An isogenic cell line panel identi es major regulators of aberrant astrocyte proliferation in Down syndrome

Keiji Kawatani Osaka University Toshihiko Nambara Osaka University Nobutoshi Nawa Tokyo Medical and Dental University https://orcid.org/0000-0001-6785-7867 Hidetaka Yoshimatsu Osaka University Haruna Kusakabe Osaka University Katsuya Hirata Osaka University https://orcid.org/0000-0003-3148-9892 Akira Tanave RIKEN Center for Biosystems Dynamics Research Kenta Sumiyama RIKEN https://orcid.org/0000-0001-8785-5439 Kimihiko Banno Nara Medical University Hidetoshi Taniguchi Osaka University https://orcid.org/0000-0002-1015-7760 Hitomi Arahori Osaka University Keiichi Ozono Osaka University Graduate School of Medicine Yasuji Kitabatake (  ykitaba@ped.med.osaka-u.ac.jp ) Osaka University

pluripotent stem cells (iPSCs) and genome-editing technologies, the complexity of transcriptional dynamics affected by the extra copy of chromosome 21 and uctuations of gene-expression pro les across individuals or cell lines hinders the identi cation of key regulators. An innovative and excellent model system for studying the biology of DS has been developed by integrating the human X-inactive speci c transcript (XIST) gene, into DS-speci c iPSCs 19 . In one such cell line, XIST was inserted into a copy of chromosome 21 in trisomy 21 iPSCs (Tri21 iPSCs), and a long noncoding RNA induced a series of chromatin modi cations that stably silenced gene transcription across the whole chromosome in cis. Chromosome silencing occurred even in differentiated cells, and various pathologies observed in DS (including proliferative defects, impaired neural differentiation, and haematopoietic abnormalities) were successfully reversed by the transcriptional inactivation of the supernumerary chromosome 20,21 . Using this genome-silencing technology, where XIST RNA expression is regulated by the tetracycline-inducible system, enables researchers to investigate the correlation between gene-expression changes and cellular phenotypes in DS, without limitations caused by transcriptional heterogeneity and differences among cell lines.
To eliminate the biological 'noise', which can result from genetic variability, we established an isogenic iPSC panel, where all cell lines share a single genetic background by combining DS-speci c iPSCs, XISTinduced chromosome silencing, and genome/chromosome-editing technologies (Fig. 1). These cell lines were subjected to astrocytic differentiation, and comparative analysis between their gene-expression pro les and proliferative phenotypes (with a common genetic background) was performed. Once XISTinduced silencing was stabilised, the transcriptional levels of most genes were continuously suppressed after the removal of doxycycline (Dox). However, the enhanced proliferative phenotype of APCs in DS, which was suppressed by chromosome silencing, returned to aberrantly accelerated conditions by Dox removal. Careful analysis of this discrepancy between transcriptional and phenotypic responses enabled us to narrow down the causative genes responsible for APC overproliferation. We further established various types of systematically designed partial trisomy 21 iPSCs (Partial-Tri21 iPSCs), leading to the identi cation of two responsible genes, namely dual-speci city tyrosine-phosphorylation-regulated kinase 1A (DYRK1A) and phosphatidylinositol glycan anchor biosynthesis, class P (PIGP).

