Scalable generation and functional classification of genetic variants in inborn errors of immunity to accelerate clinical diagnosis and treatment

Zachary H Walsh; Chris J Frangieh; Neeharika Kothapalli; Jay Levy; Clarissa K Heck; Johannes C Melms; Ron S Gejman; Parin Shah; Jared M Pollard; Akul Naik; Sarah L Grauman; Lei Haley Huang; Ashley Lee; Dusan Bogunovic; Joshua D Milner; Benjamin Izar

doi:10.1016/j.cell.2025.05.037

. Author manuscript; available in PMC: 2025 Oct 12.

Published in final edited form as: Cell. 2025 Jun 20;188(18):4861–4879.e27. doi: 10.1016/j.cell.2025.05.037

Scalable generation and functional classification of genetic variants in inborn errors of immunity to accelerate clinical diagnosis and treatment

Zachary H Walsh ^1,^2,^3,⁴, Chris J Frangieh ^1,^2,^3,⁴, Neeharika Kothapalli ^1,^2,^3,⁴, Jay Levy ⁵, Clarissa K Heck ^1,^2,^3,⁴, Johannes C Melms ^1,^2,^3,⁴, Ron S Gejman ^1,^2,^3,⁴, Parin Shah ^2,^3,⁴, Jared M Pollard ¹, Akul Naik ¹, Sarah L Grauman ¹, Lei Haley Huang ⁵, Ashley Lee ⁵, Dusan Bogunovic ⁵, Joshua D Milner ^5,^*, Benjamin Izar ^1,^2,^3,^4,^6,^7,^*

PMCID: PMC12514753 NIHMSID: NIHMS2110714 PMID: 40543502

SUMMARY

Next-generation sequencing is pivotal for diagnosing inborn errors of immunity (IEI) but predominantly yields variants of uncertain significance (VUS), creating clinical ambiguity. Activated PI3Kδ syndrome (APDS) is caused by gain-of-function (GOF) variants in PIK3CD or PIK3R1, which encode the PI3Kδ heterodimer. We performed massively parallel base editing of PIK3CD/PIK3R1 in human T cells and mapped thousands of variants to a clinically important readout (phospho-AKT/S6), nominating >100 VUS and unannotated variants for functional classification and validating 27 hits. Leniolisib, an FDA-approved PI3Kδ inhibitor, rescued aberrant signaling and dysfunction in GOF-harboring T cells and revealed partially drug-resistant PIK3R1 hotspots that responded to novel combination therapies of leniolisib with mTORC1/2 inhibition. We confirmed these findings in T cells from APDS patients spanning the functional spectrum discovered in the screen. Integrating our screens with population-level genomic studies revealed that APDS may be more prevalent than previously estimated. This work exemplifies a broadly applicable framework for removing ambiguity from sequencing in IEI.

In brief

In lieu of traditional genetic variant testing approaches, an approach using scalable variant classification in primary human T cells with a clinically relevant readout can inform rapid diagnosis and treatment of inborn errors of immunity.

Graphical Abstract

graphic file with name nihms-2110714-f0008.jpg

INTRODUCTION

Clinical diagnostics via next-generation sequencing (NGS) have become instrumental in the discovery of disease-associated genes, the identification of functional variants that cause disease, and especially in the case of inborn errors of immunity (IEI), the development of precision therapies targeting the disrupted pathway.^1–6 However, most variants detected are variants of uncertain significance (VUS).⁷ These variants cannot be definitively classified as pathogenic or benign due to the lack of clinical associations and/or functional data. Resolution of VUS frequently requires correlative data from primary patient cells, the inefficient process of identifying additional carriers of the same variants and their clinical phenotype, and cumbersome testing of individual putative gain- or loss-of-function (GOF/LOF) variants in non-physiologic cellular contexts. This process, which tends to only occur within resource-constrained academic medical systems, delays diagnosis and access to precision treatment.

Activated PI3Kδ syndrome (APDS) is an IEI that is due to variants in PIK3CD or PIK3R1, which encode the catalytic (p110δ) and regulatory (p85) subunits of the phosphoinositide 3-kinase δ (PI3Kδ) heterodimer, respectively.^8–12 These variants result in excessive PI3K/AKT signaling in lymphocytes, provoking lymphoid hyperplasia within tissues and lymphomas, and patients are also enriched for autoimmunity, atopic disease, susceptibility to herpesvirus infection, and humoral immunodeficiency associated with bronchiectasis.^13–17 T cells from APDS patients exhibit a cellular exhaustion phenotype, reduced naive T cells and receptor diversity, NK killing defects, and an increased propensity toward T follicular helper cell differentiation, and TFH1/TFH2 phenotypes.^{12,14,18–23} An understanding of the true prevalence of APDS (currently estimated at 1–2 cases per million in the United States)²⁴ and full phenotypic spectrum of many IEI, including APDS, is likely incomplete due to ascertainment and referral bias in identifying and reporting cases, and to the general propensity to dismiss VUS, which are not as rare and/or present in unaffected family members and “control” populations. Notably, recessive PIK3CD LOF is associated with variably penetrant combined immune deficiency and inflammatory disease,^25,26 while heterozygous, dominant-negative variants in PIK3R1 lead to short stature, hyperextensible joints, ocular depression with anterior segment dysgenesis (Rieger anomaly), and tooth eruption delay (SHORT syndrome).^27–30 There is a subset of patients who have both the developmental abnormalities seen in SHORT syndrome and the immune dysregulation of APDS. These patients almost exclusively have variants in the inter-SH2 (iSH2) domain of PIK3R1, which result in the unique combination of diminished PI3Kα activity—the driver of the developmental abnormalities—and reduced negative regulation of PI3Kδ—which drives immune dysregulation.^28,31

An APDS diagnosis is made when a patient presents with known associated signs and symptoms and one of few previously described pathogenic variants in PIK3CD or PIK3R1.³² If a VUS is found, research-based assays—usually employing flow cytometry to measure phosphorylation of AKT (Ser473) or S6 (Ser235/Ser236) after antigen receptor stimulation—are often used, as is overexpression of plasmid-encoded mutant vs. wild-type alleles in a cell model.^33,34 This level of evidence (clinical symptoms, pathogenic variant, and elevated PI3Kδ pathway output) is necessary to enable prescription of leniolisib, an FDA-approved, selective PI3Kδ-inhibitor that improves symptoms and immune cell functions in patients with APDS.^34–38 However, comprehensive characterization of all functional variants in PIK3CD or PIK3R1, coupled with broader genotyping in large patient cohorts with clinical metadata, is necessary to better determine the phenotypic disease spectrum and realize the full potential of precision management strategies through rapid diagnosis of patients.

Here, we leveraged CRISPR-dependent adenine base editing,³⁹ which induces site-specific deamination of adenine to generate A > G and T > C base transitions at endogenous loci, targeted by single-guide RNAs (sgRNAs),^39,40 to systematically tile variants across PIK3CD and PIK3R1 in primary human T cells. By mapping the effect of these variants on AKT/S6 phosphorylation, we functionally classified >100 VUS and novel variants as either GOF or LOF, deeply characterized their impact on T cell function and leniolisib response and found an unexpectedly high prevalence of carriers in population databases. Beyond changing our understanding of APDS, this work provides a scalable framework to systematically remove diagnostic ambiguity across IEI and to identify patients who may benefit from existing precision therapies.

RESULTS

CRISPR base editor screening coupled with a clinically relevant assay in T cells

To identify variants in PIK3CD or PIK3R1 that modulate PI3Kδ signaling, we first established the dynamic range of AKT and S6 phosphorylation in T cells measured by flow cytometry, a pathway-specific assay for diagnosing APDS.³⁴ Using our established adenine base editing (ABE) workflow,⁴¹ we used single-guide RNAs (sgRNA) to generate previously described PIK3CD GOF (p.C416R)¹³ and LOF (p.S281P)⁴¹ variants, or a non-targeting control sgRNA (NT-sgRNA) in T cells from multiple donors. Following a 20-min stimulation of these cells with cross-linked soluble CD3 and CD28 antibodies (STAR Methods), we observed variable degrees of induction of AKT phosphorylation (pS473) and S6 phosphorylation (pS235/pS236), which correlated with T cell genotype (Figure S1A). PIK3CD GOF T cells had a 12-fold greater ratio of pAKT/pS6-high to negative cells when compared with NT controls and a 136-fold greater ratio compared with PIK3CD LOF T cells (Figure S1B). This high dynamic range enables discovery and quantitative evaluation of signaling output of PIK3CD/PIK3R1 GOF and LOF.

We designed an ABE sgRNA screening library to tile variants across the full coding regions of PIK3CD and PIK3R1, including 20 bases into each intronic and untranslated region. The library was designed for compatibility with the PAM-relaxed (NG-PAM) highly efficient base editor NG-ABE8e⁴² to maximize editing activity and accessible loci.³⁹ The library included sgRNAs generating every possible ABE-mediated variant across both genes and could generate perturbations in 74% of all residues of PIK3CD and 69% of all residues in PIK3R1 (Figure S1C; Table S1; STAR Methods).⁷ We also included 100 sgRNAs targeting “pan-essential” genes⁴³ with a strong negative effect on T cell growth⁴¹ (Table S1).

We performed pooled base editing by delivering mRNA encoding NG-ABE8e to three healthy T cell donors expressing the sgRNA library (Figures S1D–S1F). Following recovery, cells were stimulated, stained for pAKT and pS6 (Figure 1A; STAR Methods), and sorted into two PI3Kδ activity bins: the top 15% of pAKT/pS6-high and bottom 15% of pAKT/pS6-negative cells (Figure 1B). We also kept unsorted edited T cells in culture and re-expanded them over 2 weeks to assess variant effects on proliferation (Figure 1A).

Figure 1. — (A) Schematic for adenine base editing (ABE) screening of *PIK3CD* and *PIK3R1* in primary T cells.

(B) FACS sort criteria for pAKT/pS6 screens. Stimulated T cells were binned into the top and bottom 15% of pAKT (pS473) and pS6 (pS235/pS236) expression. (C and D) (C) Lollipop plots for PIK3CD and (D) PIK3R1 annotated with protein domains and regions from UniProt showing log₂(Fold Change) (LFC) of base editing sgRNAs in the pAKT/pS6 high vs. negative screen. LFCs across N = 3 healthy human T cell donors calculated with MAGeCK. Each point represents an sgRNA and is mapped to the sgRNA-targeted locus. Circles represent missense variants, while “X”s represent splice-site perturbations (SA, splice acceptor; SD, splice donor). Significantly enriched or depleted sgRNAs with a two-sided p < 0.05 in MAGeCK test are shaded red and known pathogenic variants are annotated with stars. Select amino acid conversions are annotated for highly enriched and depleted sgRNAs.

Related to Figures S1, S2, and S3.

To evaluate base-editing efficiency and determine screen rigor, we filtered sgRNAs with high predicted off-target activity (STAR Methods) and first examined results from the proliferation arm of the screen. We observed depletion of all 100 sgRNAs perturbing pan-essential genes (Figure S2A; Table S2), while negative control sgRNAs had no meaningful impact on proliferation (Figure S2A). Next, we examined results from the pAKT/pS6 screen, comparing sgRNA abundance in pAKT/pS6-high vs. negative cells, which demonstrated a positive inter-donor correlation of variant effects in PIK3CD (Figure S2B) and PIK3R1 (Figure S2C).

We next analyzed pAKT/pS6 screen data across all donors and investigated whether our screens recovered known pathogenic variants associated with elevated PI3Kδ activity and APDS. The top-scoring sgRNAs in both PIK3CD and PIK3R1 precisely generated ClinVar Pathogenic variants (i.e., PIK3CD p.C416R and PIK3R1 p.L573P)^{9,10,13,14,32,44} (Figures 1C, 1D, S2D, and S2E). Multiple other known pathogenic APDS-associated variants were also significantly enriched, including PIK3CD p.E1025G and disruption of the PIK3R1 exon 11 splicedonor site^45,46 (Figures 1C, 1D, S2D, and S2E). We found strong and consistent enrichment of ClinVar Pathogenic variants (10/11 sgRNAs significant, all as GOF; MAGeCK two-sided p < 0.05) and no enrichment of ClinVar Benign/Likely Benign variants (0/33 sgRNAs scoring as significant) (Figures S2F and S2G).

Our design follows American College of Medical Genetics and Genomics (ACMG) guidelines for functional assays and PS3/BS3 classification,^47,48 and the incorporation of these ClinVar Pathogenic and Benign/Likely Benign controls also enabled us to calculate an odds of pathogenicity (OddsPath)⁴⁷ (STAR Methods). Our guides were able to produce matches for 4 distinct Pathogenic variants and 26 distinct Benign or Likely Benign variants. All 30 variants were correctly scored with the defined two-sided p < 0.05 cutoff, resulting in an OddsPath score of 26.0 for Pathogenic and 0.250 for Benign/Likely Benign variants, which are associated with “PS3” and “BS3-supporting” evidence, respectively (Table S3). These results suggest that our approach can provide substantial functional evidence to previously unannotated variants and VUS. Thus, variant curation algorithms, including those generated via the ClinGen Variant Curation Expert Panel (VCEP), can incorporate the evidence provided here to determine variant classification as encountered clinically.

To further increase confidence in our screen results, we assessed whether disruptive variants (splice-site disruptions or missense conversions to proline^41,49) were depleted in the pAKT/pS6 screen. We found that many such variants in PIK3CD were strongly depleted, and the negative effect of both splice-site disruptions and proline conversions in PIK3CD was highly significant when compared with negative control sgRNAs (Figure S2H). Importantly, those variant types in PIK3R1 were not similarly depleted, likely owing to the many examples of loss of p85α function resulting in increased p110δ activity.^11,31,45 Together, these results indicate that performed screens were rigorous, reproducible, and yielded variants known to cause APDS.

Functional classification of novel GOF/LOF variants is improved with a pathway-specific signaling assay

We next investigated the landscape of all PIK3CD and PIK3R1 variants associated with elevated or decreased PI3Kδ signaling in the pAKT/pS6 screen (Figures 1C and 1D). In PIK3CD, we identified sgRNAs generating significantly enriched (i.e., putative GOF) variants, which spanned all protein domains, including in the adaptor-binding domain (ABD) (e.g., p.Q79R, p.Q80R, p.Q83G, and p.Q93R), the Ras-binding domain (RBD) (e.g., p.S199G and p.C219R), the C2 domain (e.g., p.N334G, p.C416R, and p.S444P), the helical domain (e.g., p.Y524C, p.E525G, p.E527G, p.D529G, and p.S578G), and the catalytic domain (e.g., p.H940R, p.H976R, p.S992P, and p.E1025G) (Figure 1C). Similarly, we identified significantly depleted (i.e., putative LOF) variants spanning all domains (Figure 1C). In PIK3R1, we identified significantly enriched variants in the SH2 domain 1 (e.g., p.Y408C), the inter-SH2 domain (e.g., p.K567G, p.L570P, p.L573P, and p.R585A), and the SH2 domain 2 (e.g., p.E683G, p.Y688H, and p.S689G), as well as depleted variants (Figure 1D). Nearly all listed variants are currently either unannotated or VUS in ClinVar⁷ (Table S2). We also identified several PIK3R1 splice variants with significantly increased pAKT/pS6, including variants disrupting the exon 11 splicedonor (SD) site (a common pathogenic PIK3R1 variant) and the exon 14 splice acceptor (SA) site (Figure 1D; Table S2).

Since PI3Kδ signaling also drives T cell proliferation,^22,50 we investigated whether variant effects from the proliferation screen could identify a similar spectrum of PIK3CD and PIK3R1 variants. We performed analogous analyses for the proliferation screen and confirmed reproducibility across human donors (Figures S3A and S3B). Although we observed a significant positive correlation between variant effects in the proliferation and pAKT/pS6 screens (Figure S3C), we found that the signal-to-noise ratio overall and for many specific variants was magnitudes of effect sizes (i.e., LFCs) greater in the pAKT/pS6 screen (Figures S3D and S3E; Table S2). Furthermore, the proliferation screen failed to recover some known pathogenic GOF variants, including PIK3CD p.E1025G (0/2 sgRNAs significant in proliferation screen; 2/2 significant in pAKT/S6 screen), and had fewer sgRNAs scoring significantly for others, such as PIK3CD p.C416R (2/4 sgRNAs significant in proliferation screen; 4/4 significant in pAKT/pS6 screen) and PIK3R1 p.L573P (1/2 sgRNAs significant in proliferation screen; 2/2 significant in pAKT/pS6 screen) (Figure S3F; Table S2).

Additionally, when we compared the LFC distribution of missense and splice-site disrupting sgRNAs with the null distribution of empty-window sgRNAs in each screen, we found that 586/2,320 (25.3%) of sgRNAs fell beyond two standard deviations of the mean of the null distribution for pAKT/pS6 screen (i.e., a z score > 2 or < −2) compared with only 371/2,329 (15.9%) in the proliferation screen (Figure S3G; Table S2). These data indicate that screening variant effects using proliferation alone is not sufficient to resolve a substantial fraction of variant effects and emphasizes the need for gene/pathway-specific readouts.

Variant mapping to PI3Kδ structure reveals spatial distribution and hotspots

Two-dimensional visualization of variant effects can identify variants with close association in primary structure; however, this does not capture variant co-localization in the protein’s native folded state, or variant interactions occurring between protein subunits, and does not provide structural biochemical insight into variant effects. To address this, we superimposed variant effects from our pAKT/pS6 screen onto the solved crystal structure of the PI3Kδ protein complex,⁵¹ which consists of the bound p110δ (encoded by PIK3CD) and p85 (encoded by PIK3R1) proteins (Figure 2A). This visualization approach enabled several conclusions, which were not evident from the two-dimensional data. For example, many high-scoring putative GOF and LOF were closely associated in three-dimensional space (e.g., PIK3CD GOF p.C416R with putative LOF p.Y395C and p.L394P and PIK3CD putative GOF p.H976R, p.C991R, and p.S992P with putative LOF p.L580P), suggesting that in some regions of PI3Kδ, variant effects are not consistent and depend on the precise amino acid and conversion. Additionally, collections of putative GOF, which appeared to be distant in protein primary structure (e.g., PIK3CD p.S578G with p.H976R and PIK3CD p. F938P with p.E1025G) or occurring on different subunits altogether (e.g., PIK3CD p.C416R with PIK3R1 p.K567G, p.L570P, and p.L573P), co-localized in the protein complex (Figure 2A). Finally, both putative GOF and LOF variants were broadly distributed throughout the PI3Kδ structure, spanning internal and surface residues and p110δ/p85 interaction interfaces, indicating that there are multiple mechanisms by which variants can affect PI3Kδ signaling.

Figure 2. — (A) Solved structure of the PI3Kδ protein complex. *PIK3CD* protein product p100δ is shaded yellow and *PIK3R1* protein product p85 is shaded gray. Green and red shading represents sgRNAs with highly positive or negative LFC in the pAKT/pS6 screen, respectively, mapped to the targeted residue. LFC is integrated across N = 3 human donors. Shading darkness represents magnitude of LFC. Select enriched and depleted variants are annotated. An sgRNA FDR cutoff of <0.8 was used for visualization.

(B) Heatmap of variant effects from pAKT/pS6 screens in each of N = 3 human donors. Variants shown were selected for downstream arrayed validation experiments. Variants are color coded as ClinVar Pathogenic (orange), ClinVar VUS/conflicting interpretation of pathogenicity (blue), or previously unannotated (black). For sgRNAs which are predicted to generate more than one variant (indicated by multiple variants separated by an underscore), each variant is color coded individually.

(C) Representative chromatograms from next-generation sequencing of T cells base edited with indicated variants. Chromatogram trace of the targeted base is shaded red and percent base editing is annotated.

