A disorder-related variant (E420K) of a PP2A-regulatory subunit (PPP2R5D) causes constitutively active AKT-mTOR signaling and uncoordinated cell growth

Functional genomic approaches have facilitated the discovery of rare genetic disorders and improved efforts to decipher their underlying etiology. PPP2R5D-related disorder is an early childhood onset condition characterized by intellectual disability, hypotonia, autism-spectrum disorder, macrocephaly, and dysmorphic features. The disorder is caused by de novo single nucleotide changes in PPP2R5D, which generate heterozygous dominant missense variants. PPP2R5D is known to encode a B’-type (B’56δ) regulatory subunit of a PP2A-serine/threonine phosphatase. To help elucidate the molecular mechanisms altered in PPP2R5D-related disorder, we used a CRISPR-single-base editor to generate HEK-293 cells in which a single transition (c.1258G>A) was introduced into one allele, precisely recapitulating a clinically relevant E420K variant. Unbiased quantitative proteomic and phosphoproteomic analyses of endogenously expressed proteins revealed heterozygous-dominant changes in kinase/phosphatase signaling. These data combined with orthogonal validation studies revealed a previously unrecognized interaction of PPP2R5D with AKT in human cells, leading to constitutively active AKT-mTOR signaling, increased cell size, and uncoordinated cellular growth in E420K-variant cells. Rapamycin reduced cell size and dose-dependently reduced RPS6 phosphorylation in E420K-variant cells, suggesting that inhibition of mTOR1 can suppress both the observed RPS6 hyperphosphorylation and increased cell size. Together, our findings provide a deeper understanding of PPP2R5D and insight into how the E420K-variant alters signaling networks influenced by PPP2R5D. Our comprehensive approach, which combines precise genome editing, isobaric tandem mass tag labeling of peptides generated from endogenously expressed proteins, and concurrent liquid chromatography–mass spectrometry (LC-MS3), also provides a roadmap that can be used to rapidly explore the etiologies of additional genetic disorders.

Functional genomic approaches have facilitated the discovery of rare genetic disorders and improved efforts to decipher their underlying etiology. PPP2R5D-related disorder is an early childhood onset condition characterized by intellectual disability, hypotonia, autism-spectrum disorder, macrocephaly, and dysmorphic features. The disorder is caused by de novo single nucleotide changes in PPP2R5D, which generate heterozygous dominant missense variants. PPP2R5D is known to encode a B'-type (B'56δ) regulatory subunit of a PP2A-serine/threonine phosphatase. To help elucidate the molecular mechanisms altered in PPP2R5Drelated disorder, we used a CRISPR-single-base editor to generate HEK-293 cells in which a single transition (c.1258G>A) was introduced into one allele, precisely recapitulating a clinically relevant E420K variant. Unbiased quantitative proteomic and phosphoproteomic analyses of endogenously expressed proteins revealed heterozygousdominant changes in kinase/phosphatase signaling. These data combined with orthogonal validation studies revealed a previously unrecognized interaction of PPP2R5D with AKT in human cells, leading to constitutively active AKT-mTOR signaling, increased cell size, and uncoordinated cellular growth in E420K-variant cells. Rapamycin reduced cell size and dose-dependently reduced RPS6 phosphorylation in E420K-variant cells, suggesting that inhibition of mTOR1 can suppress both the observed RPS6 hyperphosphorylation and increased cell size. Together, our findings provide a deeper understanding of PPP2R5D and insight into how the E420K-variant alters signaling networks influenced by PPP2R5D. Our comprehensive approach, which combines precise genome editing, isobaric tandem mass tag labeling of peptides generated from endogenously expressed proteins, and concurrent liquid chromatography-mass spectrometry (LC-MS 3 ), also provides a roadmap that can be used to rapidly explore the etiologies of additional genetic disorders.
PPP2R5D-related developmental disorder (OMIM#616355) is a syndrome characterized by moderate-to-severe developmental delay, intellectual disability, seizures, macrocephaly, autism-spectrum-disorder (ASD), hypotonia, and delayed motor skill development (1)(2)(3)(4). Genomic analyses revealed that the disorder is due to a de novo single-nucleotide missense mutation in the PPP2R5D gene (1)(2)(3)5). To date, seven pathogenic variants of PPP2R5D have been reported ( Fig. S1A and Table S1). All introduce a variant PPP2R5D protein harboring a single amino acid change in a highly conserved region. The molecular mechanisms that are affected by pathogenic PPP2R5D variants remain to be elucidated.
PPP2R5D encodes a B'-type subunit (B'56δ) of the phosphoprotein phosphatase type 2A (PP2A) holoenzyme. To date, upward of 72 PP2A-holoenzymes has been described. Nearly all are heterotrimeric protein complexes contain one of 18 unique B-type regulatory/targeting proteins and a ubiquitously expressed core dimer (Fig. 1A). The shared dimer is composed of a catalytic subunit (C-subunit) that functions as a serine/ threonine protein phosphatase and a structural/scaffolding protein (A-subunit) that tethers the regulatory and catalytic subunits together. The B-subunits provide substrate specificity, affect intracellular localization, and regulate the catalytic activity of holoenzymes (6)(7)(8). In humans, two genes (PPP2CA and PPP2CB) encode nearly identical catalytic subunits and two genes (PPP2R1A and PPP2R1B) encode highly similar Asubunits. Each of the 18 known B-subunits is encoded by a distinct gene, with additional diversity generated by alternative splicing. After translation, the subunits do not simply coalesce with the core dimer to form a functional enzyme. Rather, holoenzyme biogenesis occurs through poorly characterized assembly and recycling mechanisms that pair a unique  B-subunit with the common core dimer through organized and regulated processes (7,9,10). Like many genetic developmental disorders, the etiology of PPP2R5D-related disorder is unknown. Human RNAsequencing and protein expression data indicate that PPP2R5D is expressed in most tissues, with modestly higher levels in the brain, breast, testis, and gastrointestinal tissue (11)(12)(13). All PP2A-holoenzymes act as serine/threonine protein phosphatases, and most evidence for the involvement of "PP2A" in a particular process has been derived from the use of small-molecule inhibitors including okadaic acid (14), fostriecin (15,16), and cantharidin (17,18), which target the catalytic subunit shared among all PP2A-holoenzymes (19,20), or siRNA, targeting a common core protein. Collectively, these studies have linked "PP2A" phosphatase activity to nearly all signaling networks controlled by reversible phosphorylation (21)(22)(23)(24). However, in a biological setting, individual holoenzymes are likely regulated independently, recognize limited specific substrates, and function in a limited number of intracellular processes (6)(7)(8). The challenge before the field is to decipher regulatory mechanisms and biological substrates of the >70 PP2A-holoenzymes.
To explore the biological role of PPP2R5D holoenzymes and decipher molecular mechanisms associated with a clinically relevant PPP2R5D variant, we used a CRISPR single base editor to generate HEK-293 cell lines in which a transition (c.1258G>A) introduced the E420K missense variant in PPP2R5D. The E420K variant was chosen for study because it is a reoccurring pathogenic variant (2), and the gene encoding PPP2R5D (PPP2R5D) contains an endogenous protospacer adjacent motif (PAM) ideally suited for single base editing of the target purine with BE4-Gam. Genomic editing allowed for cell-based studies of endogenously expressed proteins, and tandem mass tag (TMT) quantification and liquid chromatography coupled with mass spectrometry (LC-MS 3 ) enabled unbiased quantitative global proteomic and phosphoproteomic analyses of wild-type (WT), E420K-heterozygous, and E420K-homozygous cells. Our studies reveal heterozygous dominant changes in kinase/phosphatase signaling in E420K variant cells, providing a foundation to help decipher the etiology of PPP2R5D-related developmental disorder.

