease Models & Mechanisms • DMM • Accepted manuscript
and exon 38 (targets Taf1-201 variant).In 1loxP heterozygous female mice, no transcripts from the deleted allele were detected in any PCR products spanning the deletion.
RT-qPCR analysis of female mouse brains revealed that there were no differences in the expression level of each of these transcripts detected at the respective exons with either 2loxP or 1loxP sites compared to Wt control mice (Fig. 4A).Similarly, 2loxP hemizygous male mice showed no differences in Taf1 gene expression compared to Wt mice (Fig. 4B).Two to three RT-qPCR plates were run for each set of primers (n=2-3/genotype per plate) during the optimization stage, resulting in reproducible Ct values with no differences between plates, which were normalised to results from Wt mice then combined for analysis (total n=6/genotype).These findings suggest that there was no effect of insertion of the 2loxP sites into the mouse genome on the overall levels of Taf1 mRNA expression, and that 1loxP heterozygous female mice showed normal levels of Taf1 mRNA expression.

No variation in TAF1 protein brain expression in 1loxP heterozygous female mice
The relative level of TAF1 protein expression was determined by western blot analysis of total brain lysates isolated from Wt, 2loxP heterozygous, 2loxP homozygous, and 1loxP heterozygous female mice, and from Wt and 2loxP hemizygous male mice.No variation in TAF1 protein expression was observed between these genotypes, as detected with a TAF1specific monoclonal antibody (Capponi, 2020) (Fig. 5; Table S5).Reproducible results were obtained through multiple blots, where a representative blot is shown in Fig. 5A.
Quantification shown in Fig. 5B is from two separate blots normalised to the Wt values then combined, whereas samples were run for male mice on the same gel (Fig. 5C).This suggests that there is no impact of the loxP sites on TAF1 expression and that 1loxP heterozygous female mice express TAF1 at normal levels.

Taf1 deletion in heterozygous female mice does not affect nuclear TAF1 brain expression
Since western blotting revealed no significant differences in TAF1 expression across genetic groups, immunofluorescent staining was performed to visualise TAF1 expression in female mouse brain.Initially, antibody optimization and positive and negative controls were undertaken (Fig. S1, Fig. S2).Then, TAF1 expression was analysed across coronal sections showing the striatum and cortex (Fig. 6).A DARPP32 antibody was used to mark the striatum, and TAF1 was ubiquitously expressed in both the cortex and striatum as well as

Disease Models & Mechanisms • DMM • Accepted manuscript
other regions including the corpus callosum.Multiple coronal sections were taken from each mouse and each brain section gave reproducible results.Overall, Taf1 1loxP heterozygous female mice showed similar ubiquitous expression of TAF1 as in Wt mice.Of note, the total area of the brain and striatal and cortical areas were quantified and showed no difference between genotypes (Fig. S3A-C).
TAF1 is a transcription factor with its main function and location in the nucleus.A CTIP2 antibody was used to co-stain nuclei of medium spiny neurons (MSNs) in the striatum (Arlotta, 2008).All brains studied showed TAF1 staining that was specific to nuclei and all visible nuclei were stained with TAF1 (Fig. 7).TAF1 was highly expressed in MSNs as well as other cells in the striatum in all groups (Fig. S4A).The 1loxP heterozygous female mice did not show differences in TAF1 staining when compared to the other genetic groups.This experiment was highly reproducible and similar results were obtained with three separate sets of mice from each genotype.Since all nuclei were stained with TAF1 in heterozygous mice (Fig. S4B), this indicated that TAF1 expression was active in each cell.Of note, the number of TAF1-and CTIP2-stained nuclei, expressed as a percentage of the total nuclei, were no different between genetic groups (Fig. S4A-B).

Abnormal weight and movement in 1loxP heterozygous female mice
Behaviour analysis was done with two separate groups, at two separate times.Each group consisted of 32 female mice, the first group with 16 2loxP heterozygous controls and 16 1loxP heterozygous experimental was used for open field and rotarod, and the second group with 16 2loxP homozygous controls and 16 1loxP heterozygous experimental was used for open field only.One investigator did all the experiments.For open field, there was no significant impact of arena, nor trial.For rotarod, there was no significant impact of trial, or lane.Furthermore, all genotypes were age matched, and the 2loxP heterozygous and 2loxP homozygous controls did not differ in any of the measurements (age, weight, crosses into centre, duration in centre, or distance travelled).Therefore, data was pooled between the groups as presented in Fig. 8.
Mice from each genotype were age-matched adults (Fig. 8A), therefore it was unexpected to observe that 1loxP experimental mice were significantly heavier than the 2loxP controls (Fig. 8B, pooled, p = 0.0003; significant difference also observed in both separate groups).

