key: cord-0288472-2yn59pui
authors: Waugh, Katherine A.; Minter, Ross; Baxter, Jessica; Chi, Congwu; Tuttle, Kathryn D.; Eduthan, Neetha P.; Galbraith, Matthew D.; Kinning, Kohl T.; Andrysik, Zdenek; Araya, Paula; Dougherty, Hannah; Dunn, Lauren N.; Ludwig, Michael; Schade, Kyndal A.; Tracy, Dayna; Smith, Keith P.; Granrath, Ross E.; Busquet, Nicolas; Khanal, Santosh; Anderson, Ryan D.; Cox, Liza L.; Estrada, Belinda Enriquez; Rachubinski, Angela L.; Lyford, Hannah R.; Britton, Eleanor C.; Orlicky, David J.; Matsuda, Jennifer L.; Song, Kunhua; Cox, Timothy C.; Sullivan, Kelly D.; Espinosa, Joaquin M.
title: Interferon receptor gene dosage determines diverse hallmarks of Down syndrome
date: 2022-02-05
journal: bioRxiv
DOI: 10.1101/2022.02.03.478982
sha: 58d9df64b91693a0669467cf28b086fbca5cea07
doc_id: 288472
cord_uid: 2yn59pui

Trisomy 21 causes Down syndrome, a condition characterized by cognitive impairments, immune dysregulation, and atypical morphogenesis. Using whole blood transcriptome analysis, we demonstrate that specific overexpression of four interferon receptors encoded on chromosome 21 associates with chronic interferon hyperactivity and systemic inflammation in Down syndrome. To define the contribution of interferon receptor overexpression to Down syndrome phenotypes, we used genome editing to correct interferon receptor gene dosage in mice carrying triplication of a large genomic region orthologous to human chromosome 21. Normalization of interferon receptor copy number attenuated lethal antiviral responses, prevented heart malformations, decreased developmental delays, improved cognition and normalized craniofacial anomalies. Therefore, interferon receptor gene dosage determines major hallmarks of Down syndrome, indicating that trisomy 21 elicits an interferonopathy amenable to therapeutic intervention. One-Sentence Summary Correction of interferon receptor gene dosage rescues multiple key phenotypes in a mouse model of trisomy 21.

Triplication of human chromosome 21 (trisomy 21, T21) occurs 1 in ~700 live births, causing Down syndrome (DS) (1, 2) . People born with DS experience developmental delays, cognitive impairment, craniofacial abnormalities, and high rates of congenital heart defects (3) . These individuals also display a distinct clinical profile, exhibiting lower risk of most solid 5 malignancies, atherosclerosis, and hypertension, along with increased risk of severe respiratory viral infections, Alzheimer's disease, leukemia, autism spectrum disorders, and diverse autoimmune diseases (3) (4) (5) . Therefore, research on the mechanisms driving these phenomena will benefit not only people with DS but also those in the general population affected by conditions modulated by T21.

Interferon (IFN) signaling, a key component of the innate immune system, is hyperactive in people with DS (6). Upon receptor binding, IFN ligands induce JAK/STAT signaling and a downstream transcriptional program to elicit diverse context-dependent cellular responses, including restriction of viral replication, decreased cell proliferation, apoptosis, metabolic reprogramming, and immune activation (7). The mechanisms driving IFN hyperactivity in DS 15 are debated. Notably, four of the six IFN receptors (IFNRs) are encoded on human chromosome 21 (HSA21) and overexpressed in diverse cell types with T21: IFNAR1 and IFNAR2 for Type I IFNs, IFNGR2 for Type II IFNs, and IL10RB for Type III IFNs (6, 8) . Cells with T21 are hypersensitive to IFN stimulation (6, (9) (10) (11) and this hypersensitivity is rescued in vitro by reducing IFNR copy number (10), indicating that IFNR triplication could drive elevated IFN 20 responses in DS. However, multiple constitutive trisomies, including T21, have been shown to elevate IFN signaling through accumulation of cytosolic dsDNA and activation of the cGAS-STING pathway (12). The consequences of chronic IFN hyperactivity in DS remain to be elucidated. Importantly, mutations leading to chronic elevation in IFN signaling cause interferonopathies, a group of monogenic disorders that share some developmental and clinical features with DS (13, 14) . Therefore, elucidating the exact mechanism driving IFN hyperactivity in DS and its contribution to DS phenotypes could identify targeted therapeutic strategies to counteract the detrimental effects of T21.