Results
Generation of an isogenic iPSC panel for disease modelling of trisomy 21. We previously generated a patient-derived Tri21 iPSC line that contains one paternal copy and two maternal copies of chromosome 21 22 . Using this DS-speci c iPSC line, we further generated a corrected disomy 21 iPSC line (cDi21 iPSC), in which a single copy of chromosome 21 was arti cially removed from a Tri21 iPSC line 23 , and a partial trisomy 21 iPSC line (Partial-Tri21 iPSC) in which a 4-megabase (Mb) region corresponding to a 'Down syndrome critical region' was selectively deleted only from the paternal chromosome 21 in Tri21 iPSCs ( Fig. 1) 22 . An XIST-mediated chromosome 21-silencing system was generated by inserting Dox-inducible XIST complementary DNA (cDNA) into one copy of chromosome 21 in Tri21 iPSCs, as described previously with a modi cation in terms of the Cre recombinase-mediated cassette exchange ( Supplementary Fig. 1a -c) 19 . The resultant iPSC clone (XIST-Tri21 iPSC) exhibited a typical morphology, expression of pluripotent markers, and a trisomy 21 karyotype ( Supplementary Fig. 2a, b). Chromosome 21 in XIST-Tri21 iPSCs contained one paternal chromosome and two maternal chromosomes, which is consistent with that in the original iPSC ( Supplementary Fig. 2c). Administration of Dox for 3 weeks (D+) successfully induced XIST RNA expression and the accumulation of H3K27me3, a hallmark of heterochromatin, which led to the transcriptional silencing of genes on chromosome 21 ( Supplementary   Fig. 3a -d). However, less than 45% of XIST-Tri21 iPSCs were H3K27me3-positive even after 3 weeks of Dox administration (Supplementary Fig. 3e). This relatively low e ciency of H3K27me3 induction was continuousely observed in single-cell-derived clones of XIST-Tri iPSCs ( Supplementary Fig. 3f). Moreover, neither XIST RNA expression nor H3K27me3 marks were detected in NPCs differentiated from XIST-Tri21 iPSCs (XIST-Tri21 NPCs), which was accompanied by a lack of expression of reverse tetracycline transactivator (rtTA), which was inserted in the AAVS1 safe harbour locus on chromosome 19 ( Supplementary Fig. 4a, b). To compensate for this induction failure by the tetracycline-regulated system, the rtTA was additionally transduced into NPCs differentiated from XIST-Tri21 iPSCs, using a piggyBac (PB) transposon vector and a hyperactive PB transposase 24 (Supplementary Fig. 4b). The insertion of PB-rtTA increased the expression levels of rtTA in NPCs, leading to a signi cant elevation of XIST RNA expression after Dox administration ( Supplementary Fig. 4c, d). H3K27me3 marks were detected in approximately 90% of Dox-treated NPCs, suggesting that a su cient amount of rtTA was required to induce XIST ( Supplementary Fig. 4e, f).
XIST-mediated chromosome silencing affected the overproliferative phenotypes of DS APCs. Among the various pathological features of DS, we focused on the astrocyte population, which is aberrantly increased in the brains of individuals with DS. XIST-Tri21 NPCs transfected with an rtTA-expression vector were differentiated to the astrocyte lineage, according to a previously described protocol 25 (Fig. 2a). Glial brillary acidic protein (GFAP) and S100β (typical astrocytic markers) were detected in over 90% of the differentiated cells (Fig. 2b). Moreover, nearly all differentiated cells were positive for CD44 or vimentin (astrocyte-restricted precursor cell markers) 10,26 , but negative for SOX1 (an early marker for neural stem cells), indicating that the cells differentiated to APCs (XIST-Tri21 APCs). XIST-Tri21 APCs exhibited su cient rtTA expression and Dox-dependent induction of XIST RNA expression and H3K27me3 marks ( Fig. 2c -e, Supplementary Fig. 5a). The expression levels of genes on chromosome 21 in Dox-treated XIST-Tri21 APCs were lower than those in Dox-untreated cells, suggesting that XIST-mediated chromosome silencing was successful induced in APCs (Fig. 2f).
Cell-proliferation assays performed using 5-ethynyl-2′-deoxyuridine (EdU) in Tri21 APCs showed a higher proliferation rate than isogenic euploid APCs (cDi21 APCs). Dox administration (D+) did not affect the basal proliferative ability of simple Tri21 APCs (i.e., those without XIST cDNA), but signi cantly decreased the proliferation rate of XIST-Tri21 APCs to a level similar to that of cDi21 cells ( Fig. 2g -i, Supplementary   Fig. 5b, c). Furthermore, removing Dox (D remov ) for 3 weeks after the initial 3-week Dox treatment increased the proliferation rate of XIST-Tri21 APCs to that of Dox-untreated cells (Fig. 2g -i). This phenotypic change was accompanied by reduced XIST expression (Fig. 2c), suggesting that expression of the genes responsible for APC overexpression was suppressed by chromosome silencing, which was reversed by Dox removal.
XIST -mediated chromosome silencing was generally maintained at least 3 weeks after Dox removal.
Previous reports showed that inducible expression of murine Xist initially caused reversible chromosome inactivation in undifferentiated cells, and then irreversible inactivation after differentiation 27 , whereas forced expression of human XIST cDNA in somatic cells resulted in reversible silencing 28 . To assess how induction and depletion of the ectopic XIST affect transcriptional dynamics in human differentiated cells, gene expression and H3K27me3 histone-methylation pro les were analysed in four cell lines, i.e. cDi21 APCs or XIST-Tri21 APCs, with (D+), without (D−) Dox, and at 3 weeks after Dox removal (D remov ).