Disease Models & Mechanisms • DMM • Accepted manuscript
At this age (range 3-8 months, average 6.6), which is beyond the period of development and prior to obesity associated with extreme age, we did not expect a significant change of weight with age, as observed in the 2loxP controls (Fig. 8C, pooled, p = 0.1800, r = -0.2431;no significance observed in separate groups).Unexpectedly, we observed a significant increase in weight with age with the 1loxP experimental mice (Fig. 8D, pooled, p = 0.0001, r = 0.6267; significant difference also observed in both separate groups).
Thus, both increased weight, and increased weight with age, were significant phenotypes of the 1loxP experimental mice.
In open field testing, we observed that 1loxP mice significantly crossed into the centre zone less (Fig. 8E, pooled, p = 0.0022; significant difference also observed in both separate groups), and travelled less (Fig. 8G, pooled, p = 0.0002; significant difference in one group, trend in second group).These are two highly-related data sets that both indicate less movement in the 1loxP mice.Duration in centre zone was not significantly different between groups (Fig. 8F, pooled, p = 0.5243; no significance observed in separate groups).
In rotarod testing (tested in group 1 only), no significant difference was observed between the control and experimental mice in relation to the average time to fall (Fig. 8H, p = 0.2835).
To examine the hypothesis that decreased movement in the 1loxP experimental mice was due to their increased weight, the relationship between the two measurements was examined.No correlation was found between distance traveled and weight in the 2loxP control mice (Fig. 8I, pooled, p = 0.9576, r = -0.009785;no significance observed in separate groups) or in the 1loxP experimental mice (Fig. 8J, pooled, p = 0.6456, r = -0.08451,no significance observed in separate groups).Thus, the hypothesis that increased weight made the 1loxP mice move less was not supported.