In this study, we utilized whole blood transcriptome analysis in a large cohort of individuals with 5 DS to define associations between overexpression of individual HSA21 genes and inflammatory markers in DS, which revealed that only a select subset of triplicated genes, including the four IFNRs, associate with IFN hyperactivity and inflammation. We then employed genome editing to normalize gene dosage for all four Ifnrs in a mouse model of DS, which revealed that Ifnr copy number determines multiple phenotypes of DS. These results identify aberrant IFN signaling as a 10 common node underlying the etiology of vastly diverse hallmarks of the most common human chromosomal abnormality, with clear therapeutic implications for the clinical management of DS, while also advancing the understanding of the impacts of dysregulated cytokine signaling in human development.

In order to investigate mechanisms by which T21 causes IFN hyperactivity in DS, we analyzed the datasets generated by the Crnic Institute Human Trisome Project (HTP), which include whole blood transcriptome and matched plasma immune markers in 304 individuals with T21 versus 96 typical controls (D21) (see Methods, Fig. S1A , Data S1). These datasets, which were generated 20 from biospecimens collected from individuals ages 1-61 (Fig. S1B) , enabled us to complete a correlation study between overexpression of individual genes encoded on HSA21 and markers of immune dysregulation in both the transcriptome and proteome across the lifespan. The transcriptome analysis detected 173 mRNAs and lncRNAs encoded on HSA21, 90% of which were significantly upregulated in T21 relative to controls, with a mean fold change of ~1.5, consistent with the expected effect of increased gene dosage (Fig. 1A, Data S2) . Nevertheless, there was a wide range of expression of the triplicated genes among individuals with and without T21 (e.g., IFNAR1, DYRK1A, Fig. 1B) . Importantly, the transcriptome analysis identified 5 thousands of mRNAs encoded elsewhere in the genome that were also consistently dysregulated in DS (e.g., MYD88, COX5A, Fig. 1A DYRK1A expression did not, correlating instead with other pathways (Fig. 1E, Data S2) .

Multiple ISGs not encoded on HSA21 (e.g., MYD88, STAT3, TRIM25) showed strong positive 20 correlations with IFNRs but not with most HSA21 genes ( Fig. 1C-F, Fig. S1C , Data S2). In contrast, genes in the Oxidative Phosphorylation signature elevated in DS were negatively correlated with IFNR expression (e.g., COX5A, Fig. 1C, F, Fig. S1C , Data S2).

Next, we defined correlations between circulating protein levels of the general inflammatory marker C-reactive protein (CRP) and the pro-inflammatory cytokine IL6 versus expression of individual HSA21 genes among people with T21. Once again, expression of only a small fraction of HSA21 genes positively correlated with CRP and IL6 levels, including the four IFNRs (Fig.   1C, G, Fig. S1E , Data S3). For example, whereas mRNA expression for IFNAR1 correlates 5 positively with circulating levels of CRP and IL6, expression of DYRK1A is negatively correlated with both immune markers (Fig. 1H, Fig. S1F ).

Altogether, these results indicate that the heightened inflammatory state observed in individuals with DS is unlikely to be a general effect of the aneuploidy, being instead associated with overexpression of a small subset of specific genes on HSA21, including all four IFNRs. The B6.129S7-Dp(16Lipi-Zbtb21)1Yey/J mouse model, herein "Dp16", carries a segmental duplication of mouse chromosome 16 (MMU16) causing triplication of ~120 protein-coding genes orthologous to those on HSA21, including the gene cluster of four Ifnrs (15, 16) . Dp16 15 mice display key phenotypes of DS including increased prevalence of heart defects, craniofacial anomalies, developmental delays, cognitive impairment, a dysregulated antiviral response, and hyperactive IFN signaling (16) (17) (18) (19) (20) (21) . Importantly, gene triplication in the Dp16 model of DS is driven by segmental duplication rather than a freely segregating extra chromosome, akin to DS caused by Robertsonian translocations of HSA21, indicating that Dp16 phenotypes cannot be 20 explained by mere aneuploidy (16).

To define if increased gene dosage of Ifnrs contributes to DS phenotypes, we used CRISPR/Cas9 genome editing to concurrently knock out all four Ifnrs. Given that all four Ifnrs employ JAK/STAT signaling and that overexpression of each of them associates with similar inflammatory signatures in our transcriptome analysis, thus creating the potential for genetic redundancy, we designed a strategy to delete the entire 192 kb genomic segment on MMU16 encoding all four Ifnrs in wildtype (WT) C57BL/6 mice (see Methods, Fig. 2A, Data S4) .