(D and E) (D) Representative flow cytometry histograms of pAKT (pS473) and pS6 (pS235/pS236) in T cells base edited with known or putative gain-of-function (GOF) and (E) loss-of-function (LOF) variants discovered in the pAKT/pS6 screen, after 20-min CD3/CD28 stimulation in the presence of DMSO vehicle control. Histograms are representative of replicates from full experiment in Figures 3A–3D. Known GOF and LOF variants are shaded red and newly classified variants are shaded blue. MFI, median fluorescence intensity; FMO, fluorescence-minus-one control; and NT-sgRNA, non-targeting control sgRNA.

Related to Figures S3 and S4.

Given the broad distribution of impactful variants across PI3Kδ structure, we examined the association between variant effects in our pAKT/pS6 screens and computationally predicted features of PI3Kδ residues. Strongly scoring PIK3CD variants from our screens (GOF and LOF) tended to enrich in regions of p110δ predicted by AlphaFold⁵² to be more structurally ordered (Figure S4A; p = 8.4e 4, Levene’s test for homogeneity of variance) and to have less accessible surface area (Å²) (Figure S4B; p = 2.934e–4, Levene’s test). There was also a significant association between screen effect and AlphaMissense predicted pathogenicity score⁵³ (Figure S4C; p = 3.93e–22 for PIK3CD, p = 1.098e–04 for PIK3R1, Levene’s test). Importantly, some strong-scoring variants captured in our screen did not have a high predicted pathogenicity score in AlphaMissense (e.g., PIK3CD p. S444P, p.D529G; PIK3R1 p.V97A). Furthermore, AlphaMissense predicts damaging variants, but it cannot directly predict their functional consequence⁵³ (i.e., GOF vs. LOF; e.g., PIK3R1 p. L714P vs. p.L573P), which are readily resolved by our screen (Figure S4C). This emphasizes the critical value of empirical experimental examination of variant effects.

Novel GOF and LOF variants with functional characteristics analogous to known variants

To empirically validate variant effects from the pAKT/pS6 screen, we selected 26 consistently enriched and depleted sgRNAs (20 enriched; 6 depleted; and 1 NT-sgRNA control) (Figure 2B). These sgRNAs generate a spectrum of missense variants across nearly all annotated domains of PIK3CD and PIK3R1 and several splice-site disruptions in PIK3R1, a majority of which generate either previously unannotated variants (13 enriched sgRNAs and 6 depleted sgRNAs), or VUS⁷ (Figure 2B). Notably, these sgRNAs also exhibited a spectrum of effect sizes (i.e., magnitude of LFC) in the pAKT/pS6 screen (Figure 2B). We also included several sgRNAs generating ClinVar Pathogenic variants associated with APDS (Figure 2B). We base edited T cells from multiple human donors in an arrayed format, confirmed efficient editing (Figures 2C and S5A), and performed analogous phospho-flow cytometry assays for pAKT/pS6 (Figures 2D and 2E).

As anticipated, T cells with known pathogenic GOF variants had markedly elevated pAKT/S6 (Figure 2D, red), while known LOF variants⁴¹ had decreased AKT/S6 phosphorylation (Figure 2E, red), compared with cells receiving a NT-sgRNA. Previously unclassified variants and VUS also performed consistently with predictions from our screens, with screen-identified GOF exhibiting elevated pAKT/pS6 (Figure 2D, blue) and screen-identified LOF with decreased pAKT/pS6 (Figure 2E, blue).

Functional landscape of screen-discovered GOF and LOF variants

Next, we broadened our analysis to all variants selected for validation (Figure 2B) and examined their AKT and S6 phosphorylation profiles across multiple stimulation conditions. Consistent with screen results, arrayed base editing of these variants produced a broad spectrum of significant effects on T cell pAKT and pS6 signaling after a 20-min CD3/CD28 stimulation (Figures 3A and 3B). Remarkably, nearly all screen-identified GOF variants (19/20) exhibited significantly elevated pAKT following stimulation, including 4 variants previously annotated as VUS and 13 variants without previous annotation, and all LOF variants had significantly decreased pAKT (Figures 3A and 3C). These findings were consistent for pS6 (Figures 3B and 3C), although GOF variants had distinct impacts on pAKT and pS6 levels, suggesting variant-specific consequences for PI3Kδ signaling outputs (Figures 3C and S5B). Notably, among these validated GOF variants, all 20 were recovered by the pAKT/pS6 screen (using either a two-sided p value cutoff of 0.05 or false-discovery rate [FDR] cutoff of 0.1), while only 12/20 GOF were recovered by the proliferation screen when using the same p value threshold and only 10/20 when using the same FDR threshold (Table S2).

Figure 3. — (A and B) (A) MFI of pAKT (pS473) and (B) percent cells positive for pS6 (pS235/pS236) in T cells base edited with *PIK3CD* or *PIK3R1* variants or a NT-sgRNA. T cells were either unstimulated, stimulated for 20 min in the presence of DMSO vehicle control, or stimulated for 20 min in the presence of 100 nM leniolisib (Len). Dotted line represents mean of replicates for the stimulated plus DMSO condition for control NT-sgRNA T cells. n = 3 biological replicates.

(C) Heatmap of variant effects in pAKT/pS6 screen (normalized LFC of N = 3 donors, left), and in arrayed phospho-flow cytometry validation experiments (normalized pS6 pS235/pS236 percent positive, middle; normalized pAKT pS473 MFI, right; and all values from 20-min stimulation plus DMSO condition). n = 3 biological replicates for validation experiments with each variant are shown as individual rows.

(D) Scatterplot of variant effects in arrayed validation experiments (pAKT pS473 MFI, y axis) and pAKT/pS6 screens (LFC, x axis). Dot sizes scaled by screen p value.

(E and F) *PIK3CD* and *PIK3R1* variant effects in validation experiments in (A and B) mapped to the solved structure of PI3Kδ: (E) Normalized pAKT (pS473) MFI with 20-min stimulation plus DMSO, (F) Normalized pS6 (pS235/pS236) % positive with 20-min stimulation plus Len.

(G) Representative flow cytometry histograms for pS6 (pS235/pS236) across indicated conditions for GOF variants identified to be fully Len-sensitive or (H) partially Len-resistant.

One-way ANOVA with Dunnett’s test for multiple comparisons in (A) and (B), performed for the Stimulated + DMSO condition comparing all variants with the NT-sgRNA control (*p < 0.05). Error bars represent mean ± standard deviation. Simple linear regression in (D). All results repeated in at least N = 3 healthy human donors.

Related to Figure S6.

Noting the broad spectrum of effect sizes in PIK3CD and PIK3R1 GOF variants in the pAKT/pS6 screen and validation experiments, we assessed whether the effect size of GOF variants in the screen (i.e., LFC of sgRNAs) was predictive of the magnitude of GOF variant effects in arrayed validation experiments. We found a significant positive correlation between screen LFC and pAKT (pS473) MFI (Figure 3D), and a weaker correlation between LFC and pS6 (pS235/pS236) MFI (Figure S5C), in arrayed validation experiments. This suggests that our screen results are sufficient not only to nominate GOF and LOF variants but also to provide initial insights into their effect sizes, particularly for pAKT. Finally, we mapped the empirically validated signaling effects of each newly identified GOF and LOF variant to the solved structure of PI3Kδ (Figure 3E).

Precision base editors, rational sgRNA design, and computational deconvolution enhance resolution of variant effects

The editing window of ABE8e (i.e., positions 3–9 in the sgRNA protospacer) can result in multiple amino acid conversions associated with one sgRNA, which may occur in the same cell, posing a challenge for dissecting effects of co-occurring variants.^39,41,54,55 To address this, we leveraged a recently developed next-generation base editor, ABE9, which incorporates two amino acid changes into ABE8e which substantially narrow the editing window (i.e., positions 5–6 in the sgRNA protospacer) to improve base editing precision⁵⁶ (Figure S6A). We cloned these mutations into our existing NG-ABE8e construct, enabling the use of ABE9 with the relaxed “NG” PAM to increase sequence accessibility (STAR Methods).

We applied ABE9 to dissect variant effects in several contexts in which ABE8e editing presented ambiguity. First, we selected a validated GOF sgRNA (protospacer 1), which generated PIK3CD p.E525G and p.H526R at similar frequencies with ABE8e (Figure S6B, top). We used ABE9 with the same sgRNA to precisely generate p.E525G (Figure S6B, bottom), confirming p.E525G as a GOF variant (Figure S6C). Next, we selected a validated GOF sgRNA (protospacer 2) generating PIK3CD p.S444P and p.V445A with ABE8e. We used ABE9 to precisely generate p.V445A and rationally designed a second sgRNA (protospacer 3), which contains only p.S444P in the ABE8e editing window (Figure S6D). Profiling of PI3K signaling in these variants identified p.S444P as a GOF and p.V445A as a likely neutral “passenger” variant (Figure S6E). We also demonstrated the capability of rational ABE8e sgRNA design to dissect the effects of PIK3CD p.D529G and p.K528G (Figure S6F, protospacers 4 and 5), confirming both variants as GOF (Figure S6G). To increase the scale of ABE9-mediated variant disambiguation, we repeated the pAKT/pS6 screens with ABE9 using the same T cell donors (N = 3) and sgRNA library. Top enriched sgRNAs in the ABE9 screen included ClinVar Pathogenic/Likely Pathogenic variants (PIK3CD p.C416R, PIK3R1 exon 11 splice disruption) and other variants we have validated as GOF (e.g., PIK3R1 p.L570P), confirming robustness of the screen (Table S2). Integration of screen data from ABE9 and ABE8e supports the resolution of additional definitive variant effects from the ABE8e screen (Table S2; STAR Methods).

To further support disambiguation of variant effects from our screen, we also used a multiple linear regression model to generate a variant-level score and Wald test p value for most missense variants covered by our screen (Table S2). This approach reliably identified all three ClinVar Pathogenic missense variants in our screen as significant GOF and identified many other strongly scoring variants which we validated in our study, some of which we also disambiguated using ABE9 and rational ABE8e sgRNA design (e.g., PIK3CD p.S444P, p. D529G, and p.C991R) (Table S2). Together, these data demonstrate the power of integrating precision base editors, rational sgRNA design, and computational approaches to dissect variant effects at single-base resolution.

Sensitivity of novel GOF variants to the PI3Kδ inhibitor leniolisib

We next evaluated the sensitivity of each of our validated variants to leniolisib (Len), a small-molecule selective PI3Kδ inhibitor,³⁵ which was recently FDA-approved for the treatment of APDS.^34,36–38 We measured induction of pAKT/pS6 following T cell stimulation (STAR Methods) in cells harboring screen-identified variants and treatment with either vehicle control (DMSO) or 100 nM Len³⁴ (Figures 3A and 3B). All (20/20) screen-discovered GOF variants responded to Len as indicated by reduction in both pAKT and pS6 (Figures 3A and 3B). Most variants exhibited complete Len sensitivity, with Len treatment reducing pAKT and pS6 to the level of unstimulated cells or even below (Figures 3A, 3B, 3F, and 3G). However, several closely localized GOF variants in PIK3R1 within the inter-SH2 (iSH2) domain (p.K567G; p. L570P; p.L573P; and exon 14 SA disruption) (Figure 3F)—which accounts for many known APDS2 patients^14,31—were partially Len resistant as measured by pS6 (Figures 3B, 3F, and 3H). Relative to APDS1 (i.e., APDS caused by GOF in PIK3CD), there have been very few APDS2 (i.e., APDS caused by LOF in PIK3R1) patients reported in leniolisib trials,^34,57 and as such, this observation raises concern that PIK3R1 iSH2 domain variant carriers could have differential clinical responses to Len therapy.

Leniolisib corrects the dysfunctional T cell state in engineered GOF variants

The molecular underpinnings of Len efficacy in patients with APDS are incompletely understood. Our ability to engineer primary human T cells that reflect APDS biology and response to Len offered a unique opportunity to address this gap in knowledge. We activated and expanded T cells with each of these variants in the presence or absence of Len followed by flow-cytometric phenotyping (Figure 4A). Consistent with described patients cell states,^18,22 untreated cells harboring screen-identified GOF variants had elevated hallmarks of T cell exhaustion and activation when compared with control (NT-sgRNA) T cells. T cells with these GOF variants had elevated expression of CTLA4 (Figures 4A and S7A; 19/20 GOF), PD1 (Figures 4A and S7B; 20/20 GOF), and the transcription factor TOX (Figures 4A–4C and S7C; 18/20 GOF)—a marker of terminally differentiating T cells⁵⁸—and decreased expression of the stemness-associated transcription factor TCF1⁵⁹ (Figures 4A and S7D; 20/20 GOF). Conversely, T cells with LOF variants discovered in screens exhibited decreased expression of exhaustion markers and elevated expression of TCF1 (Figures 4A and S7A–S7D).

In addition to reducing PI3Kδ signaling (Figure 4D), Len treatment during activation and expansion of GOF-harboring T cells reduced all hallmarks of T cell activation and exhaustion, including the expression of PD1, CTLA4, and TOX (Figures 4E–4G and S7A–S7C) and rescued expression of TCF1 (Figures 4H and S7D). This Len-induced change in T cell state did not come at the expense of reduced cell proliferation (Figure S7E). While these effects were consistent across screen-identified variants, PIK3R1 iSH2 domain variants with partial resistance to Len in signaling assays (Figure 4D) also retained a relatively higher expression of TOX (Figures 4C and 4E), CTLA4 (Figure 4F), and PD1 (Figure 4G), and lower expression of TCF (Figure 4H) following Len treatment, supporting the notion that these variants are less sensitive to PI3Kδ inhibition (Figures 4I and 4J).

We also performed principal-component analysis (PCA) to integrate signaling and phenotyping readouts (including pAKT, pS6, CTLA4, PD1, TCF, and TOX) across all variants and stimulation conditions. PCA of the stimulated condition (without Len) revealed that the main determinant of variability (PC1, explaining 83.8%) separated GOF from LOF variants (Figure S7F). A separate analysis of the stimulated plus Len-treatment condition showed that PC1, explaining 82.67% of variability, separated partially Len-resistant PIK3R1 iSH2 variants from all other PIK3CD/PIK3R1 GOF variants (Figure S7G).

These results demonstrate that screen-identified variants faithfully recapitulate signaling and T cell changes seen in patients with APDS and provide opportunities to further dissect the pathophysiology and drug response mechanisms.

APDS patient samples reflect T cell signaling, differentiation, and Len responses seen in screen-discovered GOF variants

We next evaluated whether screen-identified variants were representative of phenotypes and Len responsiveness seen in patients harboring GOF variants in either PIK3CD or PIK3R1. For this purpose, we obtained peripheral blood mononuclear cells (PBMCs) from two patients with APDS (Figure 5A), including one with a heterozygous PIK3CD variant (c.3,061G>A; p. E1021K) and a second with a heterozygous PIK3R1 variant (c.1,692C>A; p.N564K) (Figure 5B). Importantly, both variants mapped to structural GOF hotspots from our screen (Figure 5C). We found significantly increased pS6 (pS235/pS236) (Figures 5D and S8A) and pAKT (pS473 and pT308) (Figures 5E, 5F, and S8B) in patient T cells compared with those from three healthy donors (HDs), which decreased with Len treatment (100 nM) (Figures 5D, 5E, S8A, and S8B).

Compared with T cells from HDs, both patient samples had significantly increased expression of surface proteins and transcription factors associated with activation (CD69) (Figures 5G and S8C) and exhaustion (PD1, CTLA4, and TOX) (Figures 5H, 5I, S8D, and S8E) and decreased expression of TCF1 (Figures 5J and S8F), in both CD4 and CD8 subsets. Consistent with our findings in GOF variant-engineered T cells, Len treatment significantly decreased CTLA4 (Figure 5K), PD1 (Figure 5L), and TOX (Figure 5M) expression. Notably, following Len treatment the PIK3R1 p.N564K patient sample retained higher expression of CTLA4 and PD1 compared with PIK3CD p.E1021K (Figures 5K and 5L). This aligned with our findings in T cells engineered with PIK3R1 variants in this domain (iSH2), including p.K567G, p.L570P, and p.L573P, which demonstrated similar features of Len resistance.

Len treatment also markedly enhanced T cell expansion (Figure 5N) and viability (Figure S8G), which was accompanied by changes in morphology from an elongated to a round appearance (Figure S8H).

Together, the effect sizes in aberrant PI3K/AKT signaling and T cell dysfunction, and restoration of these features in response to Len in patient samples, corresponded with screen-identified GOF variants, emphasizing that we achieved faithful reflection of patient biology in our screens and assays.

Combinatorial treatments to overcome Len resistance in engineered and APDS patient T cells

While Len treatment has substantially improved outcomes for patients with APDS and has significant in vitro activity for many PIK3CD and PIK3R1 variants, responsiveness to Len varies clinically, and the underlying mechanisms are poorly understood.³⁸ Our data suggest that PIK3R1 iSH2 domain variants, both engineered (Figures 3 and 4) and in patient samples (Figures 5K and 5L), had reduced sensitivity to Len, which motivated us to further characterize this phenomenon and design combination therapy strategies to improve responsiveness of cells harboring these variants. To further compare Len response between APDS patient samples, we stimulated both patient samples and a HD for 24 h in the presence of Len or DMSO and found that the PIK3R1 p.N564K sample had a significantly decreased response to Len as measured by relative reduction of S6 phosphorylation (Figure 6A) and expression of its downstream transcriptional target, the transferrin receptor CD71⁶⁰ (Figure 6B).

Figure 6. — (A and B) (A) Relative expression of pS6 (pS235/pS236) and (B) CD71 in APDS patient T cells after stimulation for 24 h in the presence of either DMSO vehicle control or 100 nM Len. For each patient sample, expression of each marker is normalized to the mean of replicates in the DMSO condition. n = 3 biological replicates.

(C) Representative schematic of combinatorial PI3K pathway targeting with drugs targeting p110δ (leniolisib), mTORC1 (rapamycin and everolimus), and mTORC2 (rapamycin).

(D and E) (D) Expression of pS6 (pS235/pS236) in a patient sample with *PIK3R1* p.N564K and (E) a patient sample with *PIK3CD* p.E1021K after 24 h of stimulation with either DMSO vehicle control or 100 nM of the indicated drug(s) (leniolisib, Len; rapamycin, Rap; and everolimus, Evr). n = 3 biological replicates.

(F and G) Same as (D) and (E) but for CD71, with representative flow cytometry histograms shown.

(H and I), Same as (D) and (E) but for TOX after 7 days of activation and expansion with the indicated drug treatment. n = 3 biological replicates. Representative flow cytometry histograms shown.

One-way ANOVA with Tukey’s test for multiple comparisons in (A) and (B). One-way ANOVA with Bonferroni’s test for multiple comparisons in (D)–(I). Error bars represent mean ± standard deviation.

Related to Figure S9.

We reasoned that deeper inhibition of the pathway by combining Len with inhibitors of mTORC1 (with rapamycin, Rap; or everolimus, Evr) or mTORC1/mTORC2 (with Rap) (Figure 6C) could provide necessary inhibition of aberrant signaling (Figure 6C). We first base edited PIK3R1 variants with features of Len resistance in healthy T cell donors, stimulated them in the presence of DMSO, Len, Rap, or Evr, or combinations thereof, and assessed CD71 expression. Consistent with our previous findings, Len-resistant variants maintained elevated CD71 following Len treatment, but we found a reduction of CD71 expression for all variants (nearly to control level) when Len was combined with either Rap or Evr (Figures S9A and S9B). We performed a similar experiment in a representative Len-resistant variant (PIK3R1 p.L573P) with a dose-matrix of Len plus Evr or Len plus Rap and confirmed that combining Len with either Evr or Rap produced additive effects on pathway inhibition⁶¹ (Figures S9C and S9D; STAR Methods).