Generation of PPP2R5D E420K variant cell lines
To gain insight into how variants in PPP2R5D alter biological processes at a molecular level, we used BE4-Gam, a fourth-generation genomic base editor, to create stable HEK-293 cell lines containing a pathogenic E420K variant (Fig. 1B). To recapitulate the missense variant, we introduced a single base-pair change (c.1258G>A) in only one of the two alleles (Fig. 1C). Base editing was facilitated by a wellpositioned endogenous PAM in PPP2R5D and the development of gRNA that positioned the cytidine deaminase precisely over the C:G target (Figs. 1D and S1B) while simultaneously positioning the nCas9 to nick the DNA strand containing the unedited guanine. Nicking the unedited DNA strand favors repair of the edited DNA using the uracilcontaining strand as a template by endogenous mismatch repair mechanisms (25). The fusion of inhibitors of base excision repair (BER) helps prevent removal of the uracil generated by the deamination of cytidine, which increases efficiency of correct base editing (26). Cells were single-cell sorted 48 h after transfection to allow time for replication, after which the base edit becomes a permanent change in the genome.
By Sanger sequencing, we confirmed the genotype of cell lines containing the desired E420K heterozygous (HET) variant (Fig. 1E). We also developed cells in which the E420K variant was introduced into both alleles (HOMO). To ensure the cell lines were homogeneous, each clonal-derived line was single-cell sorted three times, prior to further analysis (Fig. S1C). Complete genomic sequencing of the parental, E420K HET, and E420K HOMO variant cell lines was performed to characterize the genomic background, detect any off-target editing, and ensure that the repetitive single-cell sorting did not introduce spontaneous somatic mutations. Cell lines with off-target or spontaneous mutations in proteincoding regions were discarded.

The PPP2R5D E420K variant is expressed
Having established that the cell lines were homogeneous and contained the desired zygosity of the PPP2R5D c.1258G>A variant allele, we examined the expression and assembly of the endogenously expressed variant protein into a holoenzyme. Immunoprecipitates (IPs) of endogenous PPP2R5D were Coomassie-blue stained (Fig. 1F) and analyzed by LC-MS/MS revealing peptides containing Glu 420 or Lys 420 in the HET extracts ( Fig. 1, G and H). As expected, only the Glu 420 peptide was identified in extracts generated from WT, and only the Lys 420 peptide was identified in HOMO extracts. Detection of C-terminal peptides indicates that the E420K mutation does not compromise the production of full-length PPP2R5D protein. Orthogonal validation of the LC-MS/MS data was obtained through western analysis, which revealed similar levels of PPP2R5D in both total cell extracts and IPs generated from all three cell lines (Fig. 1I). Proteins that participate in PP2A biogenesis, including LCMT1, PME1, alpha4, TIPRL, and PTPA, were similarly expressed in all three cell lines (Table S2). Western and LC-MS/MS analysis revealed the presence of both catalytic (Fig. 1, I-K) and PPP2CA. J, relative amounts of PPP2CA from WT or E420K cells normalized to total PPP2R5D in IPs and represented as the mean percent of WT (N = 5 independent experiments; mean ± SD, ANOVA Kruskal-Wallis p = 0.003 and H = 9.41, Dunn's multiple comparison posthoc test, *p < 0.05 and NS: nonsignificant). K, LC-MS/MS detection of PPP2R1A, PPP2R1B, PPP2CA, and PPP2CB in PPP2R5D IPs of WT and E420K-variant cells. L, relative phosphatase activity of endogenous PPP2R5D IPs (n = 10). Endogenous PPP2R5D was immunoprecipitated from WT, HET, or HOMO cells and hydrolysis of DiFMUP was measured at Ex 360 nm and at Em 460 nm. Activity was normalized to total levels of PPP2R5D detected in IPs through immunoblotting shown in (K). WT activity was normalized at 30 min (endpoint) to 1 and variant activity is expressed relative to WT. N = 5 independent biological replicate of cells, each performed as n = 2 technical replicates.
Altered signaling in PPP2R5D-variants scaffolding (Fig. 1K) subunits in PPP2R5D IPs of WT and E420K variant cells. This observation indicates that the E420K variant does not prevent the biogenesis of holoenzymes containing PPP2R5D. Our studies also revealed that both WT and E420K variant PPP2R5D assemble with either isoform of the catalytic (PPP2CA/PPP2CB) or scaffold (PPP2R1A/PPP2R1B) subunits, suggesting that PPP2R5D can be incorporated into four distinct holoenzymes. It should be noted that reduced levels of PPP2CA/B and PPP2R1A/B were found in PPP2R5D IPs of E420K variant cells, suggesting the partial loss of PPP2R5D-holoenzyme formation or decreased stability.
We next tested the E420K variant containing holoenzymes for phosphatase activity using an established fluorescent (DiFMUP) assay (27)(28)(29)(30). In this assay, PPP2R5D was immunoprecipitated using an antibody that recognizes a near Cterminal epitope that is unique to PPP2R5D, and the phosphatase activity of the associated PP2A-holoenzyme was measured in the IP (Fig. 1L). To adjust for minor technical variation inherent in IP-based assays, the activity associated with WT and variant immunoprecipitated PPP2R5Dphosphatase complexes was normalized to total levels of PPP2R5D detected in the IPs by immunoblotting analyses. Currently, this approach represents the only viable method to measure the activity of endogenously expressed PPP2R5Dharboring holoenzymes without contaminating activity from one of the other B'-family members that share considerable homology (PPP2R5A, PPP2R5B, PPP2R5C, and PPP2R5E) ( Fig. S2A and Table S3). We observed the dephosphorylation of DiFMUP in all samples, with less activity in the IPs generated from E420K variant cells, which is consistent with observed reduced levels of immunoprecipitated catalytic subunit. When the data was normalized to the amount of catalytic subunit in the IP, the activity was similar in WT and E420K variants (Fig. S2B), suggesting that the specific activity of the variant holoenzymes was not reduced. From these studies, we concluded that the E420K variant is expressed and translated. Future studies will be needed to determine if the reduced catalytic activity in the E420K variants reflects partial assembly, altered recycling, or decreased stability of PP2A-variant holoenzymes.
Quantitative changes in the proteome and phosphoproteome in PPP2R5D E420K variant cells A major hurdle to determining how the E420K mutation alters the biological functions of PPP2R5D is that only a few studies have explored the "normal" action(s) of PPP2R5D at a cellular level (31,32). Therefore, we chose a global unbiased approach, employing TMT labeling for relative and absolute quantitation with LC-MS 3 , to obtain proteomic ( Fig. 2A) and phosphoproteomic (Fig. 2B) data sets comparing WT and E420K variant cells. Overall, we reproducibly quantified 26,731 phosphopeptides on 5926 proteins in the phosphoproteomic analysis and 9178 proteins in the proteomic analysis (Table S2 and Fig. S2C). Most proteins did not change in abundance (Fig. 2, C and D). Only 2.6% of all proteins quantified were significantly increased or decreased by twofold or more in E420K HET compared with WT cells. Notably, large changes in the abundance of PP2A catalytic, scaffolding, regulatory, or other PPP-subunits were not detected. However, approximately twofold changes in some PP1R-subunits were observed (Table S4). In contrast, 8.8% of phosphopeptides were significantly increased or decreased by twofold or more in E420K HET compared with WT cells (Table S2) suggesting that the E420K variant alters global protein phosphorylation, which is consistent with previous reports indicating that PPP2R5D acts as a regulator of PP2A-holoenzyme function (6,13,(33)(34)(35)(36). To help distinguish changes in phosphopeptide abundance resulting from changes in protein levels versus changes in protein phosphorylation, we further filtered the data set, correcting for changes in protein abundances. This reduced our data set to 24,307 reproducibly identified and quantified phosphopeptides (Fig. 2, E and F), of which 7% were significantly increased or decreased in E420K HET compared with WT cells. Compared with WT, E420K HET cells had more phosphopeptides increasing phosphorylation occupancy than phosphopeptides decreasing phosphorylation occupancy (1.9-fold). Correlation analyses of the E420K HOMO and HET proteomics and phosphoproteomics data sets revealed a strong positive correlation for both HET and HOMO samples (Fig. 2, G and H) indicating that the data sets were highly similar. This observation is consistent with the heterozygous variant acting as a dominant condition in humans.