Knockout was confirmed in potential founders by PCR (Figs. S2A-B) and Sanger sequencing 5 (Fig. S2C) . Whole genome sequencing (WGS) confirmed the deletion and did not reveal any other genomic alterations ( Fig. 2A, Fig. S2D ). Heterozygous progeny of this strain (WT 1xIFNRs ) was then intercrossed with Dp16 to normalize Ifnr copy number from three to two in a portion of Dp16 offspring (Dp16 2xIFNRs ) (Fig. S2E) . Expectedly, mRNA expression of all four Ifnrs was consistently elevated in Dp16 relative to WT littermates, but this overexpression was corrected in 10 Dp16 2xIFNRs mice (Fig. 2B) . IFNR protein expression was significantly elevated in the surface of bulk white blood cells and multiple myeloid and lymphoid immune cell lineages in Dp16 mice, but not in Dp16 2xIFNRs mice ( Fig. 2C-D, Fig. S3A-E) . Therefore, Dp16 2xIFNRs mice provide an experimental model to define the contribution of Ifnr triplication relative to the other ~120 genes triplicated in the Dp16 animal model of DS.

We previously demonstrated that peripheral immune cells from Dp16 mice are hypersensitive to IFNa and IFNg stimulation (21). Furthermore, upon chronic exposure to the viral mimetic polyinosinic:polycytidylic acid [poly(I:C)], a TLR3 agonist that induces IFNs, Dp16 mice 20 experience exacerbated weight loss and death (21). This immune hypersensitivity phenotype can be rescued by pharmacological inhibition of JAK1, a protein kinase that mediates signaling for all three types of IFNs and other cytokines (21). We thus tested the impact of reduced Ifnr gene dosage on these phenotypes. When stimulated ex vivo with IFNa or IFNg, white blood cells from Dp16 mice show significantly elevated levels of phospho-STAT1 relative to cells from WT mice, but this phenotype is rescued in Dp16 2xIFNRs mice ( Fig. S3F-G) . During chronic poly (I:C) challenge, Dp16 mice lost significantly more weight than WT littermates and had to be removed from the experiment much earlier at the human endpoint of 15% weight loss; however, 5 Dp16 2xIFNRs mice did not differ from controls in weight loss or overall survival ( Fig. 2E-F) .

Analysis of cytokine induction following poly(I:C) treatment revealed significant overproduction of TNFa in Dp16 relative to WT controls, but this was not observed in Dp16 2xIFNRs (Fig. 2G,   Fig. S3H ). TNFa is known to mediate inflammation-driven cachexia (22), and its levels correlated with weight loss in our experimental paradigm (Fig. 2G) . 10 Altogether, these results indicate that triplication of the Ifnr cluster results in increased levels of expression of all four Ifnrs, leading to dysregulated immune responses.

Congenital heart defects (CHDs) are more common in DS, afflicting around half of newborns 15 with T21 (3). To test if Ifnr gene dosage contributes to this phenotype in DS, we evaluated the frequency of heart malformations in WT, Dp16, and Dp16 2xIFNRs embryos by sectioning the entire developing heart for histological evaluation at embryonic day (E)15.5 ( Fig. 3A-C, Fig.   S4A ). In agreement with previous reports (16, 23, 24) , Dp16 mice displayed significantly elevated frequency of atrial septal defects (ASD) and ventricular septal defects (VSDs), often 20 both types of septal defects at once. Remarkably, this phenotype was corrected in Dp16 2xIFNRs mice, which did not present elevated rates of septal defects compared to WT littermates (Fig.   3D ). To define if JAK/STAT signaling is modulated by Ifnr gene dosage in the developing heart tissue, we measured levels of phospho-STAT1 by Western blot. Indeed phospho-STAT1 levels are elevated in the heart tissue of Dp16 mice relative to WT littermates, but this elevation in JAK/STAT signaling is recued in Dp16 2xIFNRs embryos (Fig. S4B) . These results indicate that Ifnr triplication disrupts normal heart development, even in the absence of obvious immune stimuli.

In order to investigate potential mechanisms underlying this phenomenon, we completed transcriptome analysis of heart tissue at E15.5, which identified ~1200 differentially expressed genes (DEGs, q<0.1 by DESeq2) in Dp16 relative to WT controls, with only 73 of these DEGs being encoded on the triplicated region ( Fig. 3E ; Data S5). GSEA of DEGs encoded outside of the triplicated region revealed enrichment of genes involved in Epithelial to Mesenchymal 10 Transition (EMT), IFN responses and other immune signatures (e.g. IL2/STAT5 signaling, IL6/JAK/STAT signaling), as well as key developmental pathways, such as Myogenesis, Notch