Next, we assessed whether these findings were predictive of responses in APDS patient samples. Indeed, while Len treatment reduced pS6 (Figures 6D, 6E, S9E, and S9F), CD71 expression (Figures 6F, 6G, S9G, and S9H), and TOX expression (Figures 6H, 6I, S9I, and S9J) in both patient samples, combination treatment with either Len plus Rap or Len plus Evr further reduced expression of each of these markers compared with Len alone (Figures 6D–6I and S9E–S9J). These effects were consistent in HD T cells (Figures S9K–S9N). Together, these results present a rational combinatorial drug treatment approach, which reveals a non-redundant role for mTORC1 and further decreases PI3Kδ pathway signaling and features of T cell exhaustion in engineered cells and APDS patient samples, including in the setting of partial therapeutic Len resistance predicted by our work.

Prevalence of screen-discovered GOF variants and associated clinical phenotypes in population databases

The current estimate for APDS prevalence is approximately one in every million individuals, without known enrichment via founder effects.⁶² To explore whether carriers of putative GOF variants of PIK3CD and PIK3R1 identified in our screen have phenotypes suggestive of APDS, we queried a cohort of patients with a suspected but undiagnosed IEI referred for clinical gene panel sequencing (which includes PIK3R1 and PIK3CD) (“IEI cohort”; N = 8,453 patients) (Figure 7A) and a large population database (All of Us [AOU]; N = 633,540 patients). We first searched for the carrier frequencies of variants identified in our screen with significant enrichment (p < 0.05, positive LFC) in any of the three donors (including the PIK3CD p.M285T variant, which we validated as a mild GOF (STAR Methods; Figure S9O). Seven carriers of newly identified and matching GOF variants were found in the APDS IEI cohort, five of whom had humoral immune deficiency, and four of whom had cytopenias (Figure 7B). Furthermore, ACMG Criteria for Variant Interpretation PS1—strong support of pathogenicity—are met when a different variant is derived from the same amino acid as a pathogenic variant.⁶³ We had detailed clinical information for 20 such patients, of whom 13 had humoral immune deficiency with recurrent sinopulmonary infections, 9 had atopic disease, 7 had cytopenias, 4 had enteropathies, 4 had lymphoproliferative disease, and 4 had autoimmune disease (Figure 7B; Table S4). Thus, our screen identified 27 relevant variant carriers in the IEI cohort with a clinical presentation consistent with APDS.

Figure 7. — (A) Integration of pAKT/pS6 screen data with *PIK3CD* and *PIK3R1* VUS encountered in patients enrolled in a clinical IEI cohort with signs and symptoms suggestive of APDS.

(B) Immuno-pathologies encountered in N = 27 patients from the IEI cohort described in (A) with either an exact variant match or different variant at the same amino acid as a GOF identified in the pAKT/pS6 screen (STAR Methods) (N = 7 and N = 20 patients, respectively).

Related to Figure S9.

In the AOU database, 123 carriers of GOF matches were identified. 12% of these patients had multiple comorbid severe immunopathologic diagnoses but cannot be described in further detail in accordance with AOU reporting rules. Next, we performed an enrichment analysis by querying the AOU database for ICD10 codes in GOF carriers and non-carriers and found enrichment of variant carriers in individuals with a variety of IEI-related diagnoses, including gastrointestinal, autoimmune, inflammatory, infectious, and lymphoproliferative phenotypes, which can be seen in APDS (Table S5). These associations provide evidence that the variants confer increased risk for immune-mediated disease. We also evaluated summary statistics of the UK BioBank cohort (UKBB)⁶⁴ for immune diagnosis enrichment in PIK3CD p.M285T carriers and controls—as that was the only variant found in the summary statistics data. We found enrichment for multiple GI, lymphoproliferative, and inflammatory phenotypes, which can be found in APDS and other IEI (Table S5).

Therefore, given the novel variants and associated clinical phenotypes we have identified, we estimate that the frequency of any PIK3CD/PIK3R1 putative GOF variant carrier is approximately 1 in 5,000—a proportion of which have immune pathology and risk for individual immune-related diagnoses. Thus, given the high frequency of putative PIK3CD/PIK3R1 GOF variants, the prevalence of APDS may be higher than previously estimated.

DISCUSSION

The number of reported genetic defects continues to grow each year, and despite the discovery of approximately 200 primary immune deficiencies associated with variants in more than 600 genes, there are around 20 new genetic diseases of the immune system reported each year.^65,66 The accompanied increase in the number of discovered VUS emphasizes the pressing need to functionally classify thousands of variants to accelerate diagnosis, and when available, precision therapies.

However, several limitations have impacted existing variant classification approaches. Existing methods are typically low throughput and rely on delivering supraphysiologic gene dosage (e.g., overexpression of a variant of interest with plasmid transfection or lentiviral transduction)^67–70 in systems that do not provide the adequate cellular context for genes of interest (e.g., study of variants in HEK293 cells rather than T cells) and are often not in compliance with consensus PS3/BS3 clinical variant interpretation guidelines for utilizing functional assay data in variant interpretation.⁴⁷ Another challenge is the retroactive nature of these studies, which are frequently performed in response to encountering a variant in a patient with a potential diagnosis, leading to significant delays in definitive diagnosis. Furthermore, some functional assays or screens use proxy readouts to study variant effects, such as proliferation and survival, rather than a direct GOF/LOF classification assay,⁶⁷ which has implications with respect to sensitivity of detecting strong and subtle GOF/LOF variants, as shown in our work comparing pAKT/pS6 over proliferation readouts.

Our work addresses several challenges in APDS diagnostics¹³ and provides an integrated framework that can be broadly applied to hundreds of IEI. Intriguingly, the high resolution of presented screens enabled assessment of variant effects beyond a binary GOF/LOF variant classification, identifying a spectrum of variant effect sizes, which we validated in an arrayed fashion and correlated quantitatively. We identified >100 GOF and LOF variants, including variants previously classified as VUS, in PIK3CD and PIK3R1 and identified previously underappreciated structure-function relationships through structural modeling of variant effects that are not robustly identified using in silico prediction models. We have also presented our functional assay and the screen data to the ClinGen Antibody Deficiencies Variant Curation Expert Panel (VCEP), with whom we have exchanged helpful discussion. We anticipate that this information will be useful and informative for crafting the PIK3CD and PIK3R1 gene-specific classification guidelines, particularly certain specifications such as PS3/BS3.

Importantly, the scale of variant discovery and dynamic range allows us to couple our findings to the identification of real-world carriers of these variants and link to clinical outcomes, which could now benefit from precision approaches. The patient populations and settings implicated here go beyond the rare and unsolved cases for whom the reclassification can have immediate benefit, but also to those with immune-related diagnoses and disorders of variable severity and comorbid states. Such patients, especially when they are adults, are rarely assessed for IEI, but our results illustrate the powerful potential of such evaluations for true precision medicine approaches. As evidenced, our study revealed clinically meaningful variants with lower magnitude GOF activity, which were present at higher allelic frequency than usually assumed for IEI, likely leading to disease with lower penetrance (e. g., whether there is an immune phenotype) and expressivity (e.g., what specific immune-mediated phenotype, how many of them, and what severity). Such variants, which have not been previously documented in preclinical or clinical studies, can drive disease and can be pharmacologically targetable when a carrier is symptomatic. This is consistent with a recent study describing an unexpectedly more frequent carriers of mild GOF variants in JAK1 that results in clinical presentation of autoimmune and inflammatory disease (JACCD syndrome) and is responsive to JAK1/JAK2 inhibitor therapy.⁷¹ Based on our integrative analyses of our screen data and publicly available whole-genome sequencing datasets, we estimate that the prevalence of carriers of variants potentially causing APDS may be one to two orders of magnitude higher than previously assumed (i.e., 1 in 10,000 rather than 1 in 1–2 million persons in the US). Such findings indicate that greater clinical awareness of presentations, which could be caused by such variants in APDS and other IEIs, is necessary and should encourage recommendations for broader comprehensive NGS testing in appropriate contexts.

Limitations of the study

Our study has important limitations: although several novel variants we identified had associated clinical phenotypes in patient databases, we were unable to engineer all possible PIK3CD and PIK3R1 amino acid substitutions due to current limitations of base editing.⁷² Development of efficient editors with expanded mutagenesis capabilities and improvements to the efficiency of other gene editing technologies (e.g., prime editors⁴⁰) may enable unbiased screening of all possible base substitutions and subsequent reclassification of new or encountered human genotypes. While current methods predominantly install biallelic edits,^41,54,73 many IEI are caused by heterozygous variants which lead to haploinsufficiency or dominant-negative effects,⁶⁶ and modeling these contexts will further improve our understanding of these disorders. Lastly, other determinants of the phenotypic profile of individuals with any given pathogenic GOF variant remain unknown and are difficult to establish with human genomic profiling data or to model given the high number of potential genetic and polygenic interactions with causative variants. These challenges are not unique to APDS and require novel experimental and analytical approaches, including large scale phenotyping tools, which capture immune-mediated disorders.⁷⁴

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to the lead contact, Benjamin Izar (bi2175@cumc.columbia.edu).

Materials availability

Plasmids generated for this study and used for IVT of NG-ABE8e and NG-ABE9 are available upon request.

Data and code availability

All base editor screening data (MAGeCK test outputs and multiple linear regression data) are provided in labeled tabs of Table S2. Code giving rise to analyses and figures in this work is available at https://github.com/cf3041/genetic-variants-IEI. Raw and processed sequencing data have been deposited in NCBI GEO under accession number GEO: GSE298853. Source data giving rise to figures in this manuscript are deposited in Dataverse: https://doi.org/10.7910/DVN/92WQLH.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Primary T cell isolation and culture

Buffy coats from healthy human donors (fully deidentified, except for blood type) were obtained from the New York Blood Center. Peripheral blood mononuclear cells (PBMCs) were isolated from buffy coats with SepMate 50 ml conical tubes (StemCell, #85450) containing 15mL of Ficoll-Paque Plus media (1.077g/mL, Cytiva, #17144002). RBCs were lysed in ACK buffer (Gibco, #A1049201) and CD3 T cells were isolated by negative selection with the EasySep human T cell isolation kit (StemCell, #17951). For patient samples, T cells were isolated from PBMCs from patients with APDS1 or APDS2 as described above. All T cell cultures were performed with OpTmizer SFM (Gibco #A1048501) supplemented with 1:40 OpTmizer Supplement (Gibco #A1048501), 1:100 GlutaMAX (Gibco #35050061) using a humidified incubator at 37 °C with 5% CO₂.

METHOD DETAILS

sgRNA library design for base editor tiling of PIK3CD and PIK3R1

All exons in PIK3CD and PIK3R1,and 20 bases into each intron, were tiled with sgRNAs using the Base Editor Design Tool.⁷³ The NG-ABE8e editing window was set from base 3–9 of each sgRNA. Although, sgRNAs with a TTTT sequence are known to be transcribed less efficiently from the hU6 promoter,⁸² we kept these sgRNAs in the final library as some have still been shown to have strong effects.⁸³ The final library included 2,195 sgRNAs generating missense variants, 138 sgRNAs generating splice site-variants, 404 sgRNAs generating intronic variants, 32 sgRNAs generating variants in the 5’ and 3’ UTRs, as well as negative-control sgRNAs generating either silent variants (202 sgRNAs) or having empty editing windows (719 sgRNAs) containing no targetable adenine base. For positive controls (i.e., dropout controls) assessed in the proliferation arm of the screen, the top 100 most depleted “pan-essential” gene perturbation sgRNAs from our previously published NG-ABE8e T cell proliferation screen were used.⁴¹

This library could generate perturbations in 74% of all residues in PIK3CD and 69% of all residues in PIK3R1. Assuming a total of 19 possible amino acid conversions per residue, this library represents 11% of all possible amino acid conversions across PIK3CD and 7.5% of all amino acid conversions across PIK3R1 that could theoretically be achieved by saturation mutagenesis. The library had a coverage of 154/1377 ClinVar 1- and 2-star missense and splice variants in these genes, including 8/38 (21%) of all pathogenic and likely pathogenic variants, and 5 benign variants.⁷

Base editor library cloning

Custom oligonucleotide pools (Twist Bioscience) encoding the individual sgRNA sequences and flanking regions for library amplification and cloning were used for pooled golden-gate cloning of sgRNA gene editing libraries into the CropSeq vector as previously described. The following sequence was used for all oligonucleotide pools: 5^′-[Forward Primer] CGTCTCACACCG [sgRNA, 20 nt] GTTTCGAGACG [Reverse Primer]-3’. The library was amplified library-specific primers and KAPA HiFi HotStart ReadyMix (Roche, #KK2601), cleaned up using the QIAquick PCR purification kit (Qiagen, #28104), and eluted in 50 μL TE buffer (Invitrogen, #12090015). 5 μg of sequencing-verified CropSeq-mTurquoise lentiviral backbone was restriction digested with BsmBI V2 (NEB, #R0739S) in a 50 μl reaction with NEB3.1 buffer (NEB, #B7203S) for 2 hours at 55 °C. After digestion the linearized backbone was dephosphorylated with 2 μl rSAP (NEB, #M0371S) at 37 °C for 1 hour, followed by heat inactivation at 80 °C for 20 minutes. The linearized backbone was extracted from the gel using the Zymogen Gel DNA recovery Kit (#D4001), cleaned up with 1x SPRI bead selection (Beckman Coulter, #B23318) and eluted in TE buffer. The linearized backbone and amplified oligonucleotide pool were combined at equimolar ratio and assembled using a 50 μl golden-gate reaction containing 1x Tango Buffer (Thermo Fisher, #BY5), 1 mM DTT (Thermo Fisher, #R0861), 1 mM ATP (NEB, # P0756S), 1 μl Esp3l (Thermo Scientific, #ER0451), T7 ligase (NEB, #M0318S) and nuclease free water. The reaction was conducted in a thermocycler with 99 cycles of incubation at 37 °C for 5 minutes followed by 20 °C for 5 minutes. The final product was cleaned using 1x SPRI beads and eluted in 11 μl TE. 10 μl of this eluate was electroporated into 50 μl Lucigen Endura electrocompetent cells (Lucigen, #602421) using 1 mm electroporation cuvettes (Biorad, #1652089) and a BioRad Gene Pulser Xcell (Biorad, #1652660) set to 10 μF, 600 Ohms, 1.8 kV. 1 ml of prewarmed recovery medium (Lucigen, #800261 was added and the cells were incubated for 1 hour at 37 °C shaking at 250 rpm. The cells were then plated on a 25×25 cm LB-agar plate with 100 μg/ml Carbenicillin and incubated for 14 hours at 32°C. Colonies were then scraped from the plates into LB buffer and collected by centrifugation and plasmid DNA was extracted with a Qiagen HiSpeed Maxi kit (Qiagen, #122663). sgRNA library distribution was assessed using next generation sequencing (targeting a Gini index <0.05) before preparation of lentivirus.

Lentiviral sgRNA library production and concentration

HEK-293T cells were passaged at least two times in DMEM (Gibco #11965092) + 10% FBS (Gibco #A5670701), splitting at ∼80% confluence, and then plated at 825,000 cells/well in 6-well plates. 24 hours later, at 70–80% confluence, cells were transfected using TransIT-LT1 (MirusBio, #MIR2304). Cells were transfected with 500 ng psPAX2 (a gift from Didier Trono, Addgene plasmid #12260), 250 ng VSV-G (a gift from Didier Trono, Addgene plasmid #12259), and 500 ng lentiviral transfer plasmid per well. 18 hours after transfection, media was replaced with DMEM + 20% FBS. 24 hours later, lentivirus-containing supernatant was harvested from cells and clarified by filtration with a 0.45 μM PES filter (Fisher Scientific, #09–720-514). Clarified lentiviral supernatant was concentrated to 10% of the original volume with Lenti-X concentrator (Alstem Bio, #631232). Concentrated lentivirus was resuspended in sterile OpTmizer medium and either flash frozen on dry ice and stored at –80 °C or used immediately for transductions.

Lentiviral transduction and selection of T cells for screens

T cells were pre-activated with Immunocult CD3/CD28/CD2 T cell activator (Stemcell, #10990) with 100 IU/ml IL2 (Chiron #53905–991-01) for 72 hours prior to transduction. T cells were then plated in flat-bottom 96-well plates at 100,000 T cells per well with 1:100 v/v LentiBOOST transduction enhancer (Mayflower Biosciences, # SBPLV10112) and 1:10 v/v of 10x concentrated lentivirus. Plates were spun at 800 x G for 90 minutes at 32 °C in a pre-warmed centrifuge and then cultured overnight at 37 °C. 24 hours after centrifugation, beads were removed using a magnetic rack and T cells were expanded in G-Rex 6-well plates (Wilson Wolf, #80240M) with 300 IU/ml IL2. 48 hours after spinoculation, T cells were selected for 3 days in 1 μg/ml Puromycin (Gibco, # A1113803) followed by a washout and expansion without puromycin for an additional 7 days to allow cells to recover. Puromycin selection was confirmed by enrichment of mTurquoise-positive T cells by flow cytometry.

In vitro transcription (IVT) of Base Editor mRNA

The base editor NG-ABE8e (a gift from David Liu, Addgene # 138491) was cloned into the IVT template vector (a gift from David Liu, Addgene #193843) as previously described.⁴¹ The DNA sequence of ABE9 was obtained from Chen et al.⁵⁶ The N108Q and L145T variants needed to generate NG-ABE9 were introduced into NG-ABE8e using Gibson cloning: briefly, PCR of the NG-ABE8e IVT backbone was performed with KOD Xtreme Hot Start DNA polymerase (MilliporeSigma, #71975–3). Two single-stranded DNA oligos (containing an ABE9 sequence fragment with the N108Q and L145T variants) were annealed by heating to 95 °C for 3 minutes followed by ramping down to 22 °C at ramp rate of 1 °C per 15 seconds. Gibson assembly of the PCR product and annealed oligos was performed using with NEB Gibson Assembly Master Mix at 50 °C for 1 hour. (NEB, #E2611L). DNA was then transformed and grown in Stbl3 bacteria (Invitrogen, #C737303), isolated by Miniprep (Qiagen, #27106), and mRNA encoding NG-ABE9 was generated by IVT as described below. All primers and DNA oligos used for Gibson assembly are included in Table S8.

The in vitro transcription protocol was adapted from Neugebauer et al.⁵⁵ The base editor IVT template was simultaneously amplified and poly-T tailed by PCR using primers IVT-F and IVT-R.^41,55 PCR was performed with Q5^® Hot Start High-Fidelity DNA Polymerase (New England Biosciences, # M0493S), then the amplicon was purified using a QIAquick PCR Purification Kit (Qiagen #28104) and in-vitro transcribed using the HiScribe^® T7 High-Yield RNA Synthesis Kit (New England Biolabs, #E2040S) in a total reaction volume of 320μl, with full substitution of N-1-methyl-pseudouridine-5’-triphosphate (TriLink Biotechnologies, #N-1081–1) for UTP and co-transcriptional capping with CleanCap^® Reagent AG (TriLink Biotechnologies, # N-7113–1). The mRNA was precipitated by mixing with 0.5 v/v Lithium Chloride (Invitrogen, #AM9480) and incubating at −20 °C for 30 minutes. mRNA was pelleted by centrifugation at 15,000xG for 20 minutes at 4 °C, washed once with ice-cold 70% ethanol, and then supernatant was removed and pellet air-dried at room temperature for 5 minutes. The mRNA pellet was resuspended in 800 μl THE RNA Storage Solution (Thermo Fisher, #AM7000), mRNA concentration was quantified using a Nanodrop Spectrophotometer (Thermo Fisher), and mRNA was further diluted to a final concentration of 1 μg/μl. Purity and size of the mRNA was confirmed using an RNA Tapestation Kit (Agilent, #5067–5579).