DISCUSSION
Here, we generated and validated the first knock-out mutant mouse strain for the Taf1 Xchromosome gene, which encodes the largest subunit of the TFIID basal transcription factor causative of human disease (Gudmundsson, 2019;O'Rawe, 2015).Given the ubiquitous and critical role of this gene (The Human Protein Atlas version 23.0, https://www.proteinatlas.org/ENSG00000147133-TAF1,accessed January 2024), lethality Disease Models & Mechanisms • DMM • Accepted manuscript was a concern; thus, we made a cre-lox conditional allele (Gu, 1993).Since this allele was created with CRISPR, an approach efficient in making the desired genomic change, but well known to also result in unwanted genome events (Tan, 2015), we extensively sequenced (44.8 kb) the final allele and found it free of spurious mutations.To mimic the human disease situation, and validate this new conditional allele, we bred the mice to a ubiquitous cre-driver allele that is expressed early in development, resulting in germline recombination of the loxP sites.
We discovered that in this mouse model, Taf1 deletion in males was embryonically lethal.Importantly, this result may provide the explanation as to why no null-variants have been reported in humans.Interestingly, without Taf1, 1loxP male embryos did not die immediately, as might have been expected of such a critical transcription factor.Instead, embryos were able to develop to the blastocyst stage (E3.5, ~140 cells), and induce uterine decidualization.Since this time frame is beyond the maternal effect (Li, 2010), it suggests that the role of TAF1 in mammals may be specialized to a subset of transcripts at this stage of embryonic development.These findings are in line with results from studies on knockout of other TAFs in mice, such as Taf7 (Gegonne, 2012), Taf8 (Voss, 2000), and Taf10 (Mohan, 2003), which also resulted in termination of development at early embryonic stages.
The blastocyst stage is when mouse ESCs can typically be derived.However, we were unable to establish 1loxP male ESCs, although we could establish ESCs from all other genotypes: Wt female and male, 2lox heterozygous female and hemizygous male, and 1lox heterozygous female.We anticipate that the 2loxP hemizygous male cell lines, when transfected with cre to induce loxP recombination and Taf1 deletion, will be an important new mammalian tool for advancing our knowledge regarding the cellular role of Taf1.
Human heterozygous female carriers of TAF1 variants demonstrate a spectrum of phenotypes from asymptomatic to dystonia-Parkinsonism features (Giovenino, 2023;Gudmundsson, 2019;Lee, 2011;Vianna, 2020).Given that the 1loxP heterozygous female mice carried an even more severe mutation (a null) than that observed in humans, and that in males it was lethal, we anticipated that the females would be strongly affected.Since this is a neurological disease, we studied the brains of Taf1 1loxP heterozygous females for gross Disease Models & Mechanisms • DMM • Accepted manuscript structural abnormalities, overall expression by quantification of TAF1 RNA and protein, and cellular localization of the protein.To our surprise, the heterozygous female mouse brains had no differences in these parameters compared to Wt mice.Building on the hypothesis based on observations in carrier human females, that varying levels of skewed X-inactivation may contribute to different disease severity (Evidente, 2004), we hypothesize that in the brains of these genetically identical mice, extreme skewing towards the non-mutant chromosome supported normal brain development, and TAF1 expression and localization.
We speculate that in strongly symptomatic human carrier females the skewing mechanism is not as robust as in mice, perhaps because it is negatively influenced by other genes in the genome, which vary between human individuals but not in the mice.
Exploring the phenotype of the 1loxP heterozygous female mice further, we discovered that they were indeed symptomatic.These mice exhibited a significant increase in weight, increased in weight with age, and reduced movement in the open field test, compared to control mice.Thus, despite hypothesizing that skewed X-inactivation normalized the brains overall, we postulate that this mechanism was not totally effective even in mice.Thus, we suggest that a small subset of neurons in the brain have been negatively impacted by the loss of Taf1, resulting in these phenotypes.Importantly, these mouse phenotypes suggest new avenues for clinical investigation of disease symptoms in human carrier females.
Beyond our studies, this new conditional cre-loxP mouse model for Taf1, and the cell lines derived from it, have the potential to reveal many aspects of TAF1 mammalian biology, which are still largely unknown.This includes, but is not limited to, the role of TAF1 dysregulation in cellular, developmental, and human disease processes.Depending on the Cre-expressing mouse strain this conditional allele is bred to, it can be used to create deletions in many different specific cell types or developmental times.Finally, and most critically, this mouse will be a platform for the advancement of therapeutics for TAF1 neurodevelopmental and neurodegenerative disease.

Animal care and generation of cKO allele
All mice for all experiments were housed and bred in the pathogen-free Transgenic Animal facility at the Centre for Molecular Medicine and Therapeutics (CMMT) of The University of British Columbia (UBC).All mouse work was performed following protocols approved by the UBC Animal Care Committee (protocol numbers A19-0204, A21-0140) in accordance with guidelines determined by the Canadian Council on Animal Care.
The online tool Benchling (https://benchling.com/crispr)was used to design guide RNAs using the Streptococcus pyogenes Cas9 nuclease, to insert loxP sites surrounding exon 8, and also exon 8' when present (Fig. 1).The best 5' (intron 7) and 3' (intron 8, or 8') guides were selected based on the predicted on-target and off-target effects; and sense and antisense templates designed for the insertion of two loxP sites.Using the 10 µL Neon transfection system (Thermo Fisher Scientific, Waltham, MA, USA), to deliver CRISPR components into ESCs as previously described in Mohanna et al. 2020(Mirjalili Mohanna, 2020), the top two 5' and 3' guides were tested in combination with four templates.
Commercially synthesized CRISPR RNA (crRNA) and trans-activating crRNA (tracrRNA) (GenScript, Piscataway, NJ, USA) were annealed at a 1:1 ratio by incubating at 95°C for 5 min to generate guide RNA.The ssODN templates had 100 bp homology on each side of the Cas9 cleavage sites, and the insertion of two loxP sites at the cut sites (Integrated DNA Technologies, Coralville, IA, USA).A purified, highly concentrated, Cas9 nuclease protein with a nuclear localization signal was used (CP02; PNA Bio, Thousand Oaks, CA, USA).An RNP complex was formed by combining Cas9 with annealed guide RNA, separately, by incubating at room temperature (RT) for 15 min.Immediately prior to electroporation, ssODN template, was added to the RNP complex, and 3 x 10 5 ESCs were used per reaction the Neon optimization setting 14.At 48 hr post-electroporation, cells were lysed in a tissue homogenization buffer (THB) as previously described (Mirjalili Mohanna, 2020), and the best combination of 5' guide, 3' guide, and template, was determined by PCR using the 5' loxP and 3' loxP primers described in Table S2.
Optimal CRISPR guide Elizabeth M. Simpson 23 (cgEMS23) and cgEMS30 along with optimal template oligonucleotide EMS6395 (oEMS6395) (Table S1) were used to generate the Taf1-5'-loxP mouse via cytoplasmic injection, as previously described (Mirjalili Mohanna, 2020).In brief, 0. Resulting pups were characterized by ear notching at postnatal day 21.Ear notches were digested in THB with protein kinase and genotyped by PCR using the 5' loxP and 3' loxP primers described in Table S2.One male founder was recovered, which only contained the 5 loxP site (Fig. 1C).This male passed the new allele on to offspring, and was used to establish the Taf1-5-loxP strain, which was then backcrossed to B6 two -three generations (N2-N3), and used for a second round of CRISPR to introduce the 3 loxP site.
Benchling was used as above to design three 3' (Intron 8, or 8') guides and three templates, with 60 bp homology on each side of the Cas9 cleavage sites, and the insertion of a loxP site.The three combinations were tested in ESCs and characterized using the 10 µL Neon transfection system as described above, and the optimal set determined by PCR.
Resulting pups were characterized as above using the 3' loxP primers described in Table S2.This resulted in three founders, one female, and two males, which contained both the 3' and 5' loxP sites (Fig. 1D).The female and one male each passed the new allele to offspring, resulting in the new strains C57BL/6J-Taf1 em1Ems and C57BL/6J-Taf1 em2Ems .
Numbers of mice used were determined by the need to obtain at least one founder capable of establishing the 2loxP strain.Statistical tests were not used.Mice were excluded only by genotype, and the investigator was not blind to the genotype.