Signaling and Hedgehog Signaling (Fig. 3F, Data S5 ). This analysis also revealed downregulation of genes involved in Oxidative Phosphorylation and multiple gene sets associated with cell growth and proliferation (e.g., MYC targets, E2F targets, mTORC 15 signaling), among others ( Fig. 3F, Data S5) . These same pathways were similarly dysregulated in the hearts of Dp16 2xIFNRs mice, albeit to a lesser degree. In fact, comparison of Dp16 2xIFNRs versus Dp16 transcriptomes revealed significant normalization of the gene signatures dysregulated in Dp16 (Fig. 3F) . Examples of EMT genes elevated in Dp16 but normalized in Dp16 2xIFNRs include multiple collagen subunits (e.g., Col3a1), collagen modifying enzymes (e.g., 20 P3h1), matrix metalloproteinases (e.g., Mmp2), and the plasminogen activator (Plaur) (Fig. 3G , Data S5). Canonical ISGs elevated in Dp16 but not in Dp16 2xIFNRs include Isg20, Lgals3bp, Trim25, and Adar (Fig. 3H) . Hallmark genes involved in Oxidative Phosphorylation decreased only in Dp16 include subunits of the cytochrome C oxidase complex (e.g., Cox7b), the NADH:Ubiquinone Oxidoreductase complex (e.g., Ndufa1), the ATP synthase complex (e.g., Atp5c1), and the pyruvate dehydrogenase complex (e.g., Pdp1) (Fig. 3I) . Genes involved in cell growth and proliferation depleted only in Dp16 include enzymes involved in nucleic acid synthesis (e.g., Ctps), DNA replication factors (e.g., Rfc4, Rpa3), and mitochondrial ribosomal proteins (e.g., Mrpl23) (Fig. 3J) . 5 Altogether, these results indicate that triplication of Ifnrs elicits a signaling cascade in the developing heart involving elevated JAK/STAT signaling, dysregulation of EMT processes, along with decreased cell growth and proliferation, all of which could contribute to heart malformations.

Children with T21 and Dp16 neonates exhibit delays in achieving developmental milestones (3, 20) . Relative to WT controls, Dp16 neonates show reduced chance of success in achieving the surface righting reflex as well as ear twitch and auditory startle sensitivities on any given day, but no differences in eye opening (Fig. 4A) . When females and males are analyzed separately, 15 this phenotype is clearly sexually dimorphic, with Dp16 females showing the most pronounced differences ( Fig. S5A-B) . Notably, Dp16 2xIFNRs mice show rescue of all three Dp16 developmental delays (Fig. 4A, Fig. S5A-B) .

Cognitive deficits in adult mice were then evaluated using contextual fear conditioning (CFC) and Morris water maze (MWM) (25, 26) . During conditioning in the CFC test for associative 20 learning and memory, mice were presented with two mild foot shocks. Upon the second shock, Dp16 displayed a significantly decreased freezing response relative to WT controls (Fig. 4B) .

When reintroduced to the shock context on day 2, both WT and Dp16 froze at a higher baseline rate relative to the beginning of day 1, yet WT mice froze at significantly higher rate than Dp16.

Throughout the experiment, both on day 1 and day 2, Dp16 2xIFNRs mice displayed a significant rescue of all phenotypes measured by the CFC test (Fig. 4B) .

Upon examination of spatial learning and memory via MWM (Fig. S5C-D) (26), adult mice of all genotypes were equally capable of learning to escape the maze during the acquisition learning 5 phase, as evidenced by decreasing latency and total distance traveled to platform over successive swimming blocks (Fig. S5E) . However, Dp16 but not Dp16 2xIFNRs males swam significantly closer to the periphery when introduced to the maze (Fig. S5F) . Although this behavior is associated with hindrance of learning (27), such thigmotaxis in Dp16 males was moderate, and they still learned to escape the maze as quickly as the other genotypes during acquisition phase 10 ( Fig. S5E) . Immediately upon change of platform location in the reversal phase, both Dp16 and Dp16 2xIFNRs presented with deficits in memory extinction (Fig. S5E, block 7) . However, only Dp16 males exhibited impaired relearning of platform location (Fig. S5E , blocks 8-9). All cohorts still improved in performance over time in the reversal phase, as measured by decreased latency and total distance traveled to platform over successive blocks (Fig. S5E) . These subtle 15 yet significant differences by genotype in the reversal phase are in line with previous publications using MWM to study Dp16 deficits in memory extinction and relearning (16, (18) (19) (20) . Next, to pursue differences by genotype in allocentric memory, we evaluated swim path efficiency. This analysis again showed no difference in acquisition learning by genotype, revealing instead a significant deficit in memory extinction and relearning in Dp16, more 20 pronounced in males, with significant rescue of this phenotype in Dp16 2xIFNRs mice ( Fig. 4C-D) .