Overall, gene expression from chromosome 21 was higher in D − APCs than in cDi21 APCs, probably due to the gene-dosage effects of trisomy 21. Consistent with previous data indicating that XIST expression was essential for the initiation of chromosome silencing, but was not required for chromosome maintenance 19,27,29 , Dox treatment effectively suppressed the expression of genes located in chromosome 21, and this suppression was maintained overall 3 weeks after Dox was removed ( The 4-Mb region on chromosome 21 was critical for aberrant APC proliferation in DS. Although the results of RNA sequencing (RNA-seq) and chromatin immunoprecipitation-sequencing (ChIP-seq) analyses indicated that XIST-mediated chromosome silencing was preserved for 3 weeks after Dox removal and that D remov APC proliferation apparently returned to the accelerated condition, similar to that of D − cells.
To explore underlying mechanisms that could explain this difference, we focused on a shape change in the violin plot for the D remov cells. In the violin plot, other cell lines showed uniform distribution, but the D remov cells showed a bimodal distribution, as evidenced by the occurrence of an additional hump in the plot (indicated by the arrowhead in Fig. 3a), which suggested the existence of a small subgroup of genes whose expression rapidly returned after the removal of Dox-induced silencing. We speculated that this phenomenon was simulated in component 1 of our principal component analysis (PCA), in which the values of D remov lines moved relatively close to those of the D − lines, while drifting away from those of the D + and cDi21 lines (Fig. 4a).
To identify causative genes among the list of genes in component 1 (Supplementary Table 2), we narrowed down the candidate genes as follows (Fig. 4b). Among 517 genes located on chromosome 21, nearly one-third of the genes (178 genes) showed positive read counts in at least one of the APC lines, and 56 genes showed > 1.5-fold increased expression, when compared with the corresponding expression levels in cDi21 APCs. The expression levels of 21 coding genes were signi cantly decreased in D + cells, suggesting that these genes were involved in the gene dosage-dependent phenotypic alteration found in XIST-Tri21 APCs.
We previously generated a genome-edited Partial-Tri21 iPSC line in which a ~ 4-Mb segment between RUNX1 and ETS2 was deleted from a single copy of chromosome 21 in Tri21 iPSCs (Fig. 4c) 22 . This 4-Mb segment corresponds to the critical region for DS pathogenesis, commonly referred to as the Down syndrome critical region. Notably, the deletion of this 4-Mb region from one chromosome 21 led to a signi cant decrease in the proliferation rate of APCs, down to the same level found with cDi21 APCs, indicating that this segment was responsible for the aberrant proliferative properties of DS APCs (Fig. 4d, e). Among the 32 genes located in this 4-Mb region, three genes (DYRK1A, DSCR3, and HLCS) were ranked in the top 10, and two genes (MORC3 and PIGP) were in the 25th and 31st positions, as essential elements of PCA component 1 (Supplementary Table 2). Consistent with these results, RNA-seq data demonstrated that the gene-expression levels of these genes reverted to trisomic levels after Dox removal, except for MORC3 (Supplementary Table 3). The gene-expression alterations were validated by quantitative reverse transcriptase-polymerase chain reaction (qRT-PCR) analysis for the DYRK1A, PIGP, and DSCR3 genes, but not for the HLCS gene (Fig. 4b, f). Signi cantly increased expression and elimination of H3K27me3 deposition were observed for these genes in D + and D remov cells, respectively ( Fig. 4g, Supplementary Fig. 9a, b).
DYRK1A potently regulated APC proliferation in a gene dosage-dependent manner. DYRK1A was identi ed as the most potent candidate gene in the PCA list (Supplementary Table 2). However, it is well established that DYRK1A exerts dose-dependent antiproliferative activity in several cell types, especially in neural precursor cells 30,31 . To investigate whether DYRK1A conversely promotes APC proliferation, and whether DYRK1A acts independently in DS pathophysiology or cooperates with other molecules synergistically, we prepared several genome-edited Tri21 iPSC lines, in which DYRK1A was targeted in one or two chromosomes in the XIST-Tri21 iPSC line (DY +/+/m -and DY +/m/m -XIST-Tri21 iPSC, respectively; Fig. 5a).