Base editing of primary T cells

Screen Format

Library-transduced T cells were pre-activated with Immunocult CD3/CD28/CD2 T cell activator with 100 IU/ml IL2 for 72 hours prior to electroporation. 30e6 T cells from each donor were then centrifuged at 200 x G for 10 minutes, resuspended in 120μl MaxCyte electroporation buffer (MaxCyte, #EPB-1), and combined with 30μl of 1 μg/μl NG-ABE8e or NG-ABE9 mRNA (30μg). T cells were electroporated in 3 wells of a MaxCyte 50×3 Processing Assembly (MaxCyte, #ER050U3–10) using program “Activated T Cell 1” on a MaxCyte GTX instrument (MaxCyte, #GTX) and then transferred to T75 flasks with pre-warmed complete OpTmizer and 300 IU/ml IL2, to a final cell concentration of 2e6/ml. 48 hours later, cells were transferred to a G-Rex 6-well plate and fresh media and IL2 were added every 48–72 hours. T cells were allowed to edit and expand for an additional 7 days prior to downstream screen assays and sorting.

Arrayed Validation Format

For arrayed base editing validation experiments, T cells were pre-activated for 72h in 100 IU/ml IL2 with Immunocult CD3/CD28/CD2 T cell activator. T cells were then washed 1x with PBS, resuspended at 1e6 / 20 μl in P3 nucleofection buffer (Lonza, #V4XP-3032), and 1e6 cells were combined with 1.5 μg ABE mRNA and 100 pmol sgRNA (Synthego) (Table S6). T cells were electroporated at 1e6 per cuvette well in a Lonza 4D nucleofector using electroporation program EO-115, and then 100 μl pre-warmed culture media was immediately added to each cuvette well. Cells were recovered in the cuvette for 15 minutes at 37 °C and then cultured in OpTmizer SFM at 1e6/ml with 300 IU/ml IL2. T cells were allowed to expand and edit for 7 days prior to downstream functional and phenotyping assays.

Flow cytometry-based assays

Surface staining

For all surface staining experiments, cells were collected, washed 1x with ice-cold PBS, then stained for 10 minutes on ice in the dark with 1:1000 Zombie NIR viability dye (Biolegend, #423105). Then, cells were washed with ice-cold FACS buffer and stained on ice for 20 minutes in FACS buffer with surface antibodies at dilutions pre-determined by titration experiments (all surface, intracellular, and phospho-flow antibody dilutions are listed in Table S7). When multiple brilliant dyes were used together, Brilliant Stain Buffer Plus (BD, #566385) was added according to the manufacturer’s instructions. Representative gating strategy for all flow cytometry and FACS assays shown in Figure S10.

Intracellular staining (non phospho-flow)

For intracellular staining experiments, surface staining was first performed as above. Next, T cells were fixed and permeabilized using the eBioscience FoxP3 / Transcription Factor Staining Buffer Set (Invitrogen, #00–5523-00), stained for intracellular proteins of interest with empirically determined antibody concentrations, and analyzed on a Cytek Aurora cytometer. When multiple brilliant dyes were used together, Brilliant Stain Buffer Plus (BD, #566385) was added according to the manufacturer’s instructions.

Phospho-flow (stimulation and staining)

Prior to downstream phospho-flow cytometry analysis, T cells were rested from exogenous stimulation (including IL2) overnight. To assess phosphorylation of AKT and S6, T cells were stained with Zombie NIR viability dye for 10 minutes in cold PBS, washed, then stained on ice with 20 μg/mL anti-CD3 (BD, #555329) and 20 μg/mL anti-CD28 (BD, #555725) solution for 20 minutes followed by 20 minutes of antibody cross linking on ice with 2 μg/mL goat-anti mouse solution (BD, #53998). T cells were then activated via a minute incubation period at 37 °C in pre-warmed OpTmizer CTS in a pre-warmed Thermocycler. All stains and the activation step were done with 1e6 T cells per well in a 96-well plate to ensure even stimulation. T cells were then immediately fixed with an equal volume of pre-warmed BD Cytofix Buffer (BD, #554655), permeabilized with BD Phosflow Perm Buffer III (BD, #558050), blocked for 10 minutes with normal mouse IgG (Santa Cruz, #SC-2025), stained with empirically determined concentrations of phospho-antibodies, and analyzed by flow cytometry as described above.

pAKT/pS6 Screen

Screen design

Our screen was designed in accordance with the guidelines outlined in Figure 2 of Brnich et al.⁴⁷ (1) We defined the disease mechanism of APDS, which is driven by hyperactive PI3K-delta signaling.^9,32,57 (2) We valuated applicability of assay classes used to stratify variants in APDS and selected a well-established^34,41 FACS-compatible assay with high dynamic range which directly measures signaling activity of the perturbed pathway and can clearly resolve established GOF and LOF (Figures S1A and S1B). (3) We incorporated multiple layers of controls including all ABE-compatible sgRNAs that could generate known ClinVar Pathogenic and Benign/Likely Benign variants (Tables S1 and S2), sgRNAs generating silent variants or having no editable bases (empty-window controls); included rigorous biological replicates (3 healthy human donors); and added a specific set of controls to confirm high base editing efficiency (“essential gene” perturbations in the proliferation screen)^41,43 (Figure S2A). (4) We calculated OddsPath scores using the functional results of Pathogenic and Benign/Likely Benign variant controls in our screen (Table S3).

Screen workflow

30e6 primary human T cells from each of N=3 human donors were transduced with the sgRNA library and electroporated with base editor mRNA (encoding either NG-ABE8e or NG-ABE9) using a MaxCyte GTX as described above. After allowing cells to recover and expand for 7 days, cells were stimulated and stained for pAKT (pS473) and pS6 (pS235/pS236) as described in Phospho-flow (stimulation and staining). A total of 100e6 library-edited T cells per donor were stimulated in 96-well plates with approximately 1e6 T cells per well, to ensure high library coverage even after sorting into pAKT/pS6-high and -negative expression bins. Cells were sorted using an MA-900 cell sorter (Sony) with the 100μm nozzle, binned into the top 15% and bottom 15% of pAKT/pS6 expression, and collected in FACS buffer. Following the sort, the DNA extraction protocol was immediately initiated.

DNA extraction and next-generation sequencing

Base editing screens

For unfixed samples (proliferation screen), DNA was extracted from library-edited T cells using a DNeasy Blood & Tissue Kit (Qiagen, #69504) per to the manufacturer’s protocols with buffer AL, and eluted in 100 μl buffer EB (Qiagen, #19086) per 5e6 T cells. For fixed samples (pAKT/pS6 screen), DNA was extracted as above with the following modification: samples were incubated overnight at 65 °C in buffer AL and proteinase K with shaking at 1000 RPM. Sample amplification and indexing PCR was performed with ExTaq polymerase (Takara Bio, #RR001A): for unfixed and unsorted samples, 10 μg gDNA input per 100 μl PCR with 24 cycles; for fixed samples, 1μg gDNA input per 100 μl PCR with 24 cycles. Each gDNA sample was split across 8 parallel PCRs and then products were pooled after PCR to minimize library skewing. Pooled PCR products for each sample were cleaned up with 1x SPRI, quantified by D5000 Tapestation (Agilent, #5067–5588), and then pooled targeting 1000x read coverage per sgRNA per library and sequenced with a CloudBreak Freestyle Kit (Element Biosciences, #860–00015), 75 bp Read 1, 8 bp Index 1 on an Element AVITI sequencer (Element Bioscience).

Base editing arrayed validation experiments

DNA was extracted from base edited T cells as described above. PCR primers were designed to amplify the targeted locus (to generate a ∼400 nucleotide amplicon centered around the edited base(s)) using NCBI PrimerBlast⁸⁴ (Table S8) and a 25 cycle PCR was performed with Q5 HotStart HiFi polymerase. PCR amplicons were cleaned up with 1x SPRI, amplicon purity and size were confirmed by D5000 Tapestation (Agilent, #5067–5588), and amplicons were sequenced by fragmentation-free long-read sequencing with v14 library prep chemistry (Oxford Nanopore Technology; Plasmidsaurus). Sequences were aligned to the reference (wildtype) amplicon and chromatograms were visualized in Geneious Prime (Dotmatics).

APDS patient samples

For APDS patient samples and a representative healthy donor, T cell DNA extraction, PCR primer design, PCR, and cleanup were performed as described in Base editing arrayed validation experiments (Table S8). Sample preparation was performed with the Nextera^® XT DNA Library Preparation Kit (Illumina, #FC-131–1024) per the manufacturer’s instructions using half-reactions, with 0.5 ng of PCR amplicon input for tagmentation. Samples were dual-indexed with Illumina DNA/RNA UD Indexes Set A, Tagmentation primers (Illumina, #20091654), quantified by D5000 Tapestation Kit (Agilent, #5067–5588), and pooled at equimolar quantities. Pooled libraries were diluted, denatured, and sequenced with the 2×75 CloudBreak Freestyle Kit (75 bp Read 1, 75 bp Read 2, 10 bp Index 1, 10 bp Index 2) (Element Biosciences, #860–00015) on an Element AVITI sequencer. WebLogos for variant frequency were visualized in Geneious Prime.

Screen quality control and evaluation of off-target effects

Quality control

FASTQ files were trimmed with Cutadapt.⁷⁶ Guide enrichment analysis was performed using MAGeCK.⁷⁵ For each sample, MAGeCK’s ‘count’ command was used to determine the number of sequencing reads mapped to each sgRNA. Across all sequenced samples, the mean reads per sample was 5,613,368 (range 2,842,507 to 9,279,031), mean percent of reads mapping to the sgRNA dictionary was 89.5% (range 88.0% to 91.7%), and mean Gini index was 0.066 (range 0.040 to 0.097). Complete read count, read mapping, and Gini index data for each sample are included in Table S2 under the “Screen_QC” tab.

Evaluation of off-target effects

sgRNA off target effects were evaluated as previously described.^41,54 Briefly, Cas-OFFinder⁷⁹ was used to search all sgRNAs across the full human genome to find off-target binding sites with up to 5 mismatches, using PAM “NG”. Each target site for each sgRNA was scored using the Cutting Frequency Determination (CFD) algorithm,⁸⁰ and sgRNAs were filtered in off-target analysis if the CFD algorithm found >5 off-target sites with a CFD score of 1. A very low fraction (4/3794; 0.10%) of sgRNAs met these established thresholds for high off-target activity, and none scored significantly in the pAKT/pS6 assay or were selected for downstream arrayed validation experiments. sgRNAs which met these thresholds are annotated “Yes” in Table S1 in a column titled “Off_Target”.

Screen analyses and statistics

Several statistical analyses were performed on the screen datasets:

MAGeCK’s ‘test’ command with default parameters was used to determine enriched and depleted guides in “treated” versus “control” conditions. Briefly, MAGeCK test takes sgRNA read counts from MAGeCK count, normalizes them across samples, and computes log fold changes between the median ‘treated’ count and the median ‘control’ count; a negative binomial model is used to compute p-values for the differential abundance of sgRNAs.⁷⁵ For the pAKT/pS6 screen, “control” was defined as the pAKT/pS6-negative sort bin and “treated” was defined as the pAKT/pS6-high sort bin. For the proliferation screen, “control” was defined as a sampling timepoint immediately post-base editing and “treated” was defined as a sampling timepoint 14 days after T cell re-expansion. sgRNAs with low coverage (defined as MAGeCK median normalized read count <30 in the control group) were filtered from downstream analyses. A two-sided MAGeCK P value cutoff rather than a log₂(fold change) cutoff was used to score significance to incorporate consideration of both sgRNA effect size and read count. sgRNAs were scored as significant if the MAGeCK two-sided P value was < 0.05.
A z-score was also calculated for each sgRNA in the library using the formula z = (x - μ) / σ, where x = observed LFC for the sgRNA of interest, μ = the mean LFC of empty window sgRNAs (the null distribution), and σ = standard deviation of the LFC of empty window sgRNAs (the null distribution). Additional supporting analyses to identify impactful sgRNAs were performed using a false discovery rate (FDR) cutoff < 0.1. These analyses are noted explicitly in the results section when they are used. Complete MAGeCK test outputs, z-scores, and ClinVar variant annotations for each sgRNA are included in Table S2.
OddsPath scores were calculated for Pathogenic and Benign/Likely Benign variants in the pAKT/pS6 screen⁴⁷ (Table S3). For Benign/Likely Benign controls, we had 33 sgRNAs from the screen which could precisely generate (i.e., exactly one edit) 26 unique ClinVar Benign/Likely Benign variants (N=24 likely benign, N=2 benign). For Pathogenic controls, we had 11 sgRNAs which could generate 4 unique ClinVar Pathogenic variants (for sgRNAs generating Pathogenic controls, we were permissive of multiple edits, which were present for 2 of the 4 Pathogenic sgRNAs, e.g., for PIK3CD p.E1025G_S1026G, where E1025G is pathogenic and S1026G is not). For OddsPath calculation, we defined our assay result as “functionally abnormal” for a variant if an sgRNA associated with the variant had a MAGeCK two-sided P < 0.05 and positive LFC (in the analysis across all N=3 donors), and functionally normal if no sgRNA associated with the variant had a significant MAGeCK P value (importantly, no sgRNAs generating either Pathogenic, Benign, or Likely Benign variants scored as LOF in the screen; i.e. with a MAGeCK two-sided P < 0.05 and negative LFC; Table S3). 4/4 ClinVar Pathogenic variants had an sgRNA that scored as significant GOF, while 0/26 ClinVar Benign/Likely Benign variants had an sgRNA that scored significantly (as either GOF or LOF). Based on this data, we calculated an OddsPath score of 26.0 for pathogenic variants (corresponding to an evidence strength equivalent of “PS3”) and an OddsPath score of 0.250 for benign variants (corresponding to an evidence strength equivalent of “BS3_supporting”).

Analyses and visualization of base editor screen data were conducted in Python and R. For lollipop plot visualization, sgRNAs associated with multiple edit types were visualized by their most impactful edit (splice>missense>silent) and sgRNAs making both splice donor/acceptor and missense edits were visualized as an “X” (indicating splice donor/acceptor site disruption).

Construction of a variant-level score

Given that some sgRNAs are capable of generating >1 edits (Table S1), we also constructed an amino acid residue-level score for each possible amino acid change introduced by our sgRNA library. We utilized a multiple linear regression model to assess the contribution of each residue-level variant on the guide-level LFCs from our screen. Significance of the estimated variant effects was determined using a Wald test and adjusted for multiple comparisons. This computational approach assumes possible edits made by a single guide are linearly additive and independent, which is an inherent limitation. Additionally, some variants could not be assigned scores as their effects were not uniquely distinguishable within the model. The residue-level scores and p-values are included in their own tab (“Variant_Level_Score”) in Table S2.

Visualization of variant effects on PI3Kδ protein structure

3D protein visualizations were generated using UCSF ChimeraX.⁸¹ The PI3Kδ structure was loaded from PDB entry PDB: 5DXU⁵¹ with the GDC-0326 small molecule compound removed from the structure. LFCs were converted to hex colors via linear interpolation along a monochromatic color scale, ranging from 0 to the maximum value from the corresponding experiment, and overlaid on the 3D protein complex.

Assessment of variant effect association with AlphaFold and AlphaMissense residue metrics

The Genomics 2 Proteins portal⁷⁸ was utilized to collect structural protein data for residues in PIK3CD (UniProt O00329) and PIK3R1 (UniProt P27986).⁸⁵ AlphaFold predicted local distance difference test (pLDDT)⁵² scores and accessible surface area scores (Å²) were evenly split into 10 bins and Levene’s test was conducted to assess the homogeneity of variance of LFCs from our pAKT/pS6 screen across pLDDT or Å² values.

AlphaMissense pathogenicity scores⁵³ corresponding to the GRCh38 human genome reference assembly were collected from the AlphaFold Protein Structure Database for PIK3CD (UniProt O00329) and PIK3R1 (UniProt P27986). All possible missense variants made by our base editing system were considered and pathogenicity scores were filtered to the corresponding combinations of locus and exact variant. For any sgRNA predicted to generate >1 possible missense variant, this sgRNA was assigned the highest pathogenicity score among the possible variants. Significance from the screen was assigned based on a two-sided P < 0.05 in MAGeCK test. AlphaMissense Pathogenicity scores were evenly split into 10 bins and Levene’s test was conducted to assess the homogeneity of variance of LFCs from our pAKT/pS6 screen across AlphaMissense pathogenicity scores.

Assessment of single and combinatorial drug effects on T cell signaling and phenotype

For short-term (20-minute) phospho-flow signaling experiments, T cells were pre-treated with 100 nM leniolisib (CDZ-173) (SelleckChem, #S8752) or an equivalent volume of DMSO vehicle control for three hours prior to assay initiation. T cells were then kept in Len or DMSO during all pre-stimulation antibody staining steps and during the 20-minute stimulation. The stimulation itself was performed as described in Phospho-flow (stimulation and staining). For longer-term phospho-flow (24h) or phenotyping experiments (24h, or 7–10d), T cells were activated with Immunocult CD3/CD28 cocktail and cultured in complete OpTmizer with 300 IU/ml IL2 and 100 nM leniolisib, 100 nM rapamycin (SelleckChem, #S1039), 100 nM everolimus (SelleckChem, #S1120), or combinations thereof. For 7- or 10-day activation plus expansion, media containing fresh drug and 300 IU/ml IL2 was refreshed every 48 hours and cells were expanded as necessary to keep the concentration between 0.5–2e6/ml.

Combinatorial drug effects were quantitatively assessed with Bliss synergy scoring in SynergyFinder 3.0.⁶¹ Briefly, T cells were stimulated with Immunocult for 24 hours as above with a dose-matrix of leniolisib (0nM, 25nM, 100nM, 250nM, 1000nM) plus everolimus (0nM, 0.01nM, 1nM, 10nM, 100nM) or plus rapamycin (0nM, 0.01nM, 1nM, 10nM, 100nM). All possible dose combinations of each drug were performed. T cells were then assessed for CD71 expression (a surface protein proxy for PI3Kδ pathway activity) by flow cytometry. An unstimulated/untreated control condition was included for baseline CD71 expression. CD71 MFI values across all drug concentration combinations were then normalized from 0 to 1, with 0 defined as the unstimulated/untreated condition and 1 defined as the stimulated + DMSO vehicle control condition. Percent pathway inhibition was then calculated with the formula [(1–normalized MFI value)*100]. Bliss scores for synergy were calculated and visualized with SynergyFinder and a box was drawn around the dose range with highest Bliss synergy score. A Bliss synergy score cutoff of >10 was considered to indicate synergistic drug effects, while a score between 0 and 10 was considered to indicate additive drug effects.⁶¹

Microscopy

Brightfield images of T cells from APDS patient samples cultured in 48-well plates in the presence of DMSO or Len for 7 days were acquired with a Celldiscoverer 7 Microscope (Zeiss).

Clinical database queries

IEI cohort

Through an industry collaboration with Pharming, we obtained data from a cohort of patients referred for clinical IEI gene panel sequencing (which includes PIK3R1 and PIK3CD) with a set of symptoms which could be consistent with APDS (N=8453 total patients). For inclusion in this cohort, each of these patients, by definition, had at least two of the following immuno-pathologies: bronchiectasis; lymphadenopathy >1 month; chronic hepatosplenomegaly; severe or recurrent herpesviral infections; enteropathy; lymphoma; elevated IgM; elevated interferon levels; reduced naïve B cells; or a personal or family history of common variable immunodeficiency.

Patients were included in the analysis if they had genetic variants that corresponded exactly (same amino acid and same substitution) or inexactly (same amino acid, but different substitution) to a variant which scored as a putative GOF in the pAKT/pS6 screen. We were inclusive of sgRNAs generating >1 variant (e.g., if a patient had variant X and a sgRNA generated variant X and Y) and have annotated these in Table S4, column “T”. For this analysis, we defined putative GOFs as significantly enriched (i.e., scoring in the screen with two-sided P <0.05 and positive LFC) in at least one donor, and not significantly depleted (i.e., scoring in the screen with a two-sided P <0.05 and negative LFC) in any donor. For these patients, we examined clinical data and identified the number of patients with the following categories of severe immuno-pathologies: recurrent sinopulmonary infection and humoral immune deficiency; herpesviral infection; lymphoproliferative disease; autoimmune disease; atopic disease; enteropathy; cytopenias; and bronchiectasis (Table S4).