Furthermore, Dp16 but not Dp16 2xIFNRs demonstrated significantly reduced target quadrant occupancy during the reversal probe trial (Fig. 4E) . Lastly, impaired Dp16 motor coordination measured by the rotarod performance test (28) was not significantly rescued in Dp16 2xIFNRs (Fig.   S5G ).

Altogether, these results indicate that Ifnr gene dosage impacts multiple neurological hallmarks of DS, including key early developmental milestones, as well as major domains of cognitive function later in development, including associative learning and memory and spatial memory.

Normalization of Ifnr copy number reduces craniofacial anomalies.

Craniofacial abnormalities are a developmental feature of all individuals with DS, with strong inter-individual variation, including brachycephaly, maxillary deficiency, and smaller cranial base (29). Given that Dp16 mice reproduce aspects of the distinct craniofacial morphology of 10 individuals with T21 (17), we evaluated the impact of Ifnr gene dosage on Dp16 skull size and shape. Euclidean Distance Matrix Analysis (EDMA) was performed using 30 craniofacial and mandibular landmarks [LMs, (30) ] that were collected on 3D-rendered micro-computed tomography (µCT) scans ( Fig. 5A-B, Fig. S6A -B, Data S6). 58% (149/259) of all inter-LM distances differed between Dp16 and WT controls, consistent with prior morphometric studies in 15 Dp16 mice (17) (Fig. 5B, Fig. S6B ). Remarkably, 23% (34/149) of these differences were normalized in Dp16 2xIFNRs mice, including 79% (11/14) of Dp16 mandibular phenotypes (Fig.   5B, Fig. S6B , Data S6). The remaining inter-LM differences that persisted between controls and Dp16 2xIFNRs were less drastic than observed in Dp16 (Fig. S6C, Data S6) . Key examples include a shortening of the basisphenoid (BS) bone (LMs 24-27), increase in the intersphenofrontal 20 width (a proxy for inter-temple width; LMs 21-22), and alterations in many mandibular inter-LM distances (e.g., LMs 18-20), all of which are observed in Dp16 but attenuated in Dp16 2xIFNRs (Fig. 5C) . These morphometric analyses not only confirm previously reported craniofacial characteristics of Dp16, such as shorter widened skulls (17), but also demonstrate substantial effect of Ifnr gene dosage on craniofacial morphology.

Many of the differences in inter-LM distances observed in the skulls of Dp16 mice (31) involved LMs at the cranial base (i.e., LMs 23 to 30), which in turn were among the most sensitive to Ifnr gene dosage (Fig. 5D, Data S6) . Qualitative inspection of the cranial base revealed a loss of 5 intersphenoidal synchondrosis (ISS) in Dp16 mice, likely due to premature fusion of the presphenoid and BS bones (Fig. 5E ). This phenotype resembles the early spheno-occipital synchondrosis mineralization observed in individuals with DS (32), and its severity was significantly reduced in Dp16 2xIFNRs mice (Fig. 5F ). Early mineralization of the anterior cranial base restricts cranial base and midfacial outgrowth and is often associated with altered calvarium 10 shape (33). Notably, Dp16 mice display significantly shortened BS length and midface length (i.e., LMs 24-27 and 1-24, respectively), but these phenotypes are ameliorated in Dp16 2xIFNRs (Fig. S6B) . Furthermore, we observed a significant inverse correlation between basisphenoid length and midface length versus ISS fusion ( Fig. 5G-H) .

Altogether, these results indicate that triplication of Ifnrs contributes to major craniofacial 15 features distinctive of DS, further supporting a role for hyperactive IFN signaling in the dysregulated development of diverse organ systems in DS, including skeletal morphogenesis.

Despite significant research efforts, the mechanisms by which T21 causes the developmental and 20 clinical hallmarks of DS remain poorly understood (3) . In addition to the possibility that multiple genes contribute to a specific phenotype, the aneuploidy itself could exert effects independent of gene content (12). Clearly, elucidation of gene-phenotype relationships in DS would accelerate therapeutic strategies to ameliorate the ill effects of the trisomy. Within this framework, deciphering the mechanisms by which T21 causes lifelong IFN hyperactivity and dysregulation of downstream signaling pathways could enable immunomodulatory strategies to improve health outcomes in DS.