Single-nucleotide polymorphism (SNP) analysis of mRNAs extracted from XIST-Tri21 APCs (with or without Dox treatment) showed that a SNP in the ETS2 gene (rs457705) derived from the paternal (P) allele (T→G) was conserved after XIST-mediated chromosome silencing. This result indicated that the XIST cDNA cassette was inserted in one of the two maternal copies (M1 and M2) of chromosome 21 (Supplementary Fig. 10; hereafter, the XIST-inserted maternal chromosome is referred to as M2). The targeting cassette was inserted into the DYRK1A locus in XIST-Tri21 iPSCs using the clustered regularly interspaced short palindromic repeats (CRISPR)-CRISPR-associated protein 9 (Cas9). Short-tandem repeat (STR) analysis of targeted and non-targeted alleles in isolated iPSC colonies revealed that six types of genome-edited cell lines were obtained (single or double targeting of the M1, M2, or P alleles; Fig. 5b, c). DYKR1A expression in APCs decreased to similar levels found in DY +/+/m -XIST-Tri21 APCs and were even lower in DY +/m/m -XIST-Tri21 APCs than in Di21 APCs (Fig. 5d). Previous reports showed that the Janus kinase-signal transducer and activator of transcription (JAK-STAT) pathway is one of the main targets of DYRK1A and that it is involved in gliogenic differentiation machinery in NPCs 13 . Consistent with the decreased expression of DYRK1A, Ser727-phosphorylated STAT3 was signi cantly decreased in DY +/+/m and DY +/m/m APCs, whereas no detectable changes in STAT3 protein levels were observed in these cell lines (Fig. 5e, Supplementary Fig. 11). Accelerated APC proliferation proportionally decreased in both lines, and the double-targeted (DY +/m/m ) APC line showed more severe proliferative impairment, indicating that DYRK1A regulated APC proliferation in an expression level-dependent manner ( Fig. 5f, g).
Intriguingly, the DY +/+/m APC line, which contains two copies of the normal DYRK1A allele, still retained higher proliferation rates than the cDi21 line, suggesting that correction of the DYRK1A gene dose was not su cient to fully reverse APC overproliferation.
To distinguish the roles of other genes on chromosome 21 from that of DYRK1A, both the DY +/+/m and DY +/m/m APC lines were subjected to XIST-mediated chromosome silencing. Given that XIST expression selectively inactivated the M2 chromosome in cis, these cell lines (with or without Dox treatment) exhibited various combinations of 'DYRK1A-expression levels' (zero, one, or two copies) and To con rm the effect of DYRK1A on proliferation, APCs and NPCs were treated with a DYRK1A inhibitor. A recently developed speci c inhibitor, folding intermediate-selective inhibitor of DYRK1A (FINDY), which interferes with DYRK1A protein folding 32 , signi cantly reduced Tri21 APC proliferation (Fig. 5m, n), without effecting that of Tri21 NPCs (Fig. 5o, p). These ndings indicate that DYRK1A potently regulated APC proliferation in a dose-dependent manner.
PIGP, but not DSCR3, was involved in the proliferative pathophysiology of APCs in DS. Next, we used APCs to investigate whether two other candidate genes (PIGP and DSCR3) were involved in pathological features associated with DS. NPCs differentiated from Partial-Tri21 iPSCs were transfected with piggyBac transposon vectors containing human PIGP or DSCR cDNA, under the regulation of the Tetinducible system, and then differentiated to the astrocyte lineage (Fig. 6a). Both PIGP and DSCR3 expression increased in the corresponding Partial-Tri21 APC transfectants following Dox administration (Fig. 6b, Supplementary Fig. 13a). Proliferation rates in APCs, which were reverted to a normal level by the deletion of a 4-Mb critical region in one copy of chromosome 21, were increased by forced expression of PIGP (Fig. 6c, d). We con rmed this result using small-interfering RNA (siRNA)-mediated knockdown of PIGP in Tri21 APCs ( Fig. 6e -g). However, the dose-dependent activity of PIGP was relatively less effective on APC proliferation, compared with that of DYRK1A, whose slight expression differences produced robust changes. However, DSCR3 did not signi cantly impact proliferation when DSCR3 was overexpressed in Partial-Tri21 APCs or when DSCR3 was knocked down in Tri21 APCs (Supplementary Fig. 13b -f). Taken together, these results imply that PIGP is another crucial molecule for APC proliferation that is involved in DS pathology, together with DYRK1A.