All of Us cohort

The All of Us Researcher Workbench cohort builder was used with version 8 of the dataset to screen hundreds of thousands of individuals with available whole genome sequence data. Patients were included in the analysis if they had genetic variants corresponding to screen variants with a reported LFC of greater than or equal to 1.0 and two-sided P value < 0.05 in at least one donor. All guides which produced only one variant were included by default. For instances where guides produced multiple variants, we applied several deconvolution approaches to define which variant to include: (1) Manual deconvolution: We first searched for instances where significant GOF or LOF multiple-edit guides contained a variant which overlapped with a single-edit guide. When the single edit guide result matched the multiple-edit guide, the additional variant(s) were not included. If the single edit guide did not lead to significant gain or loss of function, then the other variant was designated as the functional variant. (2) Experimental high-throughput deconvolution: To validate that manual strategy, and to further remove ambiguity from variant effects in this setting, we integrated screen results from the ABE8e and ABE9 pAKT/pS6 screens (Table S2). We incorporated variants which scored as a significant GOF in the ABE8e screen and had a concordant direction of effect (i.e., positive LFC) in the ABE9 screen. Finally, the PIK3CD p.M285T variant was also included as we experimentally validated it as a mild GOF (Figure S9O). The final list of variants queried is listed in the “Variants Queried” tab of Table S5.

An in-house pipeline was used to translate patient ICD10 codes to Phecodes and then cluster related Phecodes into diagnostic categories (severe type 2 allergy; autoimmune disease excluding cytopenias; rheumatologic disease; autoinflammatory disease including IBD; lymphoproliferative disease; cytopenias; immune deficiency). Patients were considered to have multiple comorbid severe immunologic diagnoses if they had either multiple Phecodes in one category or at least one Phecode in multiple categories. Because multisystem Phecode deconvolution for primary immune disorders has not been validated for enrichment studies, a burden analysis was not performed. However, an enrichment analysis was performed for individual common ICD10 diagnoses found in APDS patients within the categories of gastrointestinal disease, respiratory tract disease, lymphoproliferation, eczema, and severe, recurrent, or herpesviral infections. An in-house pipeline was constructed to select age, sex, and ethnicity matched controls at a ratio of 15:1 (N = 123 cases; N = 1,845 controls) and a case-control analysis was used to understand the increased burden of these diagnoses in patients carrying GOF variants (Tables S1 and S5).

UK Biobank cohort

Summary statistics provided by Wang et al. Nature 2021⁶⁴ at www.azphewas.com were pulled for all diagnosis enrichment in PIK3CD p.M285T carriers and controls—which was the only GOF variant found in the summary statistics data. APDS/IEI related diagnoses were highlighted. A Bonferroni correction for comparisons was not possible as the summary statistics only include enrichment burdens with P < 0.05. The genome/phenome-wide significance cutoff for that study was P < 10⁻⁸.

QUANTIFICATION AND STATISTICAL ANALYSIS

For screens, statistics (LFCs and two-sided p values for each donor, and for all N = 3 donors combined) were calculated using MAGeCK’s ‘test’ function. Z-scores for the LFC of each sgRNA were calculated against the distribution of LFCs of negative control empty window sgRNAs. Levene’s test for homogeneity of variance was used for correlative analyses of screen results with AlphaFold and AlphaMissense data. Statistics for variant-level effects were calculated with Wald’s test. For assessment of effects of different categories of sgRNAs in the screen, Kruskal-Wallis test with Dunn’s test for multiple comparisons was used. For all arrayed flow cytometry experiments, one-way ANOVA with Dunnett’s test, Bonferroni’s test, or Tukey’s test for multiple comparisons were used as indicated. N = 3 biological replicates were performed for all experiments unless otherwise indicated and error bars represent mean +/− standard deviation unless otherwise indicated. Statistics for All of Us database queries were performed with chi-squared tests on age, sex, and ethnicity matched control cohorts, and Bonferroni’s correction was applied to account for all ICD-10 codes tested (N = 39). All statistics were performed in Prism 10 (GraphPad), R, or Python.

Supplementary Material

Supplementary Table 6

NIHMS2110714-supplement-Supplementary_Table_6.xlsx^{(10.9KB, xlsx)}

Supplementary Table 7

NIHMS2110714-supplement-Supplementary_Table_7.xlsx^{(8.6KB, xlsx)}

Supplementary Table 4

NIHMS2110714-supplement-Supplementary_Table_4.xlsx^{(15.5KB, xlsx)}

Supplementary Table 5

NIHMS2110714-supplement-Supplementary_Table_5.xlsx^{(25.8KB, xlsx)}

Supplementary Table 3

NIHMS2110714-supplement-Supplementary_Table_3.xlsx^{(16.6KB, xlsx)}

Supplementary Table 1

NIHMS2110714-supplement-Supplementary_Table_1.xlsx^{(485.3KB, xlsx)}

Supplementary Table 2

NIHMS2110714-supplement-Supplementary_Table_2.xlsx^{(5.2MB, xlsx)}

Supplementary Table 8

NIHMS2110714-supplement-Supplementary_Table_8.xlsx^{(9.7KB, xlsx)}

NIHMS2110714-supplement-9.pdf^{(8.3MB, pdf)}

SUPPLEMENTAL INFORMATION

Supplemental information can be found online at https://doi.org/10.1016/j.cell.2025.05.037.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies

Mouse Monoclonal Anti-CD3 (NA/LE)	BD Biosciences	Cat# 555336
Mouse Monoclonal Anti-CD28 (NA/LE)	BD Biosciences	Cat# 555275
Goat Polyclonal Anti-Mouse Ig	BD Biosciences	Cat# 553998
Mouse Monoclonal Anti-Akt (pS473), PE-CF594	BD Biosciences	Cat# 562465
Mouse Monoclonal Anti-S6 (pS235/pS236), AF488	BD Biosciences	Cat# 560434
Mouse Monoclonal Anti-AKT (pT308), PE	BD Biosciences	Cat# 558275
Mouse Monoclonal Anti-CD8, BV421	BD Biosciences	Cat# 568217
Mouse Monoclonal Anti-CD4, BV480	BD Biosciences	Cat# 566104
Monoclonal Anti-TOX, APC	Miltenyi Biotec	Cat# 130–118-335
Rabbit Monoclonal Anti-TCF1/TCF7, AF488	Cell Signaling	Cat# 6444S
Mouse Monoclonal Anti-PD1, PE	Invitrogen	Cat# 12–9969-42
Mouse Monoclonal Anti-CTLA4, PE/Cy7	Biolegend	Cat# 349914
Mouse Monoclonal Anti-Ki67, BUV737	BD Biosciences	Cat# 567130
Mouse Monoclonal Anti-CD69, BUV563	BD Biosciences	Cat# 748764
Mouse Monoclonal Anti-CD71, PE/Cy7	Biolegend	Cat# 334112
Normal Mouse IgG	Santa Cruz	Cat# SC-2025

Bacterial and virus strains

OneShot Stbl3 Chemically Competent E. coli	Invitrogen	Cat# C737303
Endura Electrocompetent Cells	Lucigen	Cat# 602421

Biological samples

Healthy Human Buffy Coats	New York Blood Center	https://www.nybce.org/blood-products-and-services/research-products/
Human APDS Patient Sample PBMCs	This paper	N/A

Chemicals, peptides, and recombinant proteins

Leniolisib (CDZ 173)	SelleckChem	Cat# S8752
Rapamycin	SelleckChem	Cat# S1039
Everolimus	SelleckChem	Cat# S1120
Recombinant Human IL-2	Chiron	Cat# 53905–991-01
DMSO	Thermo Scientific	Cat# AA36480K2
Ficoll-Paque Plus, 1.077 g/mL	Cytiva	Cat# 17144002
ACK Lysis Buffer	Gibco	Cat# A1049201
Zombie NIR	Biolegend	Cat# 423105
KAPA HiFi HotStart ReadyMix	Roche	Cat# KK2601
rSAP	NEB	Cat# M0371S
BsmBI v2	NEB	Cat# R0739S
DTT	Thermo Fisher	Cat# R0861
T7 Ligase	NEB	Cat# M0318S
ATP	NEB	Cat# P0756S
Carbenicillin	Gibco	Cat# 10–177-012
TransIT-LT1	MirusBio	Cat# MIR2304
Lenti-X concentrator	Alstem Bio	Cat# 631232
LentiBOOST Transduction Enhancer	Mayflower Biosciences	Cat# SBPLV10112
Puromycin	Gibco	Cat# A1113803
KOD Xtreme Hot Start DNA Polymerase	MilliporeSigma	Cat# 71975–3
NEB Gibson Assembly Master Mix	NEB	Cat# E2611L
Q5 Hot Start HiFi DNA Polymerase	NEB	Cat# M0493S
N-1-methyl-pseudouridine-5’-triphosphate	TriLink Biotechnologies	Cat# N-1081–1
CleanCap Reagent AG	TriLink Biotechnologies	Cat# N-7113–1
Lithium Chloride	Invitrogen	Cat# AM9480
THE RNA Storage Solution	Thermo Fisher	Cat# AM7000
MaxCyte Electroporation Buffer	MaxCyte	Cat# EPB-1
Brilliant Stain Buffer Plus	BD Biosciences	Cat# 566385
eBioscience FoxP3/Transcription Factor Staining Buffer Set	Invitrogen	Cat# 00–5523-00
BD Cytofix Buffer	BD Biosciences	Cat# 554655
BD Perm Buffer III	BD Biosciences	Cat# 558050
ExTaq Polymerase	Takara Bio	Cat# RR001A
NG-ABE8e mRNA	This study	N/A
NG-ABE9 mRNA	This study	N/A

Critical commercial assays

EasySep human T cell isolation kit	StemCell	Cat# 17951
Immunocult CD3/CD28/CD2 T cell activator	StemCell	Cat# 10990
RNA TapeStation Kit	Agilent	Cat# 5067–5579
D5000 TapeStation Kit	Agilent	Cat# 5067–5588
HiScribe T7 High-Yield RNA Synthesis Kit	NEB	Cat# E2040S
Lonza P3 Primary Cell Nucleofection Kit	Lonza	Cat# V4XP-3032
Qiagen DNA Blood and Tissue Kit	Qiagen	Cat# 69504
CloudBreak Freestyle 2×75 High Output Sequencing Kit	Element Biosciences	Cat# 860–00015
Oxford Nanopore fragmentation-free long read sequencing with v14 library prep chemistry	Plasmidsaurus	https://plasmidsaurus.com/
Nextera XT DNA Library Preparation Kit	Illumina	Cat# FC-131–1024

Deposited data

MAGeCK outputs for all screens	This paper	Table S2
IEI cohort clinical data	This paper	Table S4
All of Us cohort clinical data	This paper	Table S5
Raw and processed sequencing data	This paper	GEO: GSE298853
Source data giving rise to figures	This paper, Dataverse	https://doi.org/10.7910/DVN/92WQLH
Code giving rise to figures	This paper	https://github.com/cf3041/genetic-variants-IEI

Experimental models: Cell lines

Human: HEK293T	ATCC	Cat# CRL-3216

Oligonucleotides

PIK3CD/PIK3R1 ABE tiling sgRNA library	This paper	Table S1
sgRNAs used in arrayed validation experiments	This paper, Synthego	Table S6
Custom Primers used in this study	This paper	Table S8
DNA/RNA UD Indexes Set A, Tagmentation Primers	Illumina	Cat# 20091654

Recombinant DNA

Plasmid: pMD2.G	Laboratory of Didier Trono	Addgene #12259
Plasmid: psPAX2	Laboratory of Didier Trono	Addgene #12260
Plasmid: CROPseq-mTurquoise lentiviral vector	This paper	N/A
Plasmid: NG-ABE8e IVT template vector	This paper	N/A
Plasmid: NG-ABE9e IVT template vector	This paper	N/A

Software and algorithms

FlowJo v10.8.1	BD Biosciences	https://www.flowjo.com
Prism v9.5.0	GraphPad	https://www.graphpad.com
MAGeCK v0.5.9.2	Li et al.⁷⁵	https://sourceforge.net/p/mageck/wiki/Home/
CutAdapt v5.0	Martin et al.⁷⁶	https://doi.org/10.14806/ej.17.1.200
SnapGene v6.0.5	Dotmatics	https://www.snapgene.com
Geneious Prime v2025.1	Dotmatics	https://www.geneious.com/
R v4.1.1	The R Project	https://www.r-project.org
ImageJ	Schneider et al.⁷⁷	https://www.imagej.net
Genomics 2 Proteins Portal (GPP)	Kwon et al.⁷⁸	https://g2p.broadinstitute.org/
AlphaFold	Google DeepMind; Jumper et al.⁵²	https://alphafold.ebi.ac.uk/
Cas-OFFinder v2.4.1	Bae et al.⁷⁹	https://github.com/snugel/cas-offinder
Cutting Frequency Determination (CFD)	Doench et al.⁸⁰	PMCID: PMC4744125
UCSF ChimeraX	Pettersen et al.⁸¹	https://www.cgl.ucsf.edu/chimerax/
Base Editor Design Tool v2.0	Hanna et al.⁷³	https://github.com/mhegde/base-editor-design-tool
SynergyFinder v3.0	Ianevski et al.⁶¹	https://synergyfinder.fimm.fi/
All of Us Researcher Workbench	National Institutes of Health	https://www.researchallofus.org/data-tools/workbench/
ICD10>Phecode>Phecode Rollup Pipeline	This study	Code repository.

Other

SPRIselect beads	Beckman Coulter	Cat# B23318
“The Big Easy” EasySep Magnet	StemCell	Cat# 18001
G-Rex 6-well culture plates	Wilson Wolf	Cat# 80240M
1 mM electroporation cuvettes	Biorad	Cat# 1652089
BioRad Gene Pulser Xcell	Biorad	Cat# 1652660
0.45 μM PES filter	Fisher Scientific	Cat# 09–720-514
SepMate 50 mL conical tubes	StemCell	Cat# 85450
MaxCyte 50×3 Processing Assembly	MaxCyte	Cat# ER050U3–10
MaxCyte GTX Instrument	MaxCyte	Cat# GTX
Lonza 4D Nucleofector X Unit	Lonza	Cat# AAF-1003X

Open in a new tab

Highlights.

Base-editing screens of PIK3CD and PIK3R1 support classification of >100 variants
Engineered cells mimic pathobiology and drug responses of T cells from APDS patients
Features of leniolisib response/resistance and new drug combinations are identified
Population data suggest APDS may be more prevalent than previously estimated

ACKNOWLEDGMENTS

Medical illustrations were prepared by Dr. Uta Mackensen. Elements of the Graphical Abstract and Figure 7A were created with BioRender.com. We thank Mike Kissner and the CSCI Flow Cytometry Core for support with FACS experiments. We gratefully acknowledge All of Us participants for their contributions, without whom this research would not have been possible. We also thank the National Institutes of Health’s All of Us Research Program for making available the participant data examined in this study. This work was supported by the National Institutes of Health (NIH), National Cancer Institute (NCI) grants R37CA258829, R01CA266446, R01CA280414, and U54CA274506 to B.I. and F30CA298572 to Z.H.W., and additional support by the Burroughs Wellcome Fund Career Award for Medical Scientists, a Velocity Fellows Award, the Louis V. Gerstner, Jr. Scholars Program, a Tara Miller Young Investigator Award by the Melanoma Research Alliance (MRA), a Tara Miller Team Science Award for Brain Metastasis Research by the MRA, and the Pershing Square Sohn Cancer Research Alliance Award to B.I. and a Melanoma Research Foundation (MRF) Medical Student Award to Z.H.W. B.I. is a CRI Lloyd J. Old STAR (CRI5579). This work was additionally supported by the Herbert Irving Comprehensive Cancer Center (HICCC) Human Tissue Immunology and Immunotherapy Initiative to B.I. This work was supported by NIH/NCI Cancer Center Support grant P30CA013696. This work was in part supported through a sponsored research agreement with Pharming to J.D.M..

DECLARATION OF INTERESTS

B.I. and J.D.M. received research support from Pharming. B.I. is a consultant for or has received honoraria from Volastra Therapeutics, Johnson & Johnson/Janssen, Novartis, GSK, EISAI, AstraZeneca, and Merck and has received research funding to Columbia University from Agenus, Alkermes, Arcus Biosciences, Checkmate Pharmaceuticals, Compugen, Immunocore, Regeneron, and Synthekine. J.D.M. is on the scientific advisory board for Blueprint Medicine and receives grant funding from Pharming.