Samplix indirect sequence capture
Samplix sequencing was performed as previously described (Devoy, 2021) using primers specific to this project (Table S3).High molecular weight genomic DNA from prospective founder mice was first encapsulated into pico-litre sized droplets using an Xdrop microfluidics device.Fluorescent in-droplet PCR (dPCR) using the primers at the target locus labelled droplets harbouring target site DNA fragments, which were subsequently isolated via flow cytometry sorting, and amplified via multiple displacement amplification prior to Oxford nanopore sequencing.In cases of DNA degradation, enrichments obtained with both the main and backup primer sets were pooled and sequenced.The investigator was blind to the mouse genotype during the analysis.

Breeding of cKO allele to ubiquitous Ella-cre
Standard mouse breeding strategies involved trios, 2 females and 1 male in a cage.

Timed matings, blastocyst genotyping, and ESC generation
Timed pregnancies were achieved as previously described (de Leeuw, 2016), using crowded females, experienced studs, and plug checking of females to determine the date of copulation.B6-Taf1-2lox and B6-Taf1-1lox heterozygous females were bred with either B6 or 129S1/SvImJ (JAX stock# 002448) males, and on E9.5 or E3.5, pregnant dams were sacrificed by cervical dislocation and their uterus removed.At E9.5, deciduum were isolated and dissected to remove the yolk sac/embryo.Deciduum and yolk sacs/embryos were imaged using a Leica MZ125 microscope with a CoolSnap-Pro CF camera (Leica Microsystems, Wetzlar, Germany).Following imaging, yolk sacs were digested in THB, and samples were genotyped for sex and loxP status, using the primers in Table S2.At E3.5, single blastocysts were isolated and either used directly for genotyping or for ESC generation.
DNA derived from a single E3.5 blastocyst was prepared and used for PCR according to the method described in Sakurai et al. 2014(Sakurai, 2014).Briefly, blastocysts were flushed from the uterine horns and cultured in EmbryoMax KSOM with 1/2 Amino Acids, glucose, and Phenol Red (MilliporeSigma) and individual blastocysts were transferred to a 0.2 mL thin-walled PCR tube, and 10 µL blastocyst lysis buffer was added (125 μg/mL proteinase K, 100 mM Tris-HCl (pH 8.3), 100 mM KCl, 0.02% gelatin, 0.45% Tween 20, and 60 μg/mL yeast tRNA (Ambion, Carlsbad, CA, USA)).PCR tubes were then incubated at 56°C for 10 min followed by 95°C for 10 min.Lysis was followed by Whole Genome Amplification using the REPLI-g Mini Kit (Qiagen, Hilden, Germany) following the manufacturer's protocol.This was followed by standard genotyping for sex and loxP status, using the primers in Table S2.
On day 7, the cells were trypsinized to one well of a 24-well plate containing 1 mL of 100% FBS-ESC media, with daily media replacement.Once confluent, wells containing ESC colonies were expanded for freezing (on MEFs) and for DNA (on gelatin).Cells were washed and harvested in THB, and samples were genotyped for sex and loxP status, using the primers in Table S2.
Numbers of female mice used in each timed pregnancy experiment was adjusted based on the expected embryo harvest and need to test loxP:Wt significance.A hypothesis of 1:1 ratio allowed us to model mouse numbers for a significance of p = 0.05.Embryo numbers and statistical tests are presented in Fig. 3 and the Statistical Analysis section.No embryos were excluded.The investigator was blind to the genotype when dissecting and scoring E9.5 decidua as being empty or containing an embryo.They were also blind to the genotype when doing PCR for E9.5, E3.5, and ESCs.Otherwise, the investigator was not blind to the genotype.