Discussion
Emerging evidence suggests that astrocytes play crucial roles in various pathological mechanisms in the central nervous system (CNS) 33 . We focused on the aberrant proliferation of DS APCs, which is a major cause of astrocyte overpopulation in the brains of DS individuals, and explored essential regulators.
Selecting a suitable model system is crucial for exploring human diseases. Although recent advances in human iPSC technologies have provided a new approach for functional disease studies, signi cant variations in phenotypes among different iPSC lines hinders the precise analysis of human diseases 34 . Therefore, we developed an isogenic cell model that combines XIST-mediated chromosome silencing and genome-editing technology using DS-speci c iPSCs, which enabled us to perform a detailed comparative analysis of genotype-phenotype correlations between dosage-sensitive genes on chromosome 21 and proliferative dynamics in DS APCs. Recent data have revealed that human iPSCs from different donors are more divergent in terms of their transcriptomes or cell phenotypes than those originating from different somatic cell types of the same donor, indicating that iPSC heterogeneity is mainly caused by genetic differences between individuals, rather than the epigenetic memory of the somatic tissue of origin 35,36 . These ndings strongly suggest that the cell lines in our isogenic iPSC panel, all of which share a single genetic background, can provide an ideal cellular model for studying DS pathology.
Tet-regulated XIST was stably expressed after Dox treatment in both iPSCs and differentiated cells in previous studies 19,20,21 . Nevertheless, XIST RNA was not detected in our differentiated XIST-Tri21 NPCs, likely due to silencing of the rtTA inserted in the AAVS1 locus. Despite many descriptions of the AAVS1 locus as a well-validated genomic safe harbour that enables stable expression of an inserted transgene 37 , data from several studies have demonstrated that the AAVS1 locus is not as a well-validated genomic 'safe' as suggested, and transgene expression varied in both iPSCs and differentiated cells, due to DNA methylation 38,39 . Although the DNA-methylation status of the locus was not analysed in this study and other unknown mechanisms may exist, additional transduction of the rtTA transgene using the piggyBac transposon system in NPCs led to e cient XIST RNA expression, followed by chromosome silencing.
To identify critical regulators of DS APC overproliferation, we focused on the variability of genetic responses after the removal of XIST. In mouse embryonic stem (ES) cells, X chromosome inactivation (XCI) was reversible and depended on continued Xist expression, but XCI became irreversible and independent of Xist after differentiation 27 . Chromosome inactivation can be introduced by ectopic expression of human XIST in somatic cells and maintained even after removal of XIST, albeit to a different extent 19,40 . Our results were consistent with these ndings in that the transcriptional levels of chromosome 21 were stably suppressed 3 weeks after Dox removal. Notably, a small subset of genes rapidly returned from silenced disomic levels to reactivated trisomic levels, which was accompanied by reoccurrence of the pathological phenotype in DS APCs. Such a multiplicity of gene reaction in XCI has been reported in several studies, indicating that 3% -7% of mouse and 12% -20% of human genes on the inactivated X chromosome escaped from XCI 41,42 . Likewise, ectopically expressed XIST RNA on autosomes can induce silencing in cis, whereas 15% of genes consistently escape from XCI, and another 15% of genes vary in terms of whether they are subject to, or escape from, inactivation 28,43 . Data from several bioinformatic studies have demonstrated that long-interspersed nuclear element repeats on the X chromosome or short-interspersed nuclear elements (such as Alu elements) are highly enriched around escape genes on autosomes, suggesting that genomic features contribute to the e ciency of XISTmediated chromosome 21 inactivation 43,44 .
Unexpectedly, our RNA-seq data showed a certain XIST-mediated reduction of gene-expression levels in several other chromosomes, especially in chromosome 18 ( Supplementary Fig. 6). One possible explanation for this phenomenon is that trisomy of chromosome 21 might cause transcriptional alteration in a speci c subset of genes on other chromosomes. It is known that the spatial organisation of chromosome territories in the cell nucleus is linked to genomic functions and regulation 45 . Gene-rich chromosomes are located preferentially at the centre of the nucleus, whereas gene-poor chromosomes, such as chromosome 18, are located at its periphery 46 . Cell-type-speci c interaction patterns among chromosome territories are correlated with genome regulation at the global level, and the presence of a supernumerary chromosome 21 may perturb the physiological positioning of other chromosomes in the nucleus, leading to transcriptional dysregulation 23,47 . XIST-mediated silencing of extra chromosome 21 reverted trisomy-induced transcriptional changes in other chromosomes, leading to downregulated expression levels, which was less noticeable but became evident by comparing the isogenic cell lines.