Enrichment of R-x-R-x-x-S/T-B phosphopeptide motif in PPP2R5D E420K variant cells
To identify signaling cascades altered in the variant cell lines, we investigated the possibility that specific kinases might differentially oppose PPP2R5D WT and E420K-variant cells. Overrepresentation analysis identifying consensus sequences either enriched or deselected, revealing an enrichment of an R-x-R-x-x-S/T-B motif (Fig. 3A) where x represents any amino acid and B represents amino acids with hydrophobic side chains. This motif is a protein kinase B (AKT/PKB) consensus sequence, which is shared by PKCα, PKCß, and RSK1 (37). We also observed a strong deselection of proline-directed motifs in phosphorylation sites differentially regulated in the E420K variant compared with WT cell lines (Fig. 3A). These observations are consistent with both the phosphorylation site consensus motif preferences and the phosphorylation site consensus motif disfavors we observed for the PPP2R5holoenzymes families (31). Performing the enrichment analysis independently for phosphorylation sites that were significantly changed revealed that the enrichment of the AKT consensus motif was due to sites increased in phosphorylation (Fig. S3A). The hyperphosphorylation of AKT, PKC, and RSK1 substrates has been previously reported following treatment with okadaic acid (38), which is often reported as a specific inhibitor of PPP2CA but actually also inhibits PPP4C, PPP5C, and PPP6C with similar potency (19,20,(39)(40)(41).
Bioinformatics tools (KinomeXplorer; PhosphoSitePlus; RegPhos) were used to predict kinases responsible for the observed changes in phosphorylation and to identify Altered signaling in PPP2R5D-variants Figure 2. Quantitative changes in the proteome and phosphoproteome in PPP2R5D E420K-variant cells. Workflow for quantitative proteomics (A) and phosphoproteomics (B). Volcano plots of proteomic (C, D) and protein-corrected phosphoproteomic (E, F) data (N = 3 independent biological replicates for WT, N = 4 independent biological replicates for E420K HET and HOMO). Volcano plots depict log 2 ratios of peptides, plotted against the negative log 10 of the p value of their fold change. Peptides above -log 10 = 1.3 (corresponding to p < 0.05) are considered statistically significant. Peptides shown in green or blue are twofold or more decreased or increased in abundance, respectively. Correlation analysis of proteomic (G) and phosphoproteomic (H) between E420K HET and HOMO datasets. kinase-substrate regulatory networks, revealing enrichment of AKT, serum and glucocorticoid-regulated kinase (SGK), and cAMP-dependent protein kinase (PKA) substrates among the phosphorylation sites increased in E420K-variant cells (Fig. 3B  and Table S5). Indeed, most known and predicted AKT, SGK1, and PKA phosphorylation sites showed increased phosphorylation in variant cell lines ( Fig. 3C and Table S5). Together, these data are consistent with the E420K variant having a dominant loss of suppression effect on AKT-mediated signaling cascades.
For interrogation of the phenotype ontology database, we filtered the number of phosphopeptides in the proteincorrected data set to those that were significantly increased or decreased by twofold in the variant versus WT group. We used the Human Phenotype Ontology database to identify enriched phenotypes (Fig. 3D). Although these experiments were conducted in HEK-293 cells, a diverse array of phenotypes were enriched in our analysis, including many observed in children diagnosed with PPP2R5D-variant disorder (4). Due to the prevalence of neurodevelopmental disorders (NDD) in PPP2R5D-variant affected children, we also investigated which genes corresponding to proteins with significantly altered phosphorylation sites are associated explicitly with NDD. Overenrichment analysis identified 262 unique entrez gene IDs that were annotated to the selected functional categories in common with the reference list (Table S6). Of the top 20 identified genes, 11 were genes associated with NDD curated by four sources: SFARI gene score 1/2; ASC FDR <0.1 DDG2P (DDD curated) genes; DDD FDR <0.025 (42)(43)(44) (Table S7). There are a total of 627 genes in the curated NDD gene list. Given a total of >19,000 protein-coding genes, 11/20 overlap with a 627-gene list gave us a p-value of 1e-11, with an odds ratio of 36. This is a conservative estimate, given how many NDD genes are putative targets of PPP2R5D. These data are consistent with clinical presentations in PPP2R5D patients including neurodevelopmental delay and macrocephaly, a phenotype associated with increased AKT activity.