Brain RNA extraction and RT-qPCR
For each mouse, total RNA was extracted from one brain hemisphere using the RNeasy kit (Qiagen) according to the manufacturer's instructions.Tissue was homogenised on ice using a TissueRuptor® (Qiagen) in Buffer RLT with 1% β-mercaptoethanol.Final extracted RNA was Disease Models & Mechanisms • DMM • Accepted manuscript eluted in DNase-and RNase-free water.Amounts of RNA were equalised and cDNA was generated using the SuperScript™ First-Strand Synthesis System for RT-PCR (11904-018;Thermo Fisher Scientific).RT-qPCR was undertaken to determine expression of all isoforms of Taf1 (Table S4).Ct values were normalised by those of β-actin (Actb) as an endogenous control.Minus template controls were run for every sample for all reactions.
Since this is the first Taf1 knockout mouse strain studied, there was no previous data on which to model effect size for significance of p = 0.5.Thus, n=6 was calculated by the "resource equation" method (Charan, 2013) and based on a previous publication showing a quantifiable effect of genetic knockout by RT-qPCR with 3-7 experimental mice (van den Hoogenhof, 2017).The experiment was repeated for individual mouse samples that were identified and excluded as outliers in the Grubb's outliers test, but no animals were excluded from the analysis.No randomization was used, and the investigator was not blind to the genotype.The data met the assumptions for the ANOVA statistical test that was used.

Protein isolation and western blotting
For protein isolation from whole-brain tissue, one brain hemisphere was dissected in icecold PBS before snap freezing.Samples were homogenized in ice-cold radioimmunoprecipitation assay buffer (150 mM sodium chloride, 50 mM Tris, 1% NP-40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulphate) plus Protease Inhibitor Cocktail I (Millipore, Billerica, MA) by mechanical disruption using a TissueRuptor II (Qiagen).
Homogenates were shaken on ice at 70 RPM for 2 h at 4°C then centrifuged for 20 mins at 15,000 x g at 4°C.Supernatants were removed and stored at -80°C until use.Total protein concentration was determined by Bradford assay using Protein Assay Dye Reagent (Bio-Rad, Hercules, CA).Equal amounts of total brain protein were denatured in LDS denaturing buffer (Thermo Fisher Scientific) and -mercaptoethanol and incubated at 95°C for 5 min prior to separation by SDS-PAGE gel electrophoresis.
Protein samples (40 μg) from individual mice, were loaded onto precast NuPAGE 7% Tris-Acetate gels (Thermo Fisher Scientific) and electrophoresed at 80 V for 4 h in an ice bath using NuPAGE Tris-Acetate Running Buffer (Thermo Fisher Scientific).Proteins were transferred to 0.2 μm nitrocellulose membranes using a Trans-Blot Turbo transfer system (Bio-Rad) High Molecular Weight Program at 1.3 A for 30 min.Membranes were stained for total protein with 0.1% (w/v) Ponceau S in 1% (v/v) acetic acid for 5 min prior to blocking for Disease Models & Mechanisms • DMM • Accepted manuscript 1 h at RT in Intercept PBS Blocking Buffer (LI-COR Biosciences, Lincoln, NE).Membranes were incubated overnight in primary antibodies diluted in Intercept PBS Blocking Buffer (LI-COR Biosciences) at 4°C (TAF1 177-4 (1:1000); -actinin (1:2000); -tubulin (1:10000)).All primary antibodies used are validated and specific monoclonal antibodies are described in Table S5.Membranes were then washed three times for 10 min in PBS containing 0.05% Tween-20 (PBS-T) and incubated for 1 h at RT with the relevant IRDye-secondary antibody (LI-COR Biosciences) diluted in Intercept PBS Blocking Buffer (LI-COR Biosciences).Finally, membranes were washed three times for 10 min in PBS-T and once for 10 min in PBS.
Antibody binding was visualised using the Odyssey CLx Imaging System (LI-COR Biosciences) and ImageJ was used for band density analysis and total protein quantification.
The same mouse brains (other hemisphere) and N values were used as in the brain RNA extraction and RT-qPCR section.Reproducible results were obtained through multiple blots, and the experiment was repeated for mouse samples that showed large variation in protein expression, which following quantification were excluded as outliers in the Grubb's outliers test, but no animals were excluded from the analysis.No randomization was used and the investigator was not blind to the genotype.
Relative signal of the TAF1 antibody compared to total protein loading, and to two different internal loading controls (-actinin and -tubulin), was calculated and the values were normalized to mean relative signal of control samples electrophoresed on the same gel.The mean of technical replicates was calculated and used for ANOVA, where biological replicates were used as the experimental unit, and the assumptions for this test were met.