Another possible explanation of the dysregulated transcription relates to the higher expression levels of XIST observed in our study. We introduced an additional rtTA into NPCs using the PB transposon vector, which generally provides robust transgene integration. After introducing the XIST transgene into mouse ES cells, silencing of autosomal genes occurred only in cell lines with high-copy transgenes, suggesting that dose-dependent regulation by XIST 48 occurred. Data from transgenic experiments indicated that XIST expression was essential, but not su cient, for XIST RNA spreading and localization 49,50 . Although XIST is a well-known cis-acting element, its trans effects on other chromosomes have not been reported. However, its robust expression, which was higher than observed physiologically, may explain the unexpected trans activity of XIST.
We identi ed two genes, DYRK1A and PIGP, as potent candidates responsible for the proliferative pathology of DS APCs. DYRK1A has been proposed to be closely involved in neural development, especially in fate speci cation and neuronal proliferation 51 . Studies conducted with human iPSCs and mouse models of DS exhibited impaired neural differentiation, which was improved by targeting DYRK1A pharmacologically or with short hairpin RNAs 52, 53 . In addition, Dyrk1a overexpression promoted astrogliogenesis in mouse cortical progenitor cells by activating the STAT-signalling pathway 13 , suggesting that DYRK1A plays a key role in cell-fate switching. However, an increased DYRK1A gene dosage attenuated neuronal proliferation rates 30 , while loss of DYRK1A function accelerated neural proliferation 51 . It is well established that DYRK1A increases the duration that cells spend in G1 phase in a dose-dependent manner by reducing cyclin D1 expression and increasing p27 KIP1 (CDKN1B) expression 31,54 . Furthermore, DYRK1A can phosphorylate p53 and subsequently induce p53-target genes such as p21 CIP1 , leading to impairment of the G1/G0-S-phase transition, resulting in attenuated proliferation 30 . These antiproliferative activities of DYRK1A are supported by a decreased number of neurons observed in humans and mouse models of DS 55,56 . In contrast to its effects on neural precursors, however, we unexpectedly found a proliferation-promoting activity of DYRK1A on APCs, which was accompanied by an alteration in the level phosphorylated STAT3.
In the CNS, STAT3 is highly expressed in astrocytes and is activated in response to multiple pathological stimuli such as ischaemia, spinal cord injury, or neurodegenerative diseases 57,58 . It remains unclear whether activated JAK-STAT signalling can stimulate astrocyte proliferation, especially during the physiological-developmental stage. Nevertheless, some ndings showed that the proliferation of reactive astrocytes in spinal cord injury was reduced by treatment with JAK inhibitors or conditional knockout of STAT3 59,60 . Although the detailed mechanism remains to be elucidated, these indirect data and our current results suggest that increased expression of DYRK1A can stimulate APC proliferation via the JAK-STAT pathway.
PIGP is a component of the glycosylphosphatidylinositol (GPI)-N-acetylglucosaminyltransferase (GPI-GnT) complex. More than 150 proteins have been identi ed as GPI-anchored proteins, which are expressed on the cell surface by being anchored to the plasma membrane 61 . Biosynthesis of mammalian GPIs is initiated by the transfer of N-acetylglucosamine (GlcNAc) to generate GlcNAc-PI through the enzymatic activity of GPI-GnT. PIGP is one of seven subunits of GPI-GnT, and mutations in PIGP that lead to reduced cell-surface expression of GPI-anchored proteins have recently been linked to early infantile encephalopathy 62 . Although the speci c function of PIGP is unclear, its overexpression has been reported to impair the membrane localisation of Wnt-signalling receptors during embryogenesis 63 . Further studies are required to elucidate how PIGP accelerates APC proliferation in DS.
Astrocytes support neuronal homoeostasis and regulate synaptic networks by promoting neuritogenesis and synaptogenesis 64 . However, DS astrocytes exert a toxic effect on the formation and maturation of neural networks and neuron survival by reducing neuronal activity, inducing morphological alterations, and promoting neuronal apoptosis 10,18,65 . In the DS brain, astrocytes may act as a primary effector in DS pathophysiology. Thus, identifying critical regulators for astrocyte overpopulation may be a critical rst step for investigating disease mechanisms and developing new therapeutic strategies for DS. Our collection of isogenic iPSC lines will provide a useful resource for conducting detailed analyses of DS.