Using whole blood transcriptome analysis in a large cohort of individuals with DS, we observed 5 that overexpression of only a small subset of genes encoded on HSA21 correlates with gene signatures indicative of IFN hyperactivity and inflammation, including the four IFNRs. This exercise demonstrated that overexpression of individual triplicated genes can be tied to dysregulation of specific pathways in DS. Although hyperactive IFN signaling has been noted in cells with T21 since the 1970's, the contribution to key phenotypes of DS has not been defined 10 (34). In the Ts16 mouse strain carrying triplication of essentially all MMU16 genes through a centric-fusion translocation, including many genes not orthologous to HSA21, reduction of IFN signaling improved some aspects of Ts16 fetal development (34). However, because these mice die shortly after birth, examination of postnatal phenotypes was not feasible (34). We therefore employed genome editing to test the impact of normalizing Ifnr dosage in the Dp16 preclinical 15 mouse model of DS bearing a segmental duplication involving ~120 protein coding genes on MMU16 orthologous to HSA21 (16, 35) . This approach revealed that Ifnr triplication contributes to a dysregulated antiviral response, septal heart malformations, developmental delays, cognitive deficits, and craniofacial abnormalities in mice. These results expand and strengthen an increasing body of work documenting harmful effects of aberrant IFN signaling in human 20 development (13), while supporting the notion that DS can be understood in part as an interferonopathy (6, 14, 36) .

Our results demonstrating that Ifnr triplication underlies an exacerbated immune response may help explain the high rate of morbidity and mortality from respiratory infections observed in DS, as well as the increased rate of autoimmune disorders (4, 5, 37) . T21 is a top risk factor for severe COVID-19, leading to greatly increased rates of hospitalization and mortality (38, 39) . Our findings define a role for the Ifnrs during embryonic heart development, even in the absence of obvious immune triggers. In mice, numerous large regions orthologous to HSA21 were 15 identified that can contribute to increased rate of heart malformations, some of which include the Ifnr gene cluster ( Fig. S7) (3, 16, 23, 24, 56, 57) . Notably, single nucleotide polymorphisms in IFNGR2 and IL10RB have been associated with risk of CHD in DS (58). Nevertheless, our results are the first demonstration that normalization of Ifnr copy number is sufficient to rescue this phenotype. Furthermore, we show that Ifnr triplication drives increased JAK/STAT Competing interests: JME serves in the COVID19 Scientific Advisory Board for Eli Lilly.

Data and materials availability: All data to evaluate conclusions in this manuscript are present.

WGS and RNAseq data were deposited in public databases. The new Ifnr knockout mouse strain will be shared upon publication.

Figs. S1 to S7 Data S1 to S6

References 66-97. C, E, and G, each dot represents an independent biological replicate with mean indicated and 10 significance determined by Mann-Whitney test. In B-C and E-G, significance is indicated as *p≤0.05, **p≤0.01, ***p≤0.001, and ****p≤0.001. midface length (H) for each skull (n=20/group, with n=6-7 from each genotype). In C and G-H, significant when *p≤0.05, **p≤0.01, ***p≤0.001, and ****p≤0.001. RNA-seq data yield was ~33-103 × 10 6 raw reads and ~21-69 × 10 6 final mapped reads per Cytokine analysis was done of 249 study participants with T21 for which a matched whole blood transcriptome analysis was available.

Gene set enrichment analysis of clinical samples.

10 GSEA (71) was carried out in R using the fgsea package (v 1.14.0; RRID:SCR_020938) (72) in R, using Hallmark gene sets (73) and either log2-transformed fold-changes (for RNA-seq) or

Spearman rho values (for correlations) as the ranking metric.

Spearman Correlation Analysis.

As previously described (74) Sanger sequencing of mutant mice.

Potential founders (F0s) were genotyped by PCR to identify those that appeared to lack the entire Ifnr gene cluster without additional large chromosomal rearrangements such as inversions or 15 duplications on MMU16. These F0 mice were bred to wildtype (WT) C57BL/6N (Taconic) mice to generate heterozygous F1 progeny (herein WT 1xIFNRs ). PCR products spanning the deleted region were generated and subjected to Sanger sequencing using a 3730xl DNA Analyzer (ThermoFisher Scientific) (3) to identify transmission of a single modified allele to progeny.

Sequence-verified F1 mice with identical deletion events were then selected to maintain the line.

Whole genome sequencing and copy number variant analysis of mutant mouse.

Whole genome sequencing (WGS) was used to confirm clean chromosomal rearrangements of a proven F0 WT 1xIFNRs by copy number variant (CNV) analysis as previously described (77). between WT and the F0 male WT 1xIFNR . As we were not interested in discovering novel CNVs, rather confirming that a) our knockout was effective and b) that there were no major off-target effects, we iteratively increased the --minimum-windows-required parameter until no CNVs 20 were called between the WT mice of identical genetic background, leading to a parameter value of 14. Given a significance level of (p), 0.01 and a CNV detection threshold ratio (r) of 0.06, the theoretical minimum window size was determined as using the default method originally described (4). The window size used for the detection of CNVs was 1.5x (default) the theoretical minimum window size. Using these parameters, we confirmed the presence of our intended deletion with no other coherent CNVs of similar size. WGS data were deposited in NCBI SRA.