Immunofluorescent staining and quantification
For histological analyses, mice were perfused with PBS so that blood was completely removed.Mice were then perfused with 50 ml of 4% PFA.Brains were removed and fixed in 4% PFA for 24 h at 4C.Samples were then washed twice in PBS and stored in PBS containing 0.05% sodium azide at 4C.Brains were sectioned on a vibrating microtome at thickness of 40 m and stored in PBS containing 0.05% sodium azide at 4C until staining.
Slides were imaged at high magnification using a Zeiss LSM980 Confocal microscope, and whole sections were imaged using an Axio Scan Z1 slide scanner.
To quantify the area of different brain regions, an H&E staining kit (ab245880, abcam) was used to stain serial coronal sections close to those in Fig. 6.ImageJ was used to quantify total brain area, including striatum and cortex.To quantify the number of CTIP2and TAF1-stained nuclei in Fig. 7, stained nuclei were counted as a percentage of the total nuclei labelled with DAPI.Quantification was performed using ImageJ.
We found little to no variability when 2 mice and 3 sections per genotype were studied.When differences were observed, defined by preestablished criteria such as photobleaching or overexposure, the image was excluded and the experiment repeated for that mouse with a Wt control for comparison.No randomization was used and the investigator was not blind to the genotype.

Behaviour Testing
The 2loxP heterozygous or 2loxP homozygous female control mice and 1loxP heterozygous female experimental mice were selected based on age matching to the 1loxP heterozygous females, and all mice were aged to adulthood.No mice within the age-matched groups were Open field locomotion was evaluated under regular light conditions and quantified by video analysis from an overhead camera, using EthoVision XT 14.0 software (Noldus, Leesburg, VA).Mice were allowed to acclimatize to the testing room for 10 min before initiating testing.Equipment was cleaned with ethanol before and in between each trial.A single 10-min trial was conducted using B6 (black) mice in white 50 cm × 50 cm × 20 cm arenas, with 4 arenas per trial.The centre zone was defined as a square covering 16% of the total arena (20 cm × 20 cm central square).Total distance travelled, duration of time in centre, and frequency of crossing into centre were quantified by video analysis software.
Motor coordination was assessed by accelerating rotarod (UGO Basile, Italy) with training as previously described in Life et al. 2023(Life, 2023).Briefly, training took place over three consecutive days, with three trials performed each day, separated by intervals of 1 h.In each trial, mice were placed on the rotarod at a fixed speed of 15 RPM for up to 2 min.Mice were then rested for 5 days before entering the testing phase.In the testing phase, mice were placed on the rotarod with acceleration set from 5 RPM to 40 RPM over 5 min, with a readout of latency to fall.A total of three trials were performed in a single day, separated by intervals of 1 h.Latency to fall over the three trials was averaged and used for analysis.
Since this is the first Taf1 knockout mouse strain studied, there was no previous data on which to model effect size for significance of p = 0.5.Thus, mouse numbers were based on the authors' previous experience with mouse model behaviour testing (Abrahams, 2005;Hossain, 2004;Wong, 2010).The experiment was separated into two groups to test reproducibility over time (months), and we report reproducible results.The standard Disease Models & Mechanisms • DMM • Accepted manuscript statistical tests used were the unpaired 2-tailed t-test, and the 2-tailed correlation test, with further details given in the Statistical Analysis section.