REFERENCES

1.von Hardenberg S, Klefenz I, Steinemann D, Di Donato N, Baumann U, Auber B, and Klemann C (2024). Current genetic diagnostics in inborn errors of immunity. Front. Pediatr 12, 1279112. 10.3389/fped.2024.1279112. [DOI] [PMC free article] [PubMed] [Google Scholar]
2.Al-Mousa H, Abouelhoda M, Monies DM, Al-Tassan N, Al-Ghonaium A, Al-Saud B, Al-Dhekri H, Arnaout R, Al-Muhsen S, Ades N, et al. (2016). Unbiased targeted next-generation sequencing molecular approach for primary immunodeficiency diseases. J. Allergy Clin. Immunol 137, 1780–1787. 10.1016/j.jaci.2015.12.1310. [DOI] [PubMed] [Google Scholar]
3.Gallo V (2016). Diagnostics of primary immunodeficiencies through next generation sequencing. 10.26226/morressier.57bc1758d462b80290b4d256. [DOI] [PMC free article] [PubMed] [Google Scholar]
4.Meyts I, Bosch B, Bolze A, Boisson B, Itan Y, Belkadi A, Pedergnana V, Moens L, Picard C, Cobat A, et al. (2016). Exome and genome sequencing for inborn errors of immunity. J. Allergy Clin. Immunol 138, 957–969. 10.1016/j.jaci.2016.08.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Cifaldi C, Brigida I, Barzaghi F, Zoccolillo M, Ferradini V, Petricone D, Cicalese MP, Lazarevic D, Cittaro D, Omrani M, et al. (2019). Targeted NGS platforms for genetic screening and gene discovery in primary Immunodeficiencies. Front. Immunol 10, 316. 10.3389/fimmu.2019.00316. [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Nijman IJ, van Montfrans JM, Hoogstraat M, Boes ML, van de Corput L, Renner ED, van Zon P, van Lieshout S, Elferink MG, van der Burg M, et al. (2014). Targeted next-generation sequencing: a novel diagnostic tool for primary immunodeficiencies. J. Allergy Clin. Immunol 133, 529–534. 10.1016/j.jaci.2013.08.032. [DOI] [PubMed] [Google Scholar]
7.Landrum MJ, Lee JM, Riley GR, Jang W, Rubinstein WS, Church DM, and Maglott DR (2014). ClinVar: public archive of relationships among sequence variation and human phenotype. Nucleic Acids Res. 42, D980–D985. 10.1093/nar/gkt1113. [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Michalovich D, and Nejentsev S (2018). Activated PI3 kinase delta syndrome: From genetics to therapy. Front. Immunol 9, 369. 10.3389/fimmu.2018.00369. [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Lucas CL, Chandra A, Nejentsev S, Condliffe AM, and Okkenhaug K (2016). PI3Kδ and primary immunodeficiencies. Nat. Rev. Immunol 16, 702–714. 10.1038/nri.2016.93. [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Angulo I, Vadas O, Garçon F, Banham-Hall E, Plagnol V, Leahy TR, Baxendale H, Coulter T, Curtis J, Wu C, et al. (2013). Phosphoinositide 3-kinase δ gene mutation predisposes to respiratory infection and airway damage. Science 342, 866–871. 10.1126/science.1243292. [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Deau M-C, Heurtier L, Frange P, Suarez F, Bole-Feysot C, Nitschke P, Cavazzana M, Picard C, Durandy A, Fischer A, et al. (2015). A human immunodeficiency caused by mutations in the PIK3R1 gene. J. Clin. Invest 125, 1764–1765. 10.1172/JCI81746. [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Lucas CL, Kuehn HS, Zhao F, Niemela JE, Deenick EK, Palendira U, Avery DT, Moens L, Cannons JL, Biancalana M, et al. (2014). Dominant-activating germline mutations in the gene encoding the PI(3)K catalytic subunit p110δ result in T cell senescence and human immunodeficiency. Nat. Immunol 15, 88–97. 10.1038/ni.2771. [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Coulter TI, Chandra A, Bacon CM, Babar J, Curtis J, Screaton N, Goodlad JR, Farmer G, Steele CL, Leahy TR, et al. (2017). Clinical spectrum and features of activated phosphoinositide 3-kinase δ syndrome: A large patient cohort study. J. Allergy Clin. Immunol 139, 597–606.e4. 10.1016/j.jaci.2016.06.021. [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Elkaim E, Neven B, Bruneau J, Mitsui-Sekinaka K, Stanislas A, Heurtier L, Lucas CL, Matthews H, Deau M-C, Sharapova S, et al. (2016). Clinical and immunologic phenotype associated with activated phosphoinositide 3-kinase δ syndrome 2: A cohort study. J. Allergy Clin. Immunol 138, 210–218.e9. 10.1016/j.jaci.2016.03.022. [DOI] [PubMed] [Google Scholar]
15.Preite S, Cannons JL, Radtke AJ, Vujkovic-Cvijin I, Gomez-Rodriguez J, Volpi S, Huang B, Cheng J, Collins N, Reilley J, et al. (2018). Hyperactivated PI3Kδ promotes self and commensal reactivity at the expense of optimal humoral immunity. Nat. Immunol 19, 986–1000. 10.1038/s41590-018-0182-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Kracker S, Curtis J, Ibrahim MAA, Sediva A, Salisbury J, Campr V, Debré M, Edgar JDM, Imai K, Picard C, et al. (2014). Occurrence of B-cell lymphomas in patients with activated phosphoinositide 3-kinase δ syndrome. J. Allergy Clin. Immunol 134, 233–236. 10.1016/j.jaci.2014.02.020. [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Oh J, Garabedian E, Fuleihan R, and Cunningham-Rundles C (2021). Clinical manifestations and outcomes of activated phosphoinositide 3-kinase δ syndrome from the USIDNET cohort. J. Allergy Clin. Immunol. Pract 9, 4095–4102. 10.1016/j.jaip.2021.07.044. [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Wentink MWJ, Mueller YM, Dalm VASH, Driessen GJ, van Hagen PM, van Montfrans JM, van der Burg M, and Katsikis PD (2018). Exhaustion of the CD8+ T cell compartment in patients with mutations in phosphoinositide 3-kinase delta. Front. Immunol 9, 446. 10.3389/fimmu.2018.00446. [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Jia Y, Yang Q, Wang Y, Li W, Chen X, Xu T, Tian Z, Feng M, Zhang L, Tang W, et al. (2021). Hyperactive PI3Kδ predisposes naive T cells to activation via aerobic glycolysis programs. Cell. Mol. Immunol 18, 1783–1797. 10.1038/s41423-020-0379-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Ruiz-García R, Vargas-Hernández A, Chinn IK, Angelo LS, Cao TN, Coban-Akdemir Z, Jhangiani SN, Meng Q, Forbes LR, Muzny DM, et al. (2018). Mutations in PI3K110δ cause impaired natural killer cell function partially rescued by rapamycin treatment. J. Allergy Clin. Immunol 142, 605–617.e7. 10.1016/j.jaci.2017.11.042. [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Cannons JL, Preite S, Kapnick SM, Uzel G, and Schwartzberg PL (2018). Genetic defects in phosphoinositide 3-kinase δ influence CD8+ T cell survival, differentiation, and function. Front. Immunol 9, 1758. 10.3389/fimmu.2018.01758. [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Edwards ESJ, Bier J, Cole TS, Wong M, Hsu P, Berglund LJ, Boztug K, Lau A, Gostick E, Price DA, et al. (2019). Activating PIK3CD mutations impair human cytotoxic lymphocyte differentiation and function and EBV immunity. J. Allergy Clin. Immunol 143, 276–291. e6. 10.1016/j.jaci.2018.04.030. [DOI] [PubMed] [Google Scholar]
23.Bier J, Rao G, Payne K, Brigden H, French E, Pelham SJ, Lau A, Lenthall H, Edwards ESJ, Smart JM, et al. (2019). Activating mutations in PIK3CD disrupt the differentiation and function of human and murine CD4+ T cells. J. Allergy Clin. Immunol 144, 236–253. 10.1016/j.jaci.2019.01.033. [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Meyts I, Bousfiha A, Duff C, Singh S, Lau YL, Condino-Neto A, Bezrodnik L, Ali A, Adeli M, and Drabwell J (2020). Primary immunodeficiencies: A decade of progress and a promising future. Front. Immunol 11, 625753. 10.3389/fimmu.2020.625753. [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Sogkas G, Fedchenko M, Dhingra A, Jablonka A, Schmidt RE, and Atschekzei F (2018). Primary immunodeficiency disorder caused by phosphoinositide 3–kinase δ deficiency. J. Allergy Clin. Immunol 142, 1650–1653.e2. 10.1016/j.jaci.2018.06.039. [DOI] [PubMed] [Google Scholar]
26.Swan DJ, Aschenbrenner D, Lamb CA, Chakraborty K, Clark J, Pandey S, Engelhardt KR, Chen R, Cavounidis A, Ding Y, et al. (2019). Immunodeficiency, autoimmune thrombocytopenia and enterocolitis caused by autosomal recessive deficiency of PIK3CD-encoded phosphoinositide 3-kinase δ. Haematologica 104, e483–e486. 10.3324/haematol.2018.208397. [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Dyment DA, Smith AC, Alcantara D, Schwartzentruber JA, Basel-Vanagaite L, Curry CJ, Temple IK, Reardon W, Mansour S, Haq MR, et al. (2013). Mutations in PIK3R1 cause SHORT syndrome. Am. J. Hum. Genet 93, 158–166. 10.1016/j.ajhg.2013.06.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Tomlinson PR, Knox R, Perisic O, Su HC, Brierley GV, Williams RL, and Semple RK (2024). Paradoxical dominant negative activity of an immunodeficiency-associated activating PIK3R1 variantPreprint as. bioRxiv. 10.1101/2023.11.02.565250. [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Chudasama KK, Winnay J, Johansson S, Claudi T, König R, Haldorsen I, Johansson B, Woo JR, Aarskog D, Sagen JV, et al. (2013). SHORT syndrome with partial lipodystrophy due to impaired phosphatidylinositol 3 kinase signaling. Am. J. Hum. Genet 93, 150–157. 10.1016/j.ajhg.2013.05.023. [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Avila M, Dyment DA, Sagen JV, St-Onge J, Moog U, Chung BHY, Mansour S, Albanese A, Garcia S, Ortiz Martin D, et al. (2015). Clinical reappraisal of SHORT syndrome with PIK3R1 mutations: towards recommendation for molecular testing and management. Clin. Genet 89, 501–506. [DOI] [PubMed] [Google Scholar]
31.Bravo García-Morato M, García-Miñaúr S, Molina Garicano J, Santos Simarro F, Del Pino Molina L, López-Granados E, Ferreira Cerdán A, and Rodríguez Pena R (2017). Mutations in PIK3R1 can lead to APDS2, SHORT syndrome or a combination of the two. Clin. Immunol 179, 77–80. 10.1016/j.clim.2017.03.004. [DOI] [PubMed] [Google Scholar]
32.Elgizouli M, Lowe DM, Speckmann C, Schubert D, Hülsdünker J, Eskandarian Z, Dudek A, Schmitt-Graeff A, Wanders J, Jørgensen SF, et al. (2016). Activating PI3Kδ mutations in a cohort of 669 patients with primary immunodeficiency. Clin. Exp. Immunol 183, 221–229. 10.1111/cei.12706. [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Takeda AJ, Zhang Y, Dornan GL, Siempelkamp BD, Jenkins ML, Matthews HF, McElwee JJ, Bi W, Seeborg FO, Su HC, et al. (2017). Novel PIK3CD mutations affecting N-terminal residues of p110δ cause activated PI3Kδ syndrome (APDS) in humans. J. Allergy Clin. Immunol 140, 1152–1156.e10. 10.1016/j.jaci.2017.03.026. [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Rao VK, Webster S, Dalm VASH, Šedivá A, van Hagen PM, Holland S, Rosenzweig SD, Christ AD, Sloth B, Cabanski M, et al. (2017). Effective “activated PI3Kδ syndrome”-targeted therapy with the PI3Kδ inhibitor leniolisib. Blood 130, 2307–2316. 10.1182/blood-2017-08-801191. [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Hoegenauer K, Soldermann N, Zécri F, Strang RS, Graveleau N, Wolf RM, Cooke NG, Smith AB, Hollingworth GJ, Blanz J, et al. (2017). Discovery of CDZ173 (leniolisib), representing a structurally novel class of PI3K delta-selective inhibitors. ACS Med. Chem. Lett 8, 975–980. 10.1021/acsmedchemlett.7b00293. [DOI] [PMC free article] [PubMed] [Google Scholar]
36.Rao VK, Kulm E, Grossman J, Buchbinder D, Chong H, Bradt J, Webster S, Šedivá A, Dalm VA, and Uzel G (2024). Long-term treatment with selective PI3Kδ inhibitor leniolisib in adults with activated PI3Kδ syndrome. Blood Adv. 8, 3092–3108. 10.1182/bloodadvances.2023011000. [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Rao VK, Kulm E, Šedivá A, Plebani A, Schuetz C, Shcherbina A, Dalm VA, Trizzino A, Zharankova Y, Webster S, et al. (2024). Interim analysis: Open-label extension study of leniolisib for patients with APDS. J. Allergy Clin. Immunol 153, 265–274.e9. 10.1016/j.jaci.2023.09.032. [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Rao VK, Webster S, Šedivá A, Plebani A, Schuetz C, Shcherbina A, Conlon N, Coulter T, Dalm VA, Trizzino A, et al. (2023). A randomized, placebo-controlled phase 3 trial of the PI3Kδ inhibitor leniolisib for activated PI3Kδ syndrome. Blood 141, 971–983. 10.1182/blood.2022018546. [DOI] [PMC free article] [PubMed] [Google Scholar]
39.Gaudelli NM, Komor AC, Rees HA, Packer MS, Badran AH, Bryson DI, and Liu DR (2017). Programmable base editing of A⋅T to G⋅C in genomic DNA without DNA cleavage. Nature 551, 464–471. 10.1038/nature24644. [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Anzalone AV, Koblan LW, and Liu DR (2020). Genome editing with CRISPR-Cas nucleases, base editors, transposases and prime editors. Nat. Biotechnol 38, 824–844. 10.1038/s41587-020-0561-9. [DOI] [PubMed] [Google Scholar]
41.Walsh ZH, Shah P, Kothapalli N, Shah SB, Nikolenyi G, Brodtman DZ, Leuzzi G, Rogava M, Mu M, Ho P, et al. (2025). Mapping variant effects on anti-tumor hallmarks of primary human T cells with base-editing screens. Nat. Biotechnol 43, 384–395. 10.1038/s41587-024-02235-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
42.Richter MF, Zhao KT, Eton E, Lapinaite A, Newby GA, Thuronyi BW, Wilson C, Koblan LW, Zeng J, Bauer DE, et al. (2020). Phage-assisted evolution of an adenine base editor with improved Cas domain compatibility and activity. Nat. Biotechnol 38, 883–891. 10.1038/s41587-020-0453-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
43.Kim E, and Hart T (2021). Improved analysis of CRISPR fitness screens and reduced off-target effects with the BAGEL2 gene essentiality classifier. Genome Med. 13, 2. 10.1186/s13073-020-00809-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
44.Deau M-C, Heurtier L, Frange P, Suarez F, Bole-Feysot C, Nitschke P, Cavazzana M, Picard C, Durandy A, Fischer A, et al. (2014). A human immunodeficiency caused by mutations in the PIK3R1 gene. J. Clin. Invest 124, 3923–3928. 10.1172/JCI75746. [DOI] [PMC free article] [PubMed] [Google Scholar]
45.Nguyen T, Lau A, Bier J, Cooke KC, Lenthall H, Ruiz-Diaz S, Avery DT, Brigden H, Zahra D, Sewell WA, et al. (2023). Human PIK3R1 mutations disrupt lymphocyte differentiation to cause activated PI3Kδ syndrome 2. J. Exp. Med 220, e20221020. 10.1084/jem.20221020. [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Zhao P, Huang J, Fu H, Xu J, Li T, Zhang X, Meng Q, Zhang L, Tan L, Zhang W, et al. (2024). Activated phosphoinositide 3-kinase δ syndrome caused by PIK3CD mutations: expanding the phenotype. Pediatr. Rheumatol. Online J 22, 24. 10.1186/s12969-024-00955-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
47.Brnich SE, Abou Tayoun AN, Couch FJ, Cutting GR, Greenblatt MS, Heinen CD, Kanavy DM, Luo X, McNulty SM, Starita LM, et al. (2019). Recommendations for application of the functional evidence PS3/BS3 criterion using the ACMG/AMP sequence variant interpretation framework. Genome Med. 12, 3. 10.1186/s13073-019-0690-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
48.Gelman H, Dines JN, Berg J, Berger AH, Brnich S, Hisama FM, James RG, Rubin AF, Shendure J, Shirts B, et al. (2019). Recommendations for the collection and use of multiplexed functional data for clinical variant interpretation. Genome Med. 11, 85. 10.1186/s13073-019-0698-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
49.Bajaj K, Madhusudhan MS, Adkar BV, Chakrabarti P, Ramakrishnan C, Sali A, and Varadarajan R (2007). Stereochemical criteria for prediction of the effects of proline mutations on protein stability. PLoS Comput. Biol 3, e241. 10.1371/journal.pcbi.0030241. [DOI] [PMC free article] [PubMed] [Google Scholar]
50.Fruman DA, Chiu H, Hopkins BD, Bagrodia S, Cantley LC, and Abraham RT (2017). The PI3K pathway in human disease. Cell 170, 605–635. 10.1016/j.cell.2017.07.029. [DOI] [PMC free article] [PubMed] [Google Scholar]
51.Heffron TP, Heald RA, Ndubaku C, Wei B, Augistin M, Do S, Edgar K, Eigenbrot C, Friedman L, Gancia E, et al. (2016). The rational design of selective benzoxazepin inhibitors of the α-isoform of phosphoinositide 3-kinase culminating in the identification of (S)-2-((2-(1-isopropyl-1H-1,2,4-triazol-5-yl)-5,6-dihydrobenzo[f]imidazo[1,2-d][1,4]oxazepin-9-yl)oxy)propanamide (GDC-0326). J. Med. Chem 59, 985–1002. 10.1021/acs.jmedchem.5b01483. [DOI] [PubMed] [Google Scholar]
52.Jumper J, Evans R, Pritzel A, Green T, Figurnov M, Ronneberger O, Tunyasuvunakool K, Bates R, Žídek A, Potapenko A, et al. (2021). Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589. 10.1038/s41586-021-03819-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
53.Cheng J, Novati G, Pan J, Bycroft C, Žemgulytė A, Applebaum T, Pritzel A, Wong LH, Zielinski M, Sargeant T, et al. (2023). Accurate proteome-wide missense variant effect prediction with AlphaMissense. Science 381, eadg7492. 10.1126/science.adg7492. [DOI] [PubMed] [Google Scholar]
54.Schmidt R, Ward CC, Dajani R, Armour-Garb Z, Ota M, Allain V, Hernandez R, Layeghi M, Xing G, Goudy L, et al. (2024). Base-editing mutagenesis maps alleles to tune human T cell functions. Nature 625, 805–812. 10.1038/s41586-023-06835-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
55.Neugebauer ME, Hsu A, Arbab M, Krasnow NA, McElroy AN, Pandey S, Doman JL, Huang TP, Raguram A, Banskota S, et al. (2023). Evolution of an adenine base editor into a small, efficient cytosine base editor with low off-target activity. Nat. Biotechnol 41, 673–685. 10.1038/s41587-022-01533-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
56.Chen L, Zhang S, Xue N, Hong M, Zhang X, Zhang D, Yang J, Bai S, Huang Y, Meng H, et al. (2023). Engineering a precise adenine base editor with minimal bystander editing. Nat. Chem. Biol 19, 101–110. 10.1038/s41589-022-01163-8. [DOI] [PubMed] [Google Scholar]
57.Rao VK, Šedivá A, Dalm VASH, Plebani A, Schuetz C, Shcherbina A, Trizzino A, Zharankova Y, Orpia A, Kulm E, et al. (2025). A randomised, placebo-controlled, phase III trial of leniolisib in activated phosphoinositide 3-kinase delta (PI3Kδ) syndrome (APDS): Adolescent and adult subgroup analysis. Clin. Immunol 270, 110400. 10.1016/j.clim.2024.110400. [DOI] [PMC free article] [PubMed] [Google Scholar]
58.Khan O, Giles JR, McDonald S, Manne S, Ngiow SF, Patel KP, Werner MT, Huang AC, Alexander KA, Wu JE, et al. (2019). TOX transcriptionally and epigenetically programs CD8+ T cell exhaustion. Nature 571, 211–218. 10.1038/s41586-019-1325-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
59.Wu T, Ji Y, Moseman EA, Xu HC, Manglani M, Kirby M, Anderson SM, Handon R, Kenyon E, Elkahloun A, et al. (2016). The TCF1-Bcl6 axis counteracts type I interferon to repress exhaustion and maintain T cell stemness. Sci. Immunol 1, eaai8593. 10.1126/sciimmunol.aai8593. [DOI] [PMC free article] [PubMed] [Google Scholar]
60.Zheng Y, Collins SL, Lutz MA, Allen AN, Kole TP, Zarek PE, and Powell JD (2007). A role for mammalian target of rapamycin in regulating T cell activation versus anergy. J. Immunol 178, 2163–2170. 10.4049/jimmunol.178.4.2163. [DOI] [PubMed] [Google Scholar]
61.Ianevski A, Giri AK, and Aittokallio T (2022). SynergyFinder 3.0: an interactive analysis and consensus interpretation of multi-drug synergies across multiple samples. Nucleic Acids Res. 50, W739–W743. 10.1093/nar/gkac382. [DOI] [PMC free article] [PubMed] [Google Scholar]
62.Vanselow S, Wahn V, and Schuetz C (2023). Activated PI3Kδ syndrome - reviewing challenges in diagnosis and treatment. Front. Immunol 14, 1208567. 10.3389/fimmu.2023.1208567. [DOI] [PMC free article] [PubMed] [Google Scholar]
63.Richards S, Aziz N, Bale S, Bick D, Das S, Gastier-Foster J, Grody WW, Hegde M, Lyon E, Spector E, et al. (2015). Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. Genet. Med 17, 405–424. 10.1038/gim.2015.30. [DOI] [PMC free article] [PubMed] [Google Scholar]
64.Wang Q, Dhindsa RS, Carss K, Harper AR, Nag A, Tachmazidou I, Vitsios D, Deevi SVV, Mackay A, Muthas D, et al. (2021). Rare variant contribution to human disease in 281,104 UK Biobank exomes. Nature 597, 527–532. 10.1038/s41586-021-03855-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
65.Bousfiha A, Moundir A, Tangye SG, Picard C, Jeddane L, Al-Herz W, Rundles CC, Franco JL, Holland SM, Klein C, et al. (2022). The 2022 update of IUIS phenotypical classification for human Inborn Errors of immunity. J. Clin. Immunol 42, 1508–1520. 10.1007/s10875-022-01352-z. [DOI] [PubMed] [Google Scholar]
66.Human Inborn Errors of Immunity: 2024 Update on the Classification from the International Union of Immunological Societies Expert Committee. [Google Scholar]
67.Meitlis I, Allenspach EJ, Bauman BM, Phan IQ, Dabbah G, Schmitt EG, Camp ND, Torgerson TR, Nickerson DA, Bamshad MJ, et al. (2020). Multiplexed functional assessment of genetic variants in CARD11. Am. J. Hum. Genet 107, 1029–1043. 10.1016/j.ajhg.2020.10.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
68.Starita LM, Islam MM, Banerjee T, Adamovich AI, Gullingsrud J, Fields S, Shendure J, and Parvin JD (2018). A multiplex homology-directed DNA repair assay reveals the impact of more than 1,000 BRCA1 missense substitution variants on protein function. Am. J. Hum. Genet 103, 498–508. 10.1016/j.ajhg.2018.07.016. [DOI] [PMC free article] [PubMed] [Google Scholar]
69.Findlay GM, Daza RM, Martin B, Zhang MD, Leith AP, Gasperini M, Janizek JD, Huang X, Starita LM, and Shendure J (2018). Accurate classification of BRCA1 variants with saturation genome editing. Nature 562, 217–222. 10.1038/s41586-018-0461-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
70.Ma K, Huang S, Ng KK, Lake NJ, Joseph S, Xu J, Lek A, Ge L, Woodman KG, Koczwara KE, et al. (2024). Saturation mutagenesis-reinforced functional assays for disease-related genes. Cell 187, 6707–6724.e22. 10.1016/j.cell.2024.08.047. [DOI] [PMC free article] [PubMed] [Google Scholar]
71.Horesh ME, Martin-Fernandez M, Gruber C, Buta S, Le Voyer T, Puzenat E, Lesmana H, Wu Y, Richardson A, Stein D, et al. (2024). Individuals with JAK1 variants are affected by syndromic features encompassing autoimmunity, atopy, colitis, and dermatitis. J. Exp. Med 221, e20232387. 10.1084/jem.20232387. [DOI] [PMC free article] [PubMed] [Google Scholar]
72.Lue NZ, and Liau BB (2023). Base editor screens for in situ mutational scanning at scale. Mol. Cell 83, 2167–2187. 10.1016/j.mol-cel.2023.06.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
73.Hanna RE, Hegde M, Fagre CR, DeWeirdt PC, Sangree AK, Szegletes Z, Griffith A, Feeley MN, Sanson KR, Baidi Y, et al. (2021). Massively parallel assessment of human variants with base editor screens. Cell 184, 1064–1080.e20. 10.1016/j.cell.2021.01.012. [DOI] [PubMed] [Google Scholar]
74.Martinson AK, Chin AT, Butte MJ, and Rider NL (2024). Artificial intelligence and machine learning for inborn errors of immunity: Current state and future promise. J. Allergy Clin. Immunol. Pract 12, 2695–2704. 10.1016/j.jaip.2024.08.012. [DOI] [PubMed] [Google Scholar]
75.Li W, Xu H, Xiao T, Cong L, Love MI, Zhang F, Irizarry RA, Liu JS, Brown M, and Liu XS (2014). MAGeCK enables robust identification of essential genes from genome-scale CRISPR/Cas9 knockout screens. Genome Biol. 15, 554. 10.1186/s13059-014-0554-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
76.Martin M (2011). Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet J. 17, 10. 10.14806/ej.17.1.200. [DOI] [Google Scholar]
77.Schneider CA, Rasband WS, and Eliceiri KW (2012). NIH Image to ImageJ: 25 years of image analysis. Nat. Methods 9, 671–675. 10.1038/nmeth.2089. [DOI] [PMC free article] [PubMed] [Google Scholar]
78.Kwon S, Safer J, Nguyen DT, Hoksza D, May P, Arbesfeld JA, Rubin AF, Campbell AJ, Burgin A, and Iqbal S (2024). Genomics 2 Proteins portal: a resource and discovery tool for linking genetic screening outputs to protein sequences and structures. Nat. Methods 21, 1947–1957. 10.1038/s41592-024-02409-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
79.Bae S, Park J, and Kim J-S (2014). Cas-OFFinder: a fast and versatile algorithm that searches for potential off-target sites of Cas9 RNA-guided endonucleases. Bioinformatics 30, 1473–1475. 10.1093/bioinformatics/btu048. [DOI] [PMC free article] [PubMed] [Google Scholar]
80.Doench JG, Fusi N, Sullender M, Hegde M, Vaimberg EW, Donovan KF, Smith I, Tothova Z, Wilen C, Orchard R, et al. (2016). Optimized sgRNA design to maximize activity and minimize off-target effects of CRISPR-Cas9. Nat. Biotechnol 34, 184–191. 10.1038/nbt.3437. [DOI] [PMC free article] [PubMed] [Google Scholar]
81.Pettersen EF, Goddard TD, Huang CC, Meng EC, Couch GS, Croll TI, Morris JH, and Ferrin TE (2021). UCSF ChimeraX: Structure visualization for researchers, educators, and developers. Protein Sci. 30, 70–82. 10.1002/pro.3943. [DOI] [PMC free article] [PubMed] [Google Scholar]
82.Riesenberg S, Helmbrecht N, Kanis P, Maricic T, and Pääbo S (2022). Improved gRNA secondary structures allow editing of target sites resistant to CRISPR-Cas9 cleavage. Nat. Commun 13, 489. 10.1038/s41467-022-28137-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
83.Martin-Rufino JD, Castano N, Pang M, Grody EI, Joubran S, Caulier A, Wahlster L, Li T, Qiu X, Riera-Escandell AM, et al. (2023). Massively parallel base editing to map variant effects in human hematopoiesis. Cell 186, 2456–2474.e24. 10.1016/j.cell.2023.03.035. [DOI] [PMC free article] [PubMed] [Google Scholar]
84.Ye J, Coulouris G, Zaretskaya I, Cutcutache I, Rozen S, and Madden TL (2012). Primer-BLAST: a tool to design target-specific primers for polymerase chain reaction. BMC Bioinformatics 13, 134. 10.1186/1471-2105-13-134. [DOI] [PMC free article] [PubMed] [Google Scholar]
85.UniProt Consortium (2025). UniProt: The universal protein knowledgebase in 2025. Nucleic Acids Res. 53, D609–D617. 10.1093/nar/gkae1010. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Table 6