Metabolic AKT-mTOR signaling is altered by the PPP2R5D E420K variant
Studies with inhibitors of the common PP2A catalytic subunit have linked "PP2A activity" (representing the combined activity of >70 PP2A-holoenzymes) to the regulation of virtually every signaling cascade influenced by serine/threonine kinases. To gain insight into processes specifically altered by the PPP2R5D E420K variant, the same filtered data set used for phenotype ontology resulted in the identification of 2457 phosphorylation sites on 1382 proteins, of which 763 and 1694 were significantly decreased or increased, respectively. This data set revealed several differences in metabolic signaling pathways between WT and E420K-variant cells ( Fig. 3E and Table S6). Specifically, we identified hyperphosphorylated proteins in insulin, mTOR, AMPK, FoxO, and ErbB signaling, along with downstream effectors such as 40S ribosomal protein S6 (RPS6), eukaryotic translation initiation factor 4E-binding protein 1 (4EBP1), and eukaryotic translation initiation factor 4B (EIF4B) that control cell size and protein synthesis (Figs. 3F and 4A; Fig. S3, B-E and Table S6). We identified three hyperphosphorylated sites (Thr 1135 , Ser 1174 , Ser 1177 ) on the rapamycin-insensitive companion of mTOR (RICTOR), which is part of the mTOR2 complex that phosphorylates AKT at Ser 473/4/2 (Fig. 4B). AKT (isoform 2 based on phosphopeptide sequence) was hyperphosphorylated at several sites (Ser 34 , Ser 302 , Ser 474 ) (Fig. 4B). P-Ser 473 (Ser 474 on isoform 2) increases the kinase activity of AKT (45). Downstream of AKT, the negative regulator of mTOR1, AKT1S1/PRAS40, was hyperphosphorylated at Thr 246 . P-Thr 246 is known to relieve AKT1S1/PRAS40 inhibitory functions toward mTOR1 (46), resulting in mTOR1 activation. Another negative regulator of mTOR1, tuberin (TSC2), was hyperphosphorylated at Ser 939 and Ser 960 . AKT phosphorylates TSC2 at Ser 939 to relieve TSC2-directed inhibition of mTOR1 (47). Downstream effectors of mTOR1, such as 40S ribosomal protein S6-Kinase (S6K) (Ser 416 ), were also hyperphosphorylated. RPS6 can be phosphorylated by S6K and was found to be hyperphosphorylated at multiple sites (Ser 236 , Ser 240 , Thr 241 , and Ser 242 ) in the E420K-variant cell lines. Another well-established substrate of AKT, glycogen synthase kinase 3α (GSK3α) (48), was hyperphosphorylated at Ser 21 in E420K-variant cells suggesting inhibited GSK3α activity. Together these phosphoproteomic data indicate that PPP2R5D E420K variants act as dominant constitutive activators of AKT-mTOR signaling.

Orthogonal validation of the LC-MS 3 data
Quantitative LC-MS 3 revealed 3.1-fold hyperphosphorylation of AKT2 on Ser 474 in E420K HET compared with WT cells. The epitope surrounding Ser 474 is conserved in AKT1, AKT2, and AKT3 (Ser 473 , Ser 474 , and Ser 472 , respectively), and a phospho-specific antibody recognizes all three phosphorylated AKT isoforms. Consistent with the LC-MS 3 data, western analysis revealed increased phosphorylation in the E420K HET cells (Fig. 5, A and B). Ser 474 phosphorylation activates AKT activity, resulting in GSK3α (Ser 21 ) and GSK3β (Ser 9 ) phosphorylation and inhibition of GSK3 activity (48). The epitopes surrounding Ser 21 and Ser 9 are identical and recognized by a single phospho-specific antibody. Western analysis revealed increased levels of P-GSK3α/ß (Ser 21/9 ) in E420K-variant cell extracts (Fig. 5, C and D), consistent with our phosphoproteomic analysis showing a 3.0-fold increase in P-GSK3α in E420K HET cells compared with WT. LC-MS 3 analysis of variant cells revealed that RPS6 phosphorylation on Ser 235 , Ser 236 , and doubly Ser 235/236 was elevated by 2.2-fold, 5.3-fold, and 3.1-fold, respectively in E420K HET compared with WT extracts. A triple phosphorylation of RPS6 with further phosphorylation on Ser 240 or Thr 241 was also increased in the E420K HET, but not the E420K HOMO variants (Table S8). Western analysis assessing doubly phosphorylated RPS6 at Ser 235/236 revealed increased phosphorylation in E420K HET extracts (Fig. 5, E and F). Western analysis of other known AKT targets also revealed increased phosphorylation in variant cells, including P-FOXO3 (Ser 253 ) (Fig. 5, G and H) and P-ß-catenin (Ser 552 ) (Fig. 5, I and J). Increased ßcatenin protein levels ( Fig. 5K and Table S2) were associated Altered signaling in PPP2R5D-variants with decreased P-ß-catenin (Ser 33/37 ) (Fig. S4, A and B). Similar results were observed in the E420K homozygous cell lines (Fig. 5, A-K). When considered in their entirety, these observations are most consistent with dominant aberrant AKT-mTOR signaling induced by the PPP2R5D E420K variant.  For western and LC-MS 3 analyses, cell extracts were generated from nearly confluent monolayers and yielded consistent differences in RPS6 phosphorylation (Ser 235/236 ). However, phosphorylation of RPS6 can be influenced by cell density (49) and growth factors (50). Therefore, changes in RPS6 phosphorylation required more rigorous analysis. We observed elevated phosphorylation of RPS6 (Ser 235/236 ) in WT cells growing at low density, which decreased with increasing cell density (Fig. 5, L and M and Fig. S4C). In contrast, E420K HET cells growing at low density displayed elevated baseline phosphorylation of RPS6 that did not change with increasing cell density (Fig. 5, L and M and Fig. S4C). Serum deprivation is known to abolish RPS6 Ser 235/236 phosphorylation and was observed in WT cells, but not in HET cells (Fig. 5, N and O). The addition of epidermal growth factor (EGF), insulin-like growth factor 1 (IGF1), or a combination of both under serum-free conditions increased P-RPS6 above the level observed in the serum-supplemented high-density WT cells. In contrast, baseline P-RPS6 was elevated to a submaximal level in serum-deprived HET cells, and subsequent treatment with either EGF, IGF, or a combination of both did not elicit an increase in P-RPS6 above basal levels. Additionally, LC-MS 3 analyses revealed 1.3-fold decrease in IGF1 receptor (IGF1R) protein levels in HET cells compared with WT cells, which may reflect feedback inhibition of IGF1 signaling due to the constitutive activation of AKT in this pathway (Table S2).
Rapamycin inhibits aberrant mTOR signaling and decreases cell size in PPP2R5D E420K variant cells mTOR signaling regulates cell size through phosphorylationdependent activation of RPS6 (51). A comparison of the relative cell size, determined by measuring mean forward scattering height (FSC-H) (51), revealed that E420K HET cells were significantly larger, evidenced by a rightward shift in FSC-H and a mean increase in FSC-H (Fig. 6, A and B). Treatment with rapamycin is known to decrease cell size by inhibiting mTOR1 signaling (51). We observed that rapamycin decreased FSC-H in both WT cells and E420K HET (Fig. 6, C-F). After 48 h, rapamycin-treated E420K and WT cells were similar in size, both smaller than respective solvent-treated controls (Fig. 6, G and H). Rapamycin also reduced P-RPS6 in both WT and HET cells in a dose-dependent manner (Fig. 6, I and J) suggesting that both the observed RPS6 hyperphosphorylation and increased cell size can be suppressed through the inhibition of mTOR1. In contrast, the elevated P-AKT and P-GSK3, which act upstream of the target of rapamycin, mTOR1, were not suppressed by rapamycin in the variant cells (Fig. 6I). This suggested that the PPP2R5D E420K variant affects signaling upstream of mTOR1.

