Heterozygous splice mutation in PIK3R1 causes human immunodeficiency with lymphoproliferation due to dominant activation of PI3K

Our evaluation of the index patient A.1, a 32-yr-old female, revealed recurrent sinopulmonary infections with associated bronchiectasis, diffuse lymphadenopathy with splenomegaly, refractory immune thrombocytopenia purpura treated with splenectomy, and development of arthritis and inflammatory bowel disease (IBD) in adulthood (Table 1). Analysis of her blood cells revealed T lymphocytosis caused by expansion of CD8 T cells with normal CD4 T cell counts and normal or low B cell counts. Serum IgG levels were historically low in this patient, leading to treatment with intravenous immunoglobulin (IVIG) since the age of 6 yr. Serum IgA and IgM were low and high, respectively. Given her history of inflammatory arthritis and IBD, we evaluated markers of inflammation and found elevated erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP). The patients in family B suffer from a similar disease with immunodeficiency, lymphoproliferation, and poor production of class-switched antibody. Patient B.1 developed Hodgkin lymphoma and has undergone hematopoietic stem cell transplantation (HSCT), and patient B.II.1 has been treated with IVIG since age 9, followed by Rituximab to deplete B cells starting at age 15. Patient C.1 is a young boy with immunodeficiency, lymphadenopathy, hepatomegaly, and severe juvenile rheumatoid arthritis. All patients have required IVIG supplementation and have a low CD4/CD8 ratio due to expansion of CD8 T cells.

Thus, the clinical features of this disease include immunodeficiency, lymphoproliferation, poor immunoglobulin responses, and expansion of CD8 T cells with susceptibility to inflammatory disease. These findings fit the clinical picture of PASLI disease (Lucas et al., 2014; also called activated p110δ syndrome; Angulo et al., 2013) caused by p110δ hyperactivation. However, mutations in the PIK3CD gene encoding p110δ were not found in the individuals described above, leading to the hypothesis that a mutation in another PI3K gene could be responsible for their disease.

The three families in our cohort demonstrate a dominant or de novo disease inheritance pattern, and whole-exome sequencing (WES) uncovered a heterozygous point mutation in PIK3R1 at a splice donor site (G > C or G > A at chr5:67589663, which corresponds to c.1425+1 in NM_181523.2) in genomic DNA from PBMCs in patients A.1, B.II.1, and C.1 (Fig. 1 A). Due to her HSCT, genomic DNA was isolated from fingernail clippings from patient B.1 to confirm the presence of the mutation. The PIK3R1 mRNA variant encoding p85α contains 16 exons beginning with a noncoding exon 1 in the 5′ UTR, and the patient splice site mutation is in the first intronic nucleotide after exon 11. The mutation results in skipping of exon 11 (delE11), as revealed by sequencing of PIK3R1 cDNA (Fig. 1 B). As previously reported (Deau et al., 2014), loss of exon 11 led to an in-frame deletion of the nucleotides encoding amino acid residues 434–475 of p85α, which falls in the inter-SH2 domain shared by the p85α, p55α, and p50α PIK3R1 isoforms (Fig. 1 C). This deletion is immediately N-terminal to the mapped binding site (residues 478–513) of the p110 catalytic subunit (Dhand et al., 1994). Structural modeling of the truncated inter-SH2 domain showed loss of part of the coiled coil structure of the inter-SH2 domain and predicted a dramatic change in the position of the nSH2 (positioned in the lower left in WT and shifted to the top in delE11) and cSH2 domains that are required for inhibitory contacts with p110 and recruitment to phosphotyrosine motifs (Fig. 1 D). Such a marked alteration in the inter-SH2 domain and the positioning of the SH2 domains strongly suggests that there will be qualitative and/or quantitative alterations in the contacts made between delE11 p85α and the p110 catalytic subunit. Moreover, all three PIK3R1 gene products, the p85α, p55α, and p50α regulatory subunits, will be affected by this structural alteration that is predicted to impair p110 inhibition (Fig. 1 C).

We next sought to directly examine PI3K signaling in these patients. PI3K activity generates PIP₃ to activate several downstream signaling pathways via the AKT serine/threonine kinase, which is recruited to the membrane at signaling sites by binding to PIP₃ through its PH domain. Once active, AKT phosphorylates the GSK3α/β, mTOR, and FOXO1/3 proteins to promote protein translation, cell growth, proliferation, and survival (Fig. 2 A). We directly examined PI3K signaling in patient T cell blasts by evaluating AKT phosphorylation on threonine 308 and serine 473 and found substantially more basal phospho-AKT in patient samples compared with controls (Figs. 2 B). Also consistent with augmented PI3K signaling, phosphorylation of S6 at serine residues 235/236 and serine residues 240/244 was constitutively increased in patient T cell blasts compared with controls (Fig. 2 C). Finally, phosphorylation of FOXO1/3a (T24/T32) and GSK3α/β (S21/9) was similarly augmented in patients compared with controls (Fig. 2 D). Thus, the heterozygous PIK3R1 mutations in these patients promote pathological, constitutive hyperactivation of PI3K and AKT, which leads to activation of the mTORC1 complex and phospho-inhibition of GSK3α/β and FOXO1/3.