Methods
Human-iPSC generation and culture. Human iPSCs were generated and cultured as reported 22  Scienti c) and used to transfect iPSCs as described 66 , until complete removal of the SeV genome was con rmed by PCR and immunostaining with an anti-SeV-NP antibody. The iPSC cultures were passaged every 6 to 9 days. Insertion of a lox cassette into the DYRK1A locus was performed using the CRISPR-Cas9 system (Supplementary Table 4). Single-guide RNA (sgRNA) oligos for the CRISPR-Cas9 system were cloned into the BbsI sites of the pX330-U6-Chimeric_BB-CBh-hSpCas9 vector (Addgene, #42230). The pEF1α-3G rtTA-pA cassette was cloned from a pEF1α-Tet3G vector (Clontech, #631167). Full-length XIST cDNA with loxP and lox5171 sites was cloned into the pTRE3G vector (Clontech, #631168). On the day before transfection, iPSC colonies were dissociated into single cells using TrypLE Express (Thermo Fisher Transfection of the piggyBac vector into NPC. To enhance rtTA, DSCR3, or PIGP expression, NPCs were transfected with a piggyBac vector harbouring an additional gene (rtTA, DSCR3, or PIGP), which was generated from the PB-TA-ERN vector (Addgene, #80474). The resulting vector (2 µg) and a pCMV-hyPBase vector (2 µg; a kind gift from the Sanger Institute) encoding transposase we co-transfected into NPCs (4.0 × 10 6 ) using the Neon Transfection System (settings: 1200 V, 20 ms, 2 pulses). After clone selection with puromycin (0.5 µg/mL), NPCs with the additional gene were established.
Maintenance and differentiation of APCs. The protocol used for differentiating APCs from NPCs has been described in detail 25 . NPCs were dissociated with TrypLE Express, and 2 ⋅ 10 4 cells/well were plated on Matrigel-coated 24-well plates with Astrocyte Medium (ScienCell) supplemented with 10 µM ROCK inhibitor. This medium was changed every 2 days, and the cells were passaged every 4-6 days, with dissociation using TrypLE Express. In this study, APCs were passaged 7-8 times before performing the analysis. When necessary, NPCs were differentiated into APCs via Dox administration for ve days. The Dox-treated cell lines were administered Dox for approximately 6 weeks, and the D remov cell lines were administered Dox for approximately 3 weeks, followed by growth in culture for 3 weeks without Dox. Two XIST-Tri 21 iPSC lines were generated from a male baby with trisomy 21. Both iPSC lines were differentiated into NPC lines. The NPCs were independently transfected using the piggyBac vector encoding an rtTA to generate three lines. The NPC-derived APCs were subjected to qRT-PCR analysis, cellproliferation assays, RNA-seq analysis, and ChIP-seq analysis. Three cDi21 lines, which were generated using a chromosome-elimination technique, were used as controls in this study.
Genome editing of the DYRK1A gene using the CRISPR-Cas9 system. DYRK1A targeting was performed using the CRISPR-Cas9 system. The sgRNA sequence was designed using CRISPR Direct  and processed using Volocity software (PerkinElmer).
qRT-PCR analysis. Total RNA was isolated from iPSCs, NPCs, and APCs with the NucleoSpin RNA II Kit (Macherey-Nagel). Reverse transcription was performed using ReverTra Ace qPCR RT Master Mix (Toyobo). qRT-PCR analysis was performed using Thunderbird SYBR qPCR Mix (Toyobo). Gene expression levels were normalised to the expression of β-actin (ACTB). The sequences of all primers used for qRT-PCR analysis are shown in Supplementary Table 6.
Single-cell cloning. XIST-Tri 21 iPSC colonies were dissociated into single cells using TrypLE Express, and then 1.0 × 10 3 cells were plated in a 10-cm dish with DR-4 IRR MEFs using iPSC culture medium containing 10 µM ROCK inhibitor, G418 (150 µg/ml) and Dox (2 µg/ml). On days 12-18, the resulting colonies were passaged to individual wells. These clones were expanded further, and on day 21 after Dox addition, they were xed and assessed using an immunocytochemistry method.