PPP2R5D-PPP2CA forms a complex with AKT
Our data are consistent with a PPP2R5D-holoenzyme directly opposing AKT, with the E420K variant cells having reduced the ability to inactivate AKT due to the failure to dephosphorylate Ser 473/4/2 . This data was further validated by IP-western analysis, in which both endogenous PPP2CA and PPP2R5D co-IP with AKT and both PPP2R5D and AKT co-IP with endogenous PPP2CA in WT and E420K cell extracts (Fig. 6, K and L and Fig. S5, A-C). Together, the data supports the existence of a protein complex containing AKT and a PPP2R5D-containing PP2A holoenzyme, with the E420K variant having a dominant loss of suppression effect on AKTmediated signaling cascades (Fig. 7A).

Discussion
Collectively PP2A-phosphatases represent up to 0.2% (52) of total cellular protein, and studies with catalytic inhibitors (i.e., okadaic acid, fostriecin, or cantharidin) or siRNA (targeting subunits of the common core dimer protein) indicate that, as a family, PP2A phosphatases are important for the vast majority of cellular processes (53)(54)(55)(56)(57)(58)(59)(60). However, little is known about the biological actions of individual holoenzymes, and nothing has been reported related to how pathogenic PPP2R5D variants alter function. Here, employing single base genomic editing, we created human cell lines that recapitulate a clinically relevant pathogenic variant. The advantage of genomic editing is that the expression, translation, and biogenesis of the E420K variant holoenzyme are controlled by endogenous mechanisms. This is important because PP2A-holoenzyme biogenesis requires the coordinated actions of many proteins (i.e., IGBP1/alpha4, PPP2R4/PTPA; LCMT1; PPPME1, and TIPRL) that are also needed for the biogenesis of the other 10 to 72 PP2A holoenzymes found in most cells (6,7). Biogenesis starts with the synthesis of an inactive catalytic subunit (Csubunit) that undergoes a number of conformational changes and modifications before it gains catalytic activity and is incorporated into the common A/C core dimer. Functional PP2A-holoenzymes are created when a unique B-subunit is incorporated into the common core dimer (6,7). Our studies revealed that both the WT and the E420K variant are expressed at similar levels. Our findings further our understanding of WT PPP2R5D and E420K-variant function, revealing its assembly with both isoforms of the catalytic (PPP2CA, PPP2CB) and scaffolding (PPP2R1A, PPP2R1B) subunits. This suggests that the E420K variant does not prevent biogenesis. However, when PPP2R5D is immunoprecipitated, IPs of E420K variant cells contained reduced levels of the catalytic subunit (PPP2CA/PPP2CB). This suggests that there may be a pool of unbound PPP2R5D in the cells. Alternatively, the variant containing holoenzymes may be less stable or some aspect of the holoenzyme biogenesis/recycling process may be altered, which warrants further investigation in future studies of PPP2R5D-related disorder.
B-subunits regulate holoenzyme catalytic activity, substrate specificity, and intracellular localization (6, 7). The c.1258G>A transition in PPP2R5D introduces a missense mutation, which replaces a negatively charged glutamic acid with a positively charged lysine (E420K) in a region predicted to regulate substrate specificity or holoenzyme catalytic activity (61) (Fig. S1A). We used unbiased functional approaches revealing heterozygous dominant changes in kinase/phosphatase signaling in E420K-variant cells (Fig. 4). Our data reveal Altered signaling in PPP2R5D-variants  (Fig. S2). L, representative immunoblot and (M) associated graph showing levels of P-S6 (Ser 235/236 ) normalized to total ß-actin. Cells were assayed in 10% serum containing media at decreasing cell densities. "+" sign indicates 100% confluent cells incubated with fresh media increases in phosphopeptide occupancy in both HET and HOMO variant cell lines. For most peptides there was a strikingly similar increase in phosphopeptide occupancy in both heterozygous and homozygous variant cell lines as compared with WT control (Fig. 2, G and H). This indicates that the heterozygous variant introduces dominant changes in protein phosphorylation, which is consistent with the heterozygous dominant actions in patients. However, we did note instances in which the heterozygous variant had a more pronounced effect when compared with control or the homozygous variant (e.g., RPS6; Fig. 5E), which is not entirely consistent with a dominant action of the heterozygous variant. One possibility for further studies is that the homozygous variant triggers a feedback mechanism, which due to the presence of some WT enzyme in the heterozygous cells does not reach a critical threshold needed to trigger feedback inhibition.
Our studies also reveal a role for PPP2R5D and its clinically relevant E420K variant in the regulation of AKT signaling (Fig. 7). We found that WT PPP2R5D forms a complex with AKT and PPP2AC, an observation consistent with reported co-IPs in murine adipose tissue (62) and studies in which okadaic acid or siRNA targeting the common A-subunit or Csubunit resulted in a "PP2A-dependent" dephosphorylation of AKT at Thr 308 and Ser 473/4/2 and reduced AKT activity. In one of the few studies that have explored the B-subunit specificity associated with AKT inactivation in dopaminergic signaling, ßarrestin specifically recruited B55α (PPP2R2A) holoenzymes to the AKT-signaling complex, resulting in the dephosphorylation of Thr 308 but not Ser 473/4/2 (63). Kuo et al. (64) have reported that the overexpression of B55α subunit preferentially impairs phosphorylation at Thr 308 in FL5.12 and NIH3T3 cells. It has also been shown that during AKT inactivation in insulin signaling, Cdc2-like kinase 2 phosphorylation of B56ß (PPP2R5B) regulates the assembly of a PPP2R5B-PP2A holoenzyme complex with AKT, leading to dephosphorylation of both AKT Ser 473 and Thr 308 (65). Finally, we have observed a strong deselection of proline-directed motifs in peptide phosphorylation that increased in the E420K variant (Figs. 3A and S3A). This is intriguing because PP2A holoenzymes containing B55-family regulatory subunits have been shown to preferentially dephosphorylate proline-directed motifs (31,66). Together, these observations suggest that PP2A holoenzymes with unique B-subunits, like PPP2R5D, may selectively target specific phosphorylation sites on the same protein contained in a signaling complex to provide cells the ability to coordinate their growth signals vis-à-vis their energetic state and nutrient availability. As our studies reveal for the PPP2R5D E420K variant, aberrations in B-subunit fine-tuning of site-directed phosphatase activity can have serious ramifications on cellular growth and development. Therefore, future studies exploring differential kinase opposition by PP2A-holoenzymes should examine regulatory B-subunit composition of the complex.