Animal Husbandry.

All animal experiments were approved by the Institutional Animal Care and Use Committee qRT-PCR to assess Ifnr transcript expression in brain hippocampi.

Mice 31-34 weeks of age were euthanized by CO2 asphyxiation and cervical dislocation then immediately perfused with 1x PBS using a Perfusion Two Automated Perfusion Instrument (Leica, Cat#39471005). Brains were removed from skulls then cut in half to divide the right and left side of the brain performing a sagittal section. The hippocampus was then manually dissected out of the right brain hemisphere and homogenized in Lysing Matrix D tubes (MP Biomedicals, Cat#6913500) containing 594 µL of lysis buffer RLT Plus (Qiagen, Cat#1048849) and 6 µL of 2-mercaptoethanol (Sigma, Cat#M3148) for 30 s using a Mini-Beadbeater-24 5 (BioSpec Products, Cat#112011) then frozen at -80°C. Upon quick freeze/thaw in a water bath at 37°C, total RNA was isolated using the AllPrep DNA/RNA/Protein Mini Kit (Qiagen, Cat#80004)). RNA was quantified using the Qubit 3 Fluorometer (Invitrogen) and cDNA generated from 100 ng of RNA using the Applied Biosystems High-Capacity cDNA Reverse

Transcription Kit (Cat#43-688-14). qRT-PCR was then performed as previously described (21) using the Applied Biosystems Viia7 384-well block real time PCR system. Briefly, qRT-PCR master mix was prepared with Applied Biosystems SYBR Select Master Mix for CFX. Standard curves were run for every primer pair in each qRT-PCR experiment to ensure efficient amplification of target transcripts within all experimental tissues. All samples were run in triplicate, averaged, and normalized to 18s rRNA. Primer sequences are provided in Data S4,

Tab D.

Spectral flow cytometry to assess IFNR surface expression on white blood cells.

Peripheral blood was collected from the submandibular vein of mice 17-25 weeks of age into tubes of lithium heparin (Sarstedt, Cat#41.1393.105) then stained as previously described with minor alterations (21). Briefly, 25 µL of fresh whole blood was pre-incubated with anti-mouse Spectral flow cytometry to assess phopho-STAT1 in white blood cells.

Peripheral blood was collected from the submandibular vein of mice 17-25 weeks of age into tubes of lithium heparin (Sarstedt, Cat#41.1393.105) then stained as previously described with the minor alterations described above for IFNR stains (21). Briefly, 25 µL of blood were subjected to RBC lysis then stimulated for 30 minutes at 37C with 10,000 units/mL of Enzyme-linked immunoassay to assess IFNAR2 protein in serum.

Peripheral blood was collected from the submandibular vein of mice 11-36 weeks of age into tubes containing serum gel with clotting activator (Sarstedt, Cat#41.1500.005). Serum was 20 isolated by spinning at 10,000 xg for 5 mins at RT, separated out from the original tube and serum gel, then frozen at -80°C. Upon first freeze/thaw on ice, serum was brought to RT then Male Dp16 were crossed overnight with 8-12-week-old synced female WT 1xIFNRs . Dams were checked daily for vaginal plugs; the first morning of visual confirmation was denoted embryonic day (E)0.5. Upon 4-chamber heart formation at E15.5 (80), mouse embryos were harvested from dam after CO2 asphyxiation and cervical dislocation then embryos were briefly allowed to 5 exsanguinate on ice in 1x PBS. A tail snip was collected from each embryo for genotyping.

Histology to detect heart malformations. To test for differences in the chance of success in achieving each developmental milestone, results were treated as time-to-event data and analyzed using a mixed effects Cox regression approach using the survival (version 3.2-7 (86)), coxme (version 2.2-16 (87) Mice of 4-5 months of age were handled two mins per day for 2-5 days leading up to the first test. When mice were run through multiple assays, they were run through rotarod, Morris water maze (MWM), then contextual fear conditioning (CFC). For all assays, mice were allowed to acclimate to the experimental room in their home cages for 30 mins before testing began each 20 day at 1200-1600 h. Equipment was cleaned in between mice.

The rotarod performance test was used to measure motor coordination as previously described for mouse models of DS (90). Briefly, mice were placed on a rotating cylinder at a set speed and had to keep pace with the rotation to stay on the cylinder (28). Each mouse was given 2 practice sessions at 16 revolutions per minute (RPM), followed by 2 test sessions at 16, 24, and 32 RPM.

Each session ended after 120 s or when the mouse fell. Inter-session intervals were 15 mins.