NIHMS2110714-supplement-Supplementary_Table_6.xlsx^{(10.9KB, xlsx)}

Supplementary Table 7

NIHMS2110714-supplement-Supplementary_Table_7.xlsx^{(8.6KB, xlsx)}

Supplementary Table 4

NIHMS2110714-supplement-Supplementary_Table_4.xlsx^{(15.5KB, xlsx)}

Supplementary Table 5

NIHMS2110714-supplement-Supplementary_Table_5.xlsx^{(25.8KB, xlsx)}

Supplementary Table 3

NIHMS2110714-supplement-Supplementary_Table_3.xlsx^{(16.6KB, xlsx)}

Supplementary Table 1

NIHMS2110714-supplement-Supplementary_Table_1.xlsx^{(485.3KB, xlsx)}

Supplementary Table 2

NIHMS2110714-supplement-Supplementary_Table_2.xlsx^{(5.2MB, xlsx)}

Supplementary Table 8

NIHMS2110714-supplement-Supplementary_Table_8.xlsx^{(9.7KB, xlsx)}

NIHMS2110714-supplement-9.pdf^{(8.3MB, pdf)}

Data Availability Statement

[R1] 1.von Hardenberg S, Klefenz I, Steinemann D, Di Donato N, Baumann U, Auber B, and Klemann C (2024). Current genetic diagnostics in inborn errors of immunity. Front. Pediatr 12, 1279112. 10.3389/fped.2024.1279112. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R2] 2.Al-Mousa H, Abouelhoda M, Monies DM, Al-Tassan N, Al-Ghonaium A, Al-Saud B, Al-Dhekri H, Arnaout R, Al-Muhsen S, Ades N, et al. (2016). Unbiased targeted next-generation sequencing molecular approach for primary immunodeficiency diseases. J. Allergy Clin. Immunol 137, 1780–1787. 10.1016/j.jaci.2015.12.1310. [DOI] [PubMed] [Google Scholar]

[R3] 3.Gallo V (2016). Diagnostics of primary immunodeficiencies through next generation sequencing. 10.26226/morressier.57bc1758d462b80290b4d256. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R4] 4.Meyts I, Bosch B, Bolze A, Boisson B, Itan Y, Belkadi A, Pedergnana V, Moens L, Picard C, Cobat A, et al. (2016). Exome and genome sequencing for inborn errors of immunity. J. Allergy Clin. Immunol 138, 957–969. 10.1016/j.jaci.2016.08.003. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R5] 5.Cifaldi C, Brigida I, Barzaghi F, Zoccolillo M, Ferradini V, Petricone D, Cicalese MP, Lazarevic D, Cittaro D, Omrani M, et al. (2019). Targeted NGS platforms for genetic screening and gene discovery in primary Immunodeficiencies. Front. Immunol 10, 316. 10.3389/fimmu.2019.00316. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R6] 6.Nijman IJ, van Montfrans JM, Hoogstraat M, Boes ML, van de Corput L, Renner ED, van Zon P, van Lieshout S, Elferink MG, van der Burg M, et al. (2014). Targeted next-generation sequencing: a novel diagnostic tool for primary immunodeficiencies. J. Allergy Clin. Immunol 133, 529–534. 10.1016/j.jaci.2013.08.032. [DOI] [PubMed] [Google Scholar]

[R7] 7.Landrum MJ, Lee JM, Riley GR, Jang W, Rubinstein WS, Church DM, and Maglott DR (2014). ClinVar: public archive of relationships among sequence variation and human phenotype. Nucleic Acids Res. 42, D980–D985. 10.1093/nar/gkt1113. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R8] 8.Michalovich D, and Nejentsev S (2018). Activated PI3 kinase delta syndrome: From genetics to therapy. Front. Immunol 9, 369. 10.3389/fimmu.2018.00369. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R9] 9.Lucas CL, Chandra A, Nejentsev S, Condliffe AM, and Okkenhaug K (2016). PI3Kδ and primary immunodeficiencies. Nat. Rev. Immunol 16, 702–714. 10.1038/nri.2016.93. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R10] 10.Angulo I, Vadas O, Garçon F, Banham-Hall E, Plagnol V, Leahy TR, Baxendale H, Coulter T, Curtis J, Wu C, et al. (2013). Phosphoinositide 3-kinase δ gene mutation predisposes to respiratory infection and airway damage. Science 342, 866–871. 10.1126/science.1243292. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R11] 11.Deau M-C, Heurtier L, Frange P, Suarez F, Bole-Feysot C, Nitschke P, Cavazzana M, Picard C, Durandy A, Fischer A, et al. (2015). A human immunodeficiency caused by mutations in the PIK3R1 gene. J. Clin. Invest 125, 1764–1765. 10.1172/JCI81746. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R12] 12.Lucas CL, Kuehn HS, Zhao F, Niemela JE, Deenick EK, Palendira U, Avery DT, Moens L, Cannons JL, Biancalana M, et al. (2014). Dominant-activating germline mutations in the gene encoding the PI(3)K catalytic subunit p110δ result in T cell senescence and human immunodeficiency. Nat. Immunol 15, 88–97. 10.1038/ni.2771. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R13] 13.Coulter TI, Chandra A, Bacon CM, Babar J, Curtis J, Screaton N, Goodlad JR, Farmer G, Steele CL, Leahy TR, et al. (2017). Clinical spectrum and features of activated phosphoinositide 3-kinase δ syndrome: A large patient cohort study. J. Allergy Clin. Immunol 139, 597–606.e4. 10.1016/j.jaci.2016.06.021. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R14] 14.Elkaim E, Neven B, Bruneau J, Mitsui-Sekinaka K, Stanislas A, Heurtier L, Lucas CL, Matthews H, Deau M-C, Sharapova S, et al. (2016). Clinical and immunologic phenotype associated with activated phosphoinositide 3-kinase δ syndrome 2: A cohort study. J. Allergy Clin. Immunol 138, 210–218.e9. 10.1016/j.jaci.2016.03.022. [DOI] [PubMed] [Google Scholar]

[R15] 15.Preite S, Cannons JL, Radtke AJ, Vujkovic-Cvijin I, Gomez-Rodriguez J, Volpi S, Huang B, Cheng J, Collins N, Reilley J, et al. (2018). Hyperactivated PI3Kδ promotes self and commensal reactivity at the expense of optimal humoral immunity. Nat. Immunol 19, 986–1000. 10.1038/s41590-018-0182-3. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R16] 16.Kracker S, Curtis J, Ibrahim MAA, Sediva A, Salisbury J, Campr V, Debré M, Edgar JDM, Imai K, Picard C, et al. (2014). Occurrence of B-cell lymphomas in patients with activated phosphoinositide 3-kinase δ syndrome. J. Allergy Clin. Immunol 134, 233–236. 10.1016/j.jaci.2014.02.020. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R17] 17.Oh J, Garabedian E, Fuleihan R, and Cunningham-Rundles C (2021). Clinical manifestations and outcomes of activated phosphoinositide 3-kinase δ syndrome from the USIDNET cohort. J. Allergy Clin. Immunol. Pract 9, 4095–4102. 10.1016/j.jaip.2021.07.044. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R18] 18.Wentink MWJ, Mueller YM, Dalm VASH, Driessen GJ, van Hagen PM, van Montfrans JM, van der Burg M, and Katsikis PD (2018). Exhaustion of the CD8+ T cell compartment in patients with mutations in phosphoinositide 3-kinase delta. Front. Immunol 9, 446. 10.3389/fimmu.2018.00446. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R19] 19.Jia Y, Yang Q, Wang Y, Li W, Chen X, Xu T, Tian Z, Feng M, Zhang L, Tang W, et al. (2021). Hyperactive PI3Kδ predisposes naive T cells to activation via aerobic glycolysis programs. Cell. Mol. Immunol 18, 1783–1797. 10.1038/s41423-020-0379-x. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R20] 20.Ruiz-García R, Vargas-Hernández A, Chinn IK, Angelo LS, Cao TN, Coban-Akdemir Z, Jhangiani SN, Meng Q, Forbes LR, Muzny DM, et al. (2018). Mutations in PI3K110δ cause impaired natural killer cell function partially rescued by rapamycin treatment. J. Allergy Clin. Immunol 142, 605–617.e7. 10.1016/j.jaci.2017.11.042. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R21] 21.Cannons JL, Preite S, Kapnick SM, Uzel G, and Schwartzberg PL (2018). Genetic defects in phosphoinositide 3-kinase δ influence CD8+ T cell survival, differentiation, and function. Front. Immunol 9, 1758. 10.3389/fimmu.2018.01758. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R22] 22.Edwards ESJ, Bier J, Cole TS, Wong M, Hsu P, Berglund LJ, Boztug K, Lau A, Gostick E, Price DA, et al. (2019). Activating PIK3CD mutations impair human cytotoxic lymphocyte differentiation and function and EBV immunity. J. Allergy Clin. Immunol 143, 276–291. e6. 10.1016/j.jaci.2018.04.030. [DOI] [PubMed] [Google Scholar]

[R23] 23.Bier J, Rao G, Payne K, Brigden H, French E, Pelham SJ, Lau A, Lenthall H, Edwards ESJ, Smart JM, et al. (2019). Activating mutations in PIK3CD disrupt the differentiation and function of human and murine CD4+ T cells. J. Allergy Clin. Immunol 144, 236–253. 10.1016/j.jaci.2019.01.033. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R24] 24.Meyts I, Bousfiha A, Duff C, Singh S, Lau YL, Condino-Neto A, Bezrodnik L, Ali A, Adeli M, and Drabwell J (2020). Primary immunodeficiencies: A decade of progress and a promising future. Front. Immunol 11, 625753. 10.3389/fimmu.2020.625753. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R25] 25.Sogkas G, Fedchenko M, Dhingra A, Jablonka A, Schmidt RE, and Atschekzei F (2018). Primary immunodeficiency disorder caused by phosphoinositide 3–kinase δ deficiency. J. Allergy Clin. Immunol 142, 1650–1653.e2. 10.1016/j.jaci.2018.06.039. [DOI] [PubMed] [Google Scholar]

[R26] 26.Swan DJ, Aschenbrenner D, Lamb CA, Chakraborty K, Clark J, Pandey S, Engelhardt KR, Chen R, Cavounidis A, Ding Y, et al. (2019). Immunodeficiency, autoimmune thrombocytopenia and enterocolitis caused by autosomal recessive deficiency of PIK3CD-encoded phosphoinositide 3-kinase δ. Haematologica 104, e483–e486. 10.3324/haematol.2018.208397. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R27] 27.Dyment DA, Smith AC, Alcantara D, Schwartzentruber JA, Basel-Vanagaite L, Curry CJ, Temple IK, Reardon W, Mansour S, Haq MR, et al. (2013). Mutations in PIK3R1 cause SHORT syndrome. Am. J. Hum. Genet 93, 158–166. 10.1016/j.ajhg.2013.06.005. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R28] 28.Tomlinson PR, Knox R, Perisic O, Su HC, Brierley GV, Williams RL, and Semple RK (2024). Paradoxical dominant negative activity of an immunodeficiency-associated activating PIK3R1 variantPreprint as. bioRxiv. 10.1101/2023.11.02.565250. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R29] 29.Chudasama KK, Winnay J, Johansson S, Claudi T, König R, Haldorsen I, Johansson B, Woo JR, Aarskog D, Sagen JV, et al. (2013). SHORT syndrome with partial lipodystrophy due to impaired phosphatidylinositol 3 kinase signaling. Am. J. Hum. Genet 93, 150–157. 10.1016/j.ajhg.2013.05.023. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R30] 30.Avila M, Dyment DA, Sagen JV, St-Onge J, Moog U, Chung BHY, Mansour S, Albanese A, Garcia S, Ortiz Martin D, et al. (2015). Clinical reappraisal of SHORT syndrome with PIK3R1 mutations: towards recommendation for molecular testing and management. Clin. Genet 89, 501–506. [DOI] [PubMed] [Google Scholar]

[R31] 31.Bravo García-Morato M, García-Miñaúr S, Molina Garicano J, Santos Simarro F, Del Pino Molina L, López-Granados E, Ferreira Cerdán A, and Rodríguez Pena R (2017). Mutations in PIK3R1 can lead to APDS2, SHORT syndrome or a combination of the two. Clin. Immunol 179, 77–80. 10.1016/j.clim.2017.03.004. [DOI] [PubMed] [Google Scholar]