Cell-proliferation assay. A Click-it EdU Alexa Fluor 488 Imaging Kit (Thermo Fisher Scienti c) was used to measure cell proliferation. At 1 or 2 days before adding the thymidine analogue EdU, APCs were dissociated with TrypLE Express, and 5 ⋅ 10 3 cells/well were plated into a Matrigel-coated 96-well plate (Greiner Bio-one). On the day of EdU treatment, the cells were cultured with EdU (10 µM) for 8 h. The cultured cells were xed with PBS containing paraformaldehyde (4%) and then permeabilised with PBS containing TritonX-100 (0.5%). The cells were stained as instructed in the manufacturer's protocols.
Images were taken with an In Cell Analyzer 6000 (GE Healthcare), and EdU-positive cells were detected using the In Cell Developer Toolbox 1.9 (GE Healthcare). Cell-counting assays were performed according to a method similar to the protocol described above. At 1 or 2 days after plating the cells ( Short-tandem repeat (STR) analysis. To perform STR genotyping, DNA was extracted from iPSCs using a DNeasy Blood & Tissue Kit (Qiagen). To assess DYRK1A-targeted alleles, junctional PCR (homologous recombination+) and outside PCR (homologous recombination−) were performed using KOD FX Neo enzyme solution (Toyobo). PCR products or genomic DNA were subjected to PCR using PrimeSTAR MAX (Takara), a uorescently labelled forward primer, and a reverse primer. The nal PCR products were mixed with an internal lane standard 600 (Promega) and HiDi formamide (Thermo Fisher Scienti c) and separated by capillary electrophoresis on an ABI 310 Genetic Analyzer (Thermo Fisher Scienti c), per the manufacturer's instructions. The primer sequences are shown in Supplementary Table 5.
Allele-speci c SNP-silencing analysis. Total RNA was isolated from XIST-Tri 21 APCs (Dox-untreated cell lines and Dox-treated cell lines) using the NucleoSpin RNA II Kit. Reverse transcription was performed using ReverTra Ace qPCR RT Mix. With the resulting total cDNA, PCR was performed using primers that ampli ed a region containing an SNP (rs457705) on exon 8 of ETS2 on chromosome 21. The sequences of the primers used for ETS2 are provided in Supplementary Table 7.
Western blotting. Cells were lysed with RIPA Buffer (Fuji lm Wako) containing a mixture of protease inhibitors (Merck) and phosphatase inhibitors (Nacalai Tesque). Lysates containing equal amounts of protein were mixed with an appropriate amount of Laemmli sample buffer (2×; Bio-Rad Laboratories) and 2-mercaptoethanol and denatured at 95 °C for 5 min. The samples were separated by 7.5% sodium dodecyl sulphate-polyacrylamide gel electrophoresis and transferred to polyvinylidene uoride membranes (Bio-Rad Laboratories). The membranes were washed with Tris-buffered saline containing Triton X-100 (0.05%) and incubated with Blocking One or Blocking One-P (Nacalai Tesque) buffer for 60 min. Mouse anti-STAT3 (1:1000; Cell Signaling Technology) and rabbit anti-Phospho-STAT3 (Ser727) (1:500; Cell Signaling Technology) were used as primary antibodies. Horseradish peroxidase-conjugated anti-rabbit or anti-mouse IgG antibodies (1:2500; Promega) were used as secondary antibodies. As a control, β-actin was detected with anti-β-actin pAb-HRP-DirecT (1:2000; MBL). Blots were visualised using Clarity Western ECL Substrate (Bio-Rad Laboratories). The stained membranes were scanned with ImageQuant LAS 4000 Biomolecular Fluorescence Image Analyzer (GE Healthcare). When necessary, the antibody was stripped with Restore Western Blot Stripping Buffer (Thermo Fisher Scienti c).
Statistical analysis. All statistical analyses were performed using the EZR software. Comparisons of two groups were made using Student's t-test or Welch's two-sample t-test. We evaluated multiple comparisons using one-way analysis of variance (ANOVA) or the Kruskal-Wallis test with Bonferroni's correction. A P value of less than 0.05 was considered to re ect a statistically signi cant difference. The data presented are expressed as the mean ± standard error of the mean (SEM) or standard deviation (SD).
Data availability. The RNA-seq data and ChIP-seq data reported here are available in the DDBJ Sequenced Archive under accession numbers DRA010528 and DRA010529, respectively.