Consistent with increased AKT-activation, our studies revealed increased phosphorylation of established AKT substrates (GSK3, ß-catenin, and FOXO3) and appropriate changes in downstream signaling proteins (AKT1S1, S6K, S6, EIF4B, ULK1, and CLIP1) in the E420K-variant cell lines. P-FOXO3 (Ser 253 ) suppresses its ability to regulate transcription of receptor tyrosine kinases (RTKs) (67), with the inhibition of AKT inducing the expression of HER3, IR, and IGF1R. Our proteomics data show a decrease in IGF1R by 1.3-fold in HET cells (Table S2), which is consistent with the associated increase in P-AKT. Consistent with reduced IGF1R, the variant cells did not display increased P-RPS6 following IGF1 treatment as observed in WT cells. However, prolonged S6K activity has been reported to lead to feedback inhibition of IRS1 signaling (68). The variant cells have constitutively elevated P-S6, but it is premature to conclude that the attenuated IGF1-S6 phosphorylation in the variant cells is due to FOXO3 suppression of IGF1R expression. Similarly, due to complex feedback mechanisms, future studies are needed to determine the contribution of feedback inhibition in the E420K variant cells.
Cells expressing the E420K variant showed increased P-ßcatenin (Ser 552 ) and increased ß-catenin protein levels, in association with decreased P-ß-catenin at Ser 33/37 (Figs. 5, I-K and S4, A and B). Previous reports in HEK-293 cells indicate that increased GSK3 activity results in decreased ß-catenin accumulation due to GSK-mediated phosphorylation of ß-catenin at Ser 33/37 , triggering phosphorylation-mediated degradation (69). The suppression of PPP2CA has been reported to increase ß-catenin levels, due to the suppression of GSK3 phosphorylation of ß-catenin at Ser 33/37 and the resulting decrease in ß-catenin degradation (69). Coexpression of B56 and GSK3ß has been shown to result in a substantial loss of ß-catenin levels (69), and overexpression of PPP2R5D also decreases ß-catenin levels (70). These studies are all consistent with our findings showing that the E420K variant is associated with increased AKT-activity (Ser 473/4/2 phosphorylation) leading to increased inhibitory P-GSK3 and, in turn, to the reduction of ß-catenin Ser 33/37 -phosphorylation-dependent degradation. Of note, another study has reported that suppression of a different B-subunit (PPP2R2A) can increase P-ß-catenin (Ser 33/37 ) (71). Treatment with okadaic acid has led to many conflicting observations that can be resolved if ßcatenin Ser 33/37 is a direct substrate for B55α-holoenzymes, and B56-holoenzymes act upstream to regulate AKT and GSK3. These studies, including ours, demonstrate that distinct roles of unique B-regulatory-containing holoenzymes need to be further defined in order to fully understand the physiological complexity of cell signaling systems.   Phenotype ontology analysis of our data sets identified top genes associated with neurodevelopmental delay, with phenotypic overlap in neurobehavior observed in individuals with PPP2R5D pathogenic variants. Albeit, most were not associated with macrocephaly, a characteristic phenotype with variable penetrance in PPP2R5D-related neurodevelopmental disorder. While we are unable to draw definitive conclusions about the etiology of PPP2R5D clinical phenotypes based solely on observations made in cultured cells, our study provides evidence of convergence of common pathways for which genetic mutations have been observed in association with other neurodevelopment-disorders (72)(73)(74). Further, our cellbased data is consistent with the recent addition of PPP2R5D to the list of autism susceptibility genes (42) related to AKT3, PIK3CA, PTEN, TSC1/2, and mTOR (72)(73)(74). A key finding in our study is the demonstration of aberrant activation of the AKT-mTOR pathway, which is known to be also dysregulated in ASD (75). Increased AKT/mTOR activity is consistent with deficiencies of FMR1, TSC1/2, or PTEN found in Fragile X, tuberous sclerosis, and Cowden syndrome. Higher activity of mTOR and S6K, but lower activity of GSK3α and TSC2 have been reported in T cells from children with ASD (75). Our finding that rapamycin suppresses aberrant hyperphosphorylation of RPS6 and decreases the size of E420K variant cells beyond untreated WT should also be noted. This suggests that the development of future therapies aimed at specifically targeting PPP2R5D-dependent effects on AKT/ mTOR signaling in PPP2R5D-related neurodevelopmental disorder should be mindful of potential salutary effects. The suppression of increased AKT/mTOR activity has been demonstrated to improve ASD-associated symptoms in mice deficient for PTEN and TSC1 (76,77). If PPP2R5D-regulated signaling is conserved in mice, future studies to define E420K signaling alterations in rodent models of PPP2R5Drelated neurodevelopmental disorder should further our understanding of this life-long and debilitating condition and aid the development of targeted therapeutic approaches.
BE4-Gam (Addgene plasmid #100806; http://n2t.net/ addgene:100806) was provided as a generous gift from Dr David Liu. Plasmids, amplified in Escherichia coli DH5α, were isolated using a Qiagen plasmid DNA purification kit (12123) according to the protocol of the manufacturer. All plasmids were sequenced to confirm the fidelity of the constructs. WT HEK-293 cells were seeded on 9.5 cm 2 plates in antibiotic-free medium and transfected at 70% confluency using XtremeGENE9 DNA Transfection Reagent (06366511011) according to the protocol of the manufacturer. Synthetic guideRNA (crRNA:tracrRNA duplex; Invitrogen A35512) was generated as two independent oligonucleotides that were annealed. The guideRNA (targeting sequence: GCACGCUCUGCCACCUGGUG) was transfected 24 h postplasmid delivery using Lipofectamine RNAiMAX (Invitrogen 13778030). GuideRNA efficiency was first assessed using a Surveyor Mutation Detection kit following the protocols provided by the manufacturer (IDT 706020). The validated guideRNA was used in transfection, and cells were clonally isolated by plating single cells into 96-well plates using BD FACSAria II. After clonal expansion, genomic DNA from clones was isolated and PCR was performed using Q5 High-Fidelity DNA Polymerase (NEB) and oligonucleotide primers pairs that target regions flanking Exon 12 (AGAATTTTCATCCC CATGCCCTC: CAGAGATCTTGCACACAGGCATC, IDT). Sanger sequencing using ABI3730XL was employed to detect single-base mutations. Prior to further use, cell lines with the desired mutations were single-cell sorted three times to ensure each cell line represented a homogeneous population.