Statistical analysis
All tests were performed in GraphPad Prism 10, which was accessed November, 2023.For Figs. 2 and 3, the binomial test was used to analyse the data since it is an exact test to compare two categories.For female:male ratios, an unpaired 2-tailed test was used because the expected probability is symmetrical.However, for loxP:Wt ratios, an unpaired 1-tailed test was used because the null hypothesis anticipated only the loss of loxP mice, embryos, or cell lines.For Figs.4A and 5B, the "resource equation" method was used to determine the number of experimental mice since it is appropriate when no previous data is available (Charan, 2013), such as for this new mouse model.For Fig. 4B and 5C, the unpaired 2-tailed t-test was used to compare male mouse groups.For these RT-qPCR and western blot quantifications, samples were excluded based on the Grubb's outliers test (alpha = 0.05, two-tailed), which is the standard method used when there is one clear outlier in the dataset (Barnett, 1994;Iglewicz, 1993).For Fig. 8 (A,B, E-H), the unpaired 2-tailed t-test was used to quantify the difference between the means for each variable from the two groups.S2.
To assay for mosaicism, notches were studied from both ears (mouse number 1: L, left; R, right).(C) Male 5 loxP founder.Primers oEMS6445/oEMS6446 indicated the presence of the 5 loxP by a size shift (354 (not present) to 388 bp).Primer oEMS6443 specifically binds the 3 loxP and with oEMS6444 indicated its presence.L, 1kb DNA ladder.(D) Female (1) and male (2) founders also carried the 3 loxP, with the female founding strain C57BL/6J-Taf1 em2ems .Primers oEMS6447/oEMS6448 indicated the presence of the 3 loxP by a size shift (362 to 396 bp (faint indicating mosaics)).Primer oEMS6449 specifically binds the 3 loxP and with oEMS6447 indicated its presence in both founders.L, 100 bp DNA ladder; NT, no template.was generated and then maintained on the C57BL/6J background by backcrossing the Xlinked allele typically through the female.Over all generations N2-5, statistical analysis showed the female:male ratio was not significantly different from expected, and regardless

Disease Models & Mechanisms • DMM • Accepted manuscript
and showed the expected DNA sequence in a ~44.8 kb stretch of the targeted locus Disease Models & Mechanisms • DMM • Accepted manuscript (chrX:101522350-101567153), including ~34.8 kb with coverage of >100x (chrX: 101524973-101559808), with >1,000X over the region of interest.
Numbers of mice were determined by the needs of the subsequent experiments and are presented in Fig. 2. Statistical tests are also presented in Fig 2 and the Statistical Analysis section.No mice were excluded, nor was the investigator blind to the genotype.
excluded.Two rounds of behaviour testing were carried out, each with 16 2loxP Disease Models & Mechanisms • DMM • Accepted manuscript heterozygous or 2loxP homozygous female control mice and 16 1lox heterozygous female experimental mice.The first round included both open field and rotarod testing, the second round only included open field.Two weeks prior to the testing, mice were singly housed and constrained randomization was used to assign them to a trial.Each trial consisted of 4 mice (n=2 2loxP and n=2 1loxP), and within each trial a random number generator was used to assign each mouse a number from 1 to 4, which indicated which of the four arenas (open field) or four lanes (rotarod) the mouse would be placed in.The day before each open field trial, mice were weighed.For all experiments, the investigator was blind to the genotype and ID of the mice during testing, and to the genotypes of the groups during statistical analysis.

For
Fig. 8 (C,D, I,J), the 2-tailed correlation test was used to quantify the correlation between two variables.The value of the Pearson correlation coefficient (r) was calculated, in addition to a p value indicative of a significant different from no correlation (r = 0).

Fig. 2 .
Fig. 2. Breeding produced Taf1 1loxP heterozygous females, but no 1loxP hemizygous males were born.Statistical analyses used the Binomial Test; 2 tailed for sex ratios, but 1tailed for loxP:Wt since only loss of loxP animals was anticipated.(A) The cKO 2loxP allele