[R32] 32.Elgizouli M, Lowe DM, Speckmann C, Schubert D, Hülsdünker J, Eskandarian Z, Dudek A, Schmitt-Graeff A, Wanders J, Jørgensen SF, et al. (2016). Activating PI3Kδ mutations in a cohort of 669 patients with primary immunodeficiency. Clin. Exp. Immunol 183, 221–229. 10.1111/cei.12706. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R33] 33.Takeda AJ, Zhang Y, Dornan GL, Siempelkamp BD, Jenkins ML, Matthews HF, McElwee JJ, Bi W, Seeborg FO, Su HC, et al. (2017). Novel PIK3CD mutations affecting N-terminal residues of p110δ cause activated PI3Kδ syndrome (APDS) in humans. J. Allergy Clin. Immunol 140, 1152–1156.e10. 10.1016/j.jaci.2017.03.026. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R34] 34.Rao VK, Webster S, Dalm VASH, Šedivá A, van Hagen PM, Holland S, Rosenzweig SD, Christ AD, Sloth B, Cabanski M, et al. (2017). Effective “activated PI3Kδ syndrome”-targeted therapy with the PI3Kδ inhibitor leniolisib. Blood 130, 2307–2316. 10.1182/blood-2017-08-801191. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R35] 35.Hoegenauer K, Soldermann N, Zécri F, Strang RS, Graveleau N, Wolf RM, Cooke NG, Smith AB, Hollingworth GJ, Blanz J, et al. (2017). Discovery of CDZ173 (leniolisib), representing a structurally novel class of PI3K delta-selective inhibitors. ACS Med. Chem. Lett 8, 975–980. 10.1021/acsmedchemlett.7b00293. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R36] 36.Rao VK, Kulm E, Grossman J, Buchbinder D, Chong H, Bradt J, Webster S, Šedivá A, Dalm VA, and Uzel G (2024). Long-term treatment with selective PI3Kδ inhibitor leniolisib in adults with activated PI3Kδ syndrome. Blood Adv. 8, 3092–3108. 10.1182/bloodadvances.2023011000. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R37] 37.Rao VK, Kulm E, Šedivá A, Plebani A, Schuetz C, Shcherbina A, Dalm VA, Trizzino A, Zharankova Y, Webster S, et al. (2024). Interim analysis: Open-label extension study of leniolisib for patients with APDS. J. Allergy Clin. Immunol 153, 265–274.e9. 10.1016/j.jaci.2023.09.032. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R38] 38.Rao VK, Webster S, Šedivá A, Plebani A, Schuetz C, Shcherbina A, Conlon N, Coulter T, Dalm VA, Trizzino A, et al. (2023). A randomized, placebo-controlled phase 3 trial of the PI3Kδ inhibitor leniolisib for activated PI3Kδ syndrome. Blood 141, 971–983. 10.1182/blood.2022018546. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R39] 39.Gaudelli NM, Komor AC, Rees HA, Packer MS, Badran AH, Bryson DI, and Liu DR (2017). Programmable base editing of A⋅T to G⋅C in genomic DNA without DNA cleavage. Nature 551, 464–471. 10.1038/nature24644. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R40] 40.Anzalone AV, Koblan LW, and Liu DR (2020). Genome editing with CRISPR-Cas nucleases, base editors, transposases and prime editors. Nat. Biotechnol 38, 824–844. 10.1038/s41587-020-0561-9. [DOI] [PubMed] [Google Scholar]

[R41] 41.Walsh ZH, Shah P, Kothapalli N, Shah SB, Nikolenyi G, Brodtman DZ, Leuzzi G, Rogava M, Mu M, Ho P, et al. (2025). Mapping variant effects on anti-tumor hallmarks of primary human T cells with base-editing screens. Nat. Biotechnol 43, 384–395. 10.1038/s41587-024-02235-x. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R42] 42.Richter MF, Zhao KT, Eton E, Lapinaite A, Newby GA, Thuronyi BW, Wilson C, Koblan LW, Zeng J, Bauer DE, et al. (2020). Phage-assisted evolution of an adenine base editor with improved Cas domain compatibility and activity. Nat. Biotechnol 38, 883–891. 10.1038/s41587-020-0453-z. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R43] 43.Kim E, and Hart T (2021). Improved analysis of CRISPR fitness screens and reduced off-target effects with the BAGEL2 gene essentiality classifier. Genome Med. 13, 2. 10.1186/s13073-020-00809-3. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R44] 44.Deau M-C, Heurtier L, Frange P, Suarez F, Bole-Feysot C, Nitschke P, Cavazzana M, Picard C, Durandy A, Fischer A, et al. (2014). A human immunodeficiency caused by mutations in the PIK3R1 gene. J. Clin. Invest 124, 3923–3928. 10.1172/JCI75746. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R45] 45.Nguyen T, Lau A, Bier J, Cooke KC, Lenthall H, Ruiz-Diaz S, Avery DT, Brigden H, Zahra D, Sewell WA, et al. (2023). Human PIK3R1 mutations disrupt lymphocyte differentiation to cause activated PI3Kδ syndrome 2. J. Exp. Med 220, e20221020. 10.1084/jem.20221020. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R46] 46.Zhao P, Huang J, Fu H, Xu J, Li T, Zhang X, Meng Q, Zhang L, Tan L, Zhang W, et al. (2024). Activated phosphoinositide 3-kinase δ syndrome caused by PIK3CD mutations: expanding the phenotype. Pediatr. Rheumatol. Online J 22, 24. 10.1186/s12969-024-00955-7. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R47] 47.Brnich SE, Abou Tayoun AN, Couch FJ, Cutting GR, Greenblatt MS, Heinen CD, Kanavy DM, Luo X, McNulty SM, Starita LM, et al. (2019). Recommendations for application of the functional evidence PS3/BS3 criterion using the ACMG/AMP sequence variant interpretation framework. Genome Med. 12, 3. 10.1186/s13073-019-0690-2. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R48] 48.Gelman H, Dines JN, Berg J, Berger AH, Brnich S, Hisama FM, James RG, Rubin AF, Shendure J, Shirts B, et al. (2019). Recommendations for the collection and use of multiplexed functional data for clinical variant interpretation. Genome Med. 11, 85. 10.1186/s13073-019-0698-7. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R49] 49.Bajaj K, Madhusudhan MS, Adkar BV, Chakrabarti P, Ramakrishnan C, Sali A, and Varadarajan R (2007). Stereochemical criteria for prediction of the effects of proline mutations on protein stability. PLoS Comput. Biol 3, e241. 10.1371/journal.pcbi.0030241. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R50] 50.Fruman DA, Chiu H, Hopkins BD, Bagrodia S, Cantley LC, and Abraham RT (2017). The PI3K pathway in human disease. Cell 170, 605–635. 10.1016/j.cell.2017.07.029. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R51] 51.Heffron TP, Heald RA, Ndubaku C, Wei B, Augistin M, Do S, Edgar K, Eigenbrot C, Friedman L, Gancia E, et al. (2016). The rational design of selective benzoxazepin inhibitors of the α-isoform of phosphoinositide 3-kinase culminating in the identification of (S)-2-((2-(1-isopropyl-1H-1,2,4-triazol-5-yl)-5,6-dihydrobenzo[f]imidazo[1,2-d][1,4]oxazepin-9-yl)oxy)propanamide (GDC-0326). J. Med. Chem 59, 985–1002. 10.1021/acs.jmedchem.5b01483. [DOI] [PubMed] [Google Scholar]

[R52] 52.Jumper J, Evans R, Pritzel A, Green T, Figurnov M, Ronneberger O, Tunyasuvunakool K, Bates R, Žídek A, Potapenko A, et al. (2021). Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589. 10.1038/s41586-021-03819-2. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R53] 53.Cheng J, Novati G, Pan J, Bycroft C, Žemgulytė A, Applebaum T, Pritzel A, Wong LH, Zielinski M, Sargeant T, et al. (2023). Accurate proteome-wide missense variant effect prediction with AlphaMissense. Science 381, eadg7492. 10.1126/science.adg7492. [DOI] [PubMed] [Google Scholar]

[R54] 54.Schmidt R, Ward CC, Dajani R, Armour-Garb Z, Ota M, Allain V, Hernandez R, Layeghi M, Xing G, Goudy L, et al. (2024). Base-editing mutagenesis maps alleles to tune human T cell functions. Nature 625, 805–812. 10.1038/s41586-023-06835-6. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R55] 55.Neugebauer ME, Hsu A, Arbab M, Krasnow NA, McElroy AN, Pandey S, Doman JL, Huang TP, Raguram A, Banskota S, et al. (2023). Evolution of an adenine base editor into a small, efficient cytosine base editor with low off-target activity. Nat. Biotechnol 41, 673–685. 10.1038/s41587-022-01533-6. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R56] 56.Chen L, Zhang S, Xue N, Hong M, Zhang X, Zhang D, Yang J, Bai S, Huang Y, Meng H, et al. (2023). Engineering a precise adenine base editor with minimal bystander editing. Nat. Chem. Biol 19, 101–110. 10.1038/s41589-022-01163-8. [DOI] [PubMed] [Google Scholar]

[R57] 57.Rao VK, Šedivá A, Dalm VASH, Plebani A, Schuetz C, Shcherbina A, Trizzino A, Zharankova Y, Orpia A, Kulm E, et al. (2025). A randomised, placebo-controlled, phase III trial of leniolisib in activated phosphoinositide 3-kinase delta (PI3Kδ) syndrome (APDS): Adolescent and adult subgroup analysis. Clin. Immunol 270, 110400. 10.1016/j.clim.2024.110400. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R58] 58.Khan O, Giles JR, McDonald S, Manne S, Ngiow SF, Patel KP, Werner MT, Huang AC, Alexander KA, Wu JE, et al. (2019). TOX transcriptionally and epigenetically programs CD8+ T cell exhaustion. Nature 571, 211–218. 10.1038/s41586-019-1325-x. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R59] 59.Wu T, Ji Y, Moseman EA, Xu HC, Manglani M, Kirby M, Anderson SM, Handon R, Kenyon E, Elkahloun A, et al. (2016). The TCF1-Bcl6 axis counteracts type I interferon to repress exhaustion and maintain T cell stemness. Sci. Immunol 1, eaai8593. 10.1126/sciimmunol.aai8593. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R60] 60.Zheng Y, Collins SL, Lutz MA, Allen AN, Kole TP, Zarek PE, and Powell JD (2007). A role for mammalian target of rapamycin in regulating T cell activation versus anergy. J. Immunol 178, 2163–2170. 10.4049/jimmunol.178.4.2163. [DOI] [PubMed] [Google Scholar]

[R61] 61.Ianevski A, Giri AK, and Aittokallio T (2022). SynergyFinder 3.0: an interactive analysis and consensus interpretation of multi-drug synergies across multiple samples. Nucleic Acids Res. 50, W739–W743. 10.1093/nar/gkac382. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R62] 62.Vanselow S, Wahn V, and Schuetz C (2023). Activated PI3Kδ syndrome - reviewing challenges in diagnosis and treatment. Front. Immunol 14, 1208567. 10.3389/fimmu.2023.1208567. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R63] 63.Richards S, Aziz N, Bale S, Bick D, Das S, Gastier-Foster J, Grody WW, Hegde M, Lyon E, Spector E, et al. (2015). Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. Genet. Med 17, 405–424. 10.1038/gim.2015.30. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R64] 64.Wang Q, Dhindsa RS, Carss K, Harper AR, Nag A, Tachmazidou I, Vitsios D, Deevi SVV, Mackay A, Muthas D, et al. (2021). Rare variant contribution to human disease in 281,104 UK Biobank exomes. Nature 597, 527–532. 10.1038/s41586-021-03855-y. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R65] 65.Bousfiha A, Moundir A, Tangye SG, Picard C, Jeddane L, Al-Herz W, Rundles CC, Franco JL, Holland SM, Klein C, et al. (2022). The 2022 update of IUIS phenotypical classification for human Inborn Errors of immunity. J. Clin. Immunol 42, 1508–1520. 10.1007/s10875-022-01352-z. [DOI] [PubMed] [Google Scholar]

[R66] 66.Human Inborn Errors of Immunity: 2024 Update on the Classification from the International Union of Immunological Societies Expert Committee. [Google Scholar]

[R67] 67.Meitlis I, Allenspach EJ, Bauman BM, Phan IQ, Dabbah G, Schmitt EG, Camp ND, Torgerson TR, Nickerson DA, Bamshad MJ, et al. (2020). Multiplexed functional assessment of genetic variants in CARD11. Am. J. Hum. Genet 107, 1029–1043. 10.1016/j.ajhg.2020.10.015. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R68] 68.Starita LM, Islam MM, Banerjee T, Adamovich AI, Gullingsrud J, Fields S, Shendure J, and Parvin JD (2018). A multiplex homology-directed DNA repair assay reveals the impact of more than 1,000 BRCA1 missense substitution variants on protein function. Am. J. Hum. Genet 103, 498–508. 10.1016/j.ajhg.2018.07.016. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R69] 69.Findlay GM, Daza RM, Martin B, Zhang MD, Leith AP, Gasperini M, Janizek JD, Huang X, Starita LM, and Shendure J (2018). Accurate classification of BRCA1 variants with saturation genome editing. Nature 562, 217–222. 10.1038/s41586-018-0461-z. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R70] 70.Ma K, Huang S, Ng KK, Lake NJ, Joseph S, Xu J, Lek A, Ge L, Woodman KG, Koczwara KE, et al. (2024). Saturation mutagenesis-reinforced functional assays for disease-related genes. Cell 187, 6707–6724.e22. 10.1016/j.cell.2024.08.047. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R71] 71.Horesh ME, Martin-Fernandez M, Gruber C, Buta S, Le Voyer T, Puzenat E, Lesmana H, Wu Y, Richardson A, Stein D, et al. (2024). Individuals with JAK1 variants are affected by syndromic features encompassing autoimmunity, atopy, colitis, and dermatitis. J. Exp. Med 221, e20232387. 10.1084/jem.20232387. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R72] 72.Lue NZ, and Liau BB (2023). Base editor screens for in situ mutational scanning at scale. Mol. Cell 83, 2167–2187. 10.1016/j.mol-cel.2023.06.009. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R73] 73.Hanna RE, Hegde M, Fagre CR, DeWeirdt PC, Sangree AK, Szegletes Z, Griffith A, Feeley MN, Sanson KR, Baidi Y, et al. (2021). Massively parallel assessment of human variants with base editor screens. Cell 184, 1064–1080.e20. 10.1016/j.cell.2021.01.012. [DOI] [PubMed] [Google Scholar]

[R74] 74.Martinson AK, Chin AT, Butte MJ, and Rider NL (2024). Artificial intelligence and machine learning for inborn errors of immunity: Current state and future promise. J. Allergy Clin. Immunol. Pract 12, 2695–2704. 10.1016/j.jaip.2024.08.012. [DOI] [PubMed] [Google Scholar]

[R75] 75.Li W, Xu H, Xiao T, Cong L, Love MI, Zhang F, Irizarry RA, Liu JS, Brown M, and Liu XS (2014). MAGeCK enables robust identification of essential genes from genome-scale CRISPR/Cas9 knockout screens. Genome Biol. 15, 554. 10.1186/s13059-014-0554-4. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R76] 76.Martin M (2011). Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet J. 17, 10. 10.14806/ej.17.1.200. [DOI] [Google Scholar]

[R77] 77.Schneider CA, Rasband WS, and Eliceiri KW (2012). NIH Image to ImageJ: 25 years of image analysis. Nat. Methods 9, 671–675. 10.1038/nmeth.2089. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R78] 78.Kwon S, Safer J, Nguyen DT, Hoksza D, May P, Arbesfeld JA, Rubin AF, Campbell AJ, Burgin A, and Iqbal S (2024). Genomics 2 Proteins portal: a resource and discovery tool for linking genetic screening outputs to protein sequences and structures. Nat. Methods 21, 1947–1957. 10.1038/s41592-024-02409-0. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R79] 79.Bae S, Park J, and Kim J-S (2014). Cas-OFFinder: a fast and versatile algorithm that searches for potential off-target sites of Cas9 RNA-guided endonucleases. Bioinformatics 30, 1473–1475. 10.1093/bioinformatics/btu048. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R80] 80.Doench JG, Fusi N, Sullender M, Hegde M, Vaimberg EW, Donovan KF, Smith I, Tothova Z, Wilen C, Orchard R, et al. (2016). Optimized sgRNA design to maximize activity and minimize off-target effects of CRISPR-Cas9. Nat. Biotechnol 34, 184–191. 10.1038/nbt.3437. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R81] 81.Pettersen EF, Goddard TD, Huang CC, Meng EC, Couch GS, Croll TI, Morris JH, and Ferrin TE (2021). UCSF ChimeraX: Structure visualization for researchers, educators, and developers. Protein Sci. 30, 70–82. 10.1002/pro.3943. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R82] 82.Riesenberg S, Helmbrecht N, Kanis P, Maricic T, and Pääbo S (2022). Improved gRNA secondary structures allow editing of target sites resistant to CRISPR-Cas9 cleavage. Nat. Commun 13, 489. 10.1038/s41467-022-28137-7. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R83] 83.Martin-Rufino JD, Castano N, Pang M, Grody EI, Joubran S, Caulier A, Wahlster L, Li T, Qiu X, Riera-Escandell AM, et al. (2023). Massively parallel base editing to map variant effects in human hematopoiesis. Cell 186, 2456–2474.e24. 10.1016/j.cell.2023.03.035. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R84] 84.Ye J, Coulouris G, Zaretskaya I, Cutcutache I, Rozen S, and Madden TL (2012). Primer-BLAST: a tool to design target-specific primers for polymerase chain reaction. BMC Bioinformatics 13, 134. 10.1186/1471-2105-13-134. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R85] 85.UniProt Consortium (2025). UniProt: The universal protein knowledgebase in 2025. Nucleic Acids Res. 53, D609–D617. 10.1093/nar/gkae1010. [DOI] [PMC free article] [PubMed] [Google Scholar]

PERMALINK

Scalable generation and functional classification of genetic variants in inborn errors of immunity to accelerate clinical diagnosis and treatment

Zachary H Walsh

Chris J Frangieh

Neeharika Kothapalli

Jay Levy

Clarissa K Heck

Johannes C Melms

Ron S Gejman

Parin Shah

Jared M Pollard

Akul Naik

Sarah L Grauman

Lei Haley Huang

Ashley Lee

Dusan Bogunovic

Joshua D Milner

Benjamin Izar

SUMMARY

In brief

Graphical Abstract

INTRODUCTION

RESULTS

CRISPR base editor screening coupled with a clinically relevant assay in T cells

Figure 1. Saturation adenine base editor screening of PIK3CD and PIK3R1 coupled with a clinically informed GOF/LOF classification assay.

Functional classification of novel GOF/LOF variants is improved with a pathway-specific signaling assay

Variant mapping to PI3Kδ structure reveals spatial distribution and hotspots

Figure 2. Base editor screens enable robust functional classification of novel GOF and LOF variants.

Novel GOF and LOF variants with functional characteristics analogous to known variants

Functional landscape of screen-discovered GOF and LOF variants

Figure 3. PI3Kδ signaling effects and leniolisib sensitivity of novel GOF and LOF variants discovered in screens.

Precision base editors, rational sgRNA design, and computational deconvolution enhance resolution of variant effects

Sensitivity of novel GOF variants to the PI3Kδ inhibitor leniolisib

Leniolisib corrects the dysfunctional T cell state in engineered GOF variants

Figure 4. Phenotypic consequences of novel GOF and LOF variants, correction with Len, and discovery of GOF variants with partial Len resistance.

APDS patient samples reflect T cell signaling, differentiation, and Len responses seen in screen-discovered GOF variants

Figure 5. Functional effects and Len sensitivity of engineered PIK3CD and PIK3R1 GOF variants are reflected in APDS patient samples.

Combinatorial treatments to overcome Len resistance in engineered and APDS patient T cells

Figure 6. Combinatorial drug targeting of PI3Kδ pathway proteins enhances patient sample responses.

Prevalence of screen-discovered GOF variants and associated clinical phenotypes in population databases

Figure 7. Identification of screen-discovered GOF/LOF variants in clinical databases identifies potential cases of APDS and risk for immunopathology.

DISCUSSION

Limitations of the study

RESOURCE AVAILABILITY

Lead contact

Materials availability

Data and code availability

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Primary T cell isolation and culture

METHOD DETAILS

sgRNA library design for base editor tiling of PIK3CD and PIK3R1

Base editor library cloning

Lentiviral sgRNA library production and concentration

Lentiviral transduction and selection of T cells for screens

In vitro transcription (IVT) of Base Editor mRNA

Base editing of primary T cells

Screen Format

Arrayed Validation Format

Flow cytometry-based assays

Surface staining

Intracellular staining (non phospho-flow)

Phospho-flow (stimulation and staining)

pAKT/pS6 Screen

Screen design

Screen workflow

DNA extraction and next-generation sequencing

Base editing screens

Base editing arrayed validation experiments

APDS patient samples

Screen quality control and evaluation of off-target effects

Quality control

Evaluation of off-target effects

Screen analyses and statistics

Construction of a variant-level score

Visualization of variant effects on PI3Kδ protein structure

Assessment of variant effect association with AlphaFold and AlphaMissense residue metrics

Assessment of single and combinatorial drug effects on T cell signaling and phenotype

Microscopy

Clinical database queries