We next sought to examine expression of the mutant p85α protein to determine if the in-frame deletion produces a stable protein product. Detection of endogenous p85α by immunoblot in lysates from patient T cell blasts revealed a reduction in full-length p85α compared with controls (Fig. 3 A). Importantly, a band at the predicted size of 78 kD was detectable in patient lanes (Fig. 3 A, bottom arrow). However, the truncated delE11 mutant form of p50α was undetectable (Fig. 3 B). The low-level mutant p50α protein expression despite the clear mRNA expression (unpublished data) suggests that the mutant protein is unstable. Examination of PI3K signaling after siRNA-mediated knockdown of PIK3R1 showed no evidence for haploinsufficiency of PIK3R1 contributing to hyperactivation of PI3K in primary human T cells (unpublished data). Moreover, we observed normal expression of the PTEN and SHIP1 phosphatases in patient T cell blasts, excluding the possibility that reduction in expression of the enzymes that counteract PI3K accounted for increased signaling (unpublished data). We next performed overexpression experiments in which WT versus delE11 5x-myc–tagged p85α was ectopically expressed in T cells from a healthy control subject and found that the mutant exerted a dominant, gain-of-function effect on PI3K signaling, causing increased AKT serine 473 phosphorylation (Fig. 3 C, compare lanes 1 and 3). We further found that WT and delE11 p85α-myc were similarly recruited to linker of activated T cells (LAT), a protein known to recruit p85α upon tyrosine phosphorylation (unpublished data).

Because the deletion removes residues near the p110-binding site, we next determined whether or not delE11 p85α associates normally with p110δ. To test this, we overexpressed myc-tagged WT or delE11 p85α in T cells from a healthy control subject and immunoprecipitated p110δ from cell lysates. The p110δ catalytic subunit strongly associated with myc-tagged WT p85α but exhibited reduced association with myc-tagged delE11 p85α despite robust expression in the input lysate (Fig. 3 D). Impaired association with p110δ offers a possible explanation for the reduced expression of delE11 p85α in patient cells because the p85α monomer has been shown to be unstable (Brachmann et al., 2005; Zhao et al., 2006). These data, together with the observation that p110δ protein levels are relatively normal in these patients, indicate that the delE11 p85α protein expressed in patient cells can associate with p110δ and stabilize it, though the affinity of the interaction is reduced compared with WT because the complexes fail to remain intact in immunoprecipitation procedures.

The experiments above shed light on the stability of the association between p85α and p110δ, which dimerize via stable contacts mediated by the inter-SH2 domain of p85α and the adaptor-binding domain of p110δ. However, the data above do not address the additional inhibitory contacts normally made between the SH2/inter-SH2 domains of p85α and various regions of p110δ. We assessed the contribution of the SH2/inter-SH2 domains of the delE11 p85α mutant in driving PI3K signaling by overexpressing myc-tagged p50α protein containing only these domains. When WT and delE11 forms of the C-terminal p50α fragment were overexpressed, we observed that delE11 p50α was sufficient to increase AKT phosphorylation on serine 473 (Fig. 3 E). These data rule out a role for the N-terminal domains (i.e., the SH3, PRR, and BH domains) of p85α in mediating the dominant gain-of-function effect on PI3K activity and instead point to a failure of delE11 p85α to inhibit p110δ activity due to alterations in the structure of the SH2 and inter-SH2 domains that normally form inhibitory contacts with the p110 subunit. Thus, the heterozygous PIK3R1 splice mutation results in expression of mutant p85α that is sufficient to dominantly hyperactivate the PI3K pathway due to qualitatively and quantitatively different binding and impaired inhibition of p110δ.

Because the PIK3R1 mutation in these patients drives PI3K signaling, we expected to see similar cellular derangements as previously observed in PASLI patients with gain-of-function mutations in PIK3CD encoding p110δ. We therefore evaluated patient PBMCs and found several of the key cellular features of PASLI disease. CD8 T cells were increased in frequency and deficient in naive (CD45RA⁺CCR7⁺) cells with overrepresentation of CCR7-negative effector-type T cells in PBMCs from patients compared with controls (Fig