Latency to fall was tracked using Rotarod Version1.4.1 (Copyright 2002 -2010 and averaged across both test sessions for each speed.

MWM was performed on mice to assess spatial learning and memory as previously described for S5D ). Throughout both phases, more efficient navigation to the platform was interpreted as superior allocentric memory of the platform location (Fig. S5C) .

Swim data were collected using the video tracking system Ethovision (Noldus) v8.5. Nesting was applied to swim paths in Ethovision prior to analysis where the center point (mouse) must be between start and stop threshold velocities of 2 and 1.75 cm/s to avoid giving weight of initial 5 interaction of animal to arena and tracking of experimenters' hand during mouse drop into the maze.

The CFC test of associative learning and memory was performed as previously described for interpreted as fear in the testing phase.

Morphometric analysis skull form.

Mice were decapitated at 7-8 weeks of age then whole heads imaged using a SkyScan 1275 (µCT). Scanning was performed at 17.6-micron resolution using the following parameters: 55 kV, 180 mA, 0.5 mm Al filter; 0.3˚ rotation step over 180˚, and 3-frame averaging. All raw scan 20 data were reconstructed to multiplanar slice data using NRecon V1.7.4.6 software (Bruker, Belgium). Reconstructed data were then rendered in 3D with consistent thresholding parameters using Drishti V2.6.5 Volume Exploration software (Limaye, 2012) for gross visual assessment of the craniofacial skeleton. Representative rendered images were captured and processed using Photoshop (Adobe Creative Cloud).

Morphometric analysis of craniofacial landmarks (LMs) was then used to compare skull form (e.g. skull size and shape) between genotypes as previously described (30). Briefly, coordinates for 30 homologous LMs were independently collected from each 3D rendered skull within 5 Drishti V2.6.5 by two investigators blinded to genotype ( Fig. S6A and Data S6). LM datasets from two investigators were averaged and then normalized by their respective root centroid size (RCS) values to remove overall skull size as a variable and are reported in voxel units (Data S6).

The WinEDMA package was used to conduct Euclidean Distance Matrix Analysis (EDMA), which analyzes morphological differences between two groups of specimens by assessing the 10 change in ratio values between respective LM pairs (31). Following Lele and Richtsmeier (95), the 90% CIs were calculated by bootstrapping the shape difference matrix 10,000 times (Data S6). The FORM procedure was employed to find inter-LM distances that differed between populations, as well as LMs influencing or driving those differences (95). Inter-LM distances were deemed different between populations if the CIs did not cross 1.

Qualitative assessment of the skull.

Except for calvaria rounding, which was done on mice 7-10 weeks of age, all qualitative assessment of skulls from decapitated mice were done at 7-8 weeks of age by an expert in craniofacial morphology blind to genotypes. As mineralization of spheno-occipital synchondrosis (SOS) typically begins around postnatal D28 to ultimately mediate fusion of the 20 basisphenoid (BS) and basioccipital bones by 12 weeks of age (30), unusually pronounced ossification of the intersphenoid synchondrosis (ISS) was formally assessed. An ISS severity score was assigned using the following criteria: 0 = normal unfused appearance; 1 = <1/3 rd of the ISS width bridged by ossification; 2 = between 1/3 rd and 2/3 rd of the ISS bridged by ossification; compared to WT controls (n=6-7/group followed by bootstrapping 10,000x). Mean population estimates (circles) are on confidence intervals (CIs, lines) and colored according to differences 10 where green represents larger distance for the numerator (i.e. CIs>1), blue represents smaller (HSA21) that may contribute to risk of congenital heart defects (CHDs) in humans with trisomy 21 (T21, yellow), genes with functional evidence whose triplication is necessary or sufficient to 5 increase incidence of CHDs in mouse models of Down syndrome (DS, blue), and genes with supporting evidence in both humans and mouse models of DS (green). Relative cytogenetic locations and number of protein-coding genes (bold) are indicated along ideogram of the q arm for HSA21 colored according to Giemsa banding (3, 16, 23, 24, (56) (57) (58) .

Data S1. Demographics of human research cohort (separate file).

Tab A. RNA sequencing (RNA-seq) results of whole blood compared between people with T21 and typical D21 controls.

Tab B. Gene Set Enrichment Analysis (GSEA) of whole blood transcriptome after comparison 5 between people with T21 and typical D21 controls.

Tab C. GSEA of HSA21 transcripts from whole blood transcriptome after comparison between people with T21 and typical D21 controls. 

National population-based estimates for major birth defects

Down syndrome

Log (LN) of form difference ratio of all mean population estimates for inter-LM distances of the skull proper and mandible shown in panel (B) that were smaller (blue) or larger (green) in WT animals