Immunoprecipitations and phosphatase activity assay
Cells were plated at 1.5 × 10 6 in 150 mm dishes and allowed to grow for 48 h. Cell lysis was achieved by the addition of icecold Triton X-100 Lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% Triton X-100, 5 mM EDTA), containing Protease inhibitor Cocktail (ThermoScientific). Cells were mechanically sheared by gentle passage through a 28G syringe needle and centrifuged at 14,000g for 15 min. The supernatant was transferred to a new tube, and protein concentrations were determined using RC DC Protein Assay (Bio-Rad 5000121). TrueBlot Anti-Rabbit Ig IP Agarose Beads (Rockland Altered signaling in PPP2R5D-variants Immunochemicals) were used to immunoprecipitate endogenous AKT (Cell Signaling 9272) as previously described (62), and Dynabeads Protein A (100001D) was used to immunoprecipitate endogenous PPP2CA (Millipore clone 1D6 05-421) and PPP2R5D (ab188323) according to protocols provided by the manufacturer. Immunoprecipitated samples were eluted into near-boiling 2× SDS-PAGE sample buffer and analyzed on a SDS-PAGE as described above. For phosphatase assay, immunoprecipitated samples were eluted into 50 mM Tris HCl, 150 mM NaCl, pH 7.4. Phosphatase activity assay was performed as described previously (78). Briefly, 20 μl of IPs was dispensed in 1.33× assay buffer (0.15 M NaCl, 30 mM HEPES pH 7.0, 1 mM DTT, 0.1 mg/ml BSA, 1 mM ascorbate, 1 mM MnCl 2 ) with a final assay concentration of 75 μM 6,8-difluoro-4-methylumbelliferyl phosphate (DiFMUP) (Invitrogen D6567). 6,8-difluro-4methylumbelliferone (DiFMU) was measured every minute for 30 min with a BioTek Synergy (Ex 360 nm, Em 460 nm). Aliquots of the IPs were eluted with near-boiling 2× SDS-PAGE sample buffer and were used for western analysis to detect total levels of PPP2R5D, which was used to normalize activity measured.
Trypsin digest of PPP2R5D peptides PPP2R5D IPs were digested overnight with sequencing grade trypsin (0.2 μg/μl) (Promega) at 37 C and subjected to centrifugation (10,000g for 5 min). The supernatant was transferred into a MS sample vial for analysis. LC-MS/MS analysis was performed with a Thermo Q-Exactive Plus mass spectrometer (Thermo Fisher Scientific). The mobile phases for the HPLC consisted of 3% ACN and 0.2% formic acid in water (solvent A) and 3% water and 0.2% formic acid in acetonitrile (solvent B). The peptide mixture was loaded onto a C18 reverse-phase column (ES800) equipped with a guard column and chromatographically separated using a liner gradient of 2%-30% solvent B over a period of 40 min at a flow rate of 300 nl/min. The peptides were subjected to electrospray ionization with positive polarity using the Thermo EASY-Spray source, with a voltage of 1.5 kV and a capillary temperature of 300 C. Samples were analyzed in a datadependent manner, with one MS1 scan from 400 to 2000 m/ z at a resolution of 70,000, followed by the top six multiply charged ions per MS1 scan selected for MS2 scans at 17,500 resolution using HCD fragmentation. Charge states of 1, 5-8, and greater than 8 as well as those that were unassigned were excluded. The resulting files were searched using Mascot against the RefNCBInr_human protein entry databases (gi| 5453954|ref|NP_006236.1| serine/threonine-protein phosphatase 2A) or Custom database (gi|5453954E420K|ref| NP_006236.1E420K| serine/threonine-protein phosphatase 2A) with homo sapiens taxonomy.

Western analysis
Cells (2 × 10 5 ) were plated in 60 mm dishes. For orthogonal validation of MS 3 , cells were lysed at confluent monolayers.

Rapamycin treatment
Rapamycin (MCE, HY-102019) was prepared in DMSO. 2 × 10 5 cells were added to 60 mm dishes, allowed to grow for 24 h, and then treated with the indicated concentrations of rapamycin. In total, 10 nM concentrations were used in all Altered signaling in PPP2R5D-variants FSC-H analyses. 48 h after treatment, cells were either treated with trypsin and resuspended in media to be analyzed in FACS or lysed near-boiling 2× SDS-PAGE sample buffer for western blotting analysis, as described above. FACS analysis was achieved using BD FACSAria II and Software FACSDiva Version 6.1.3. Relative FSC-H was analyzed using FlowJo.

Quantitative proteomics and phosphoproteomics
Cell pellets from biological replicates were resuspended in of ice-cold lysis buffer [8 M urea, 25 mM Tris-HCl pH 8.6, 150 mM NaCl, containing phosphatase inhibitors (2.5 mM beta-glycerophosphate, 1 mM sodium fluoride, 1 mM sodium orthovanadate, 1 mM sodium molybdate) and protease inhibitors (1 mini-Complete EDTA-free tablet per 10 ml lysis buffer; Roche Life Sciences), and the cells were then lysed by sonication (three times for 15 s each with intermittent cooling on ice). Lysates were subjected to centrifugation (15,000g for 30 min at 4 C). Supernatants were transferred to a new tube, and the protein concentration was determined using a BCA assay (Pierce/ThermoFisher Scientific), DTT was then added to the lysates to a final concentration of 5 mM to reduce disulfide bond, and DTT-lysate was incubated for 30 min at 55 C. Lysates were then cooled to room temperate, and 15 mM iodoacetamide (room temperature) was added to achieve alkylation. After a 45-min room temperature incubation in the dark, alkylation was quenched by the addition of 500 mM DTT (1:100 dilution). After sixfold dilution with 25 mM Tris-HCl pH 8, the samples were incubated overnight at 37 C with 1:100 (w/w) trypsin. The next day, the trypsin digest was stopped by the addition of 0.25% TFA (final v/v). Precipitated lipids were removed by centrifugation (3500g for 15 min at room temperature), and the peptides in the supernatant were desalted over an Oasis HLB 60 mg plate (Waters). An aliquot containing 200 μg of peptides was removed for proteomics analysis. The remaining peptide containing supernatant was lyophilized and stored at -80 C until further use (e.g., phosphopeptide enrichment).
Phosphopeptide enrichment was achieved using a Fe-NTA phosphopeptide enrichment kit (Thermo Fisher) according to instructions provided by the manufacture and desalted over an Oasis HLB 60 mg plate (Waters). Phosphopeptides were resuspended in 133 mM HEPES (SIGMA) pH 8.5, and TMT reagent (ThermoFisher Scientific) stored in 100% acetonitrile (ACN) (Burdick & Jackson) was added, vortexed to mix reagent and peptides. After 1 h at room temperature, an aliquot was withdrawn to check for labeling efficiency while the remaining reaction was stored at -80 C. Once labeling efficiency was confirmed to be at least 95%, each reaction was quenched by the addition of ammonium bicarbonate to a final concentration of 50 mM for 10 min, mixed, diluted with 0.1% TFA in water, and desalted. The desalted multiplex was dried by vacuum centrifugation and separated by offline Pentafluorophenyl (PFP)-based reversed-phase HPLC fractionation performed as previously described (79).
For proteomics analysis, peptides were labeled with TMT reagent, label checked, mixed, and off-line separated as described above. Samples were analyzed on the Orbitrap Lumos operating in data-dependent, SPS-MS3 quantification mode wherein an Orbitrap MS1 scan was taken (scan range = 350-1400 m/z, R = 120K, AGC target = 2.5e5, max ion injection time = 50 ms), followed by data-dependent ion trap MS2 scans on the most abundant precursors for 1.8 s. Ion selection; charge state = 2: minimum intensity 2e5, precursor selection range 600-1400 m/z; charge state 3-5: minimum intensity 4e5). Quadrupole isolation = 0.8 m/z, CID collision energy = 35%, CID activation time = 10 ms, activation Q = 0.25, scan range mode = m/z normal, ion trap scan rate = rapid, AGC target = 4e3, max ion injection time = 40 ms). Orbitrap MS3 scans for quantification (R = 50K, AGC target = 5e4, max ion injection time = 100 ms, HCD collision energy = 65%, scan range = 100-500 m/z, synchronous precursors selected = 10). The raw data files were searched using COMET with a static mass of 229.162932 Da on peptide N-termini and Altered signaling in PPP2R5D-variants lysines and 57.02146 Da on cysteines, and a variable mass of 15.99491 Da on methionines against the target-decoy version of the human proteome sequence database (UniProt; downloaded 2/2013, 40,482 entries of forward and reverse protein sequences), maximum of three missed cleavages allowed, precursor ion mass tolerance 1 Da, fragment ion mass tolerance ±8 ppm, and filtered to a <1% FDR at the peptide level. Quantification and data analysis were carried out as described above.

Bioinformatics analysis
Icelogos were generated using singly phosphorylated sites with a phosphorylation localization score of 0.95 or higher that significantly increased or decreased by at least twofold after protein correction in both heterozygous and homozygous variant cells (84). Kinase enrichment analysis using RegPhos was performed within Webgestalt (85) using singly phosphorylated sites with a phosphorylation localization score of 0.95 or higher that significantly increased or decreased by at least twofold after protein correction. Kinase-substrate relationships were investigated using the PhosphoSitePlus database (86). Upstream kinase prediction was carried out using NetworKIN (87). Gene Ontology (GO), and KEGG pathway analysis was performed on proteins with phosphorylation sites with significantly increased or decreased by at least twofold after protein correction in Webgestalt and visualized in Cytoscape (88). Protein-protein interactions were identified using the STRING database (89). Off-target mutations were detected using a pipeline derived from methods for detecting de novo mutations (90). Specifically, we used bwa-mem (91) to map the whole genome sequencing data to human reference genome version hg38, called single nucleotide variants and small insertion and deletions by GATK (92), and annotated the candidate variants by VEP (93). We define candidate off-target mutations as variants carried by the mutant but not WT cells. To exclude likely technical artifacts, we discard candidate mutations meeting any following exclusion criteria: the mutant allele is supported by fewer than six reads; allele fraction (# of reads supporting the mutant allele/# of reads covering the site) <0.3; the mutant allele is supported by reads from only one of the sequencing strands; and the genomic site is located in segmental duplication regions with score >0.95.

Statistical analysis
GraphPad Prism version 8.2.1 was used for all statistical analyses. Image J software was used for quantification of band density in western blotting analyses. N is defined as independent biological replicates of the same clonal cell line. Passage number of the cells used in this study at the time of transfection were p3. WT control and variant cells at the time of identification of desired genetic edits were p12. All experiments shown in this study were performed at passages 12-28 (different N's were generated between different passages to ensure no difference between cell passages were observed).
All statistical analyses are reported with standard deviation (SD). Statistical analyses were performed as unpaired t tests for analysis between two sample groups or Kruskal-Wallis with Dunn's multiple comparison for multiple sample groups as indicated in figure legends. Numerical normality tests were performed to determine nonparametric statistical analysis when no assumption about the distribution of the data was possible due to small sample sizes (N ≤ 6).

Data availability
Raw MS data for this study are available at MassIVE (MSV000084960) and PRIDE accession (PXD017562). All other data needed to evaluate the conclusions in the paper are present in the paper or the supporting information. Funding and additional information-The research was supported by the Jordan's Guardians Angels (JGA) and National Institutes of Health/National Institute of General Medical Sciences, United States, R35GM119455 to A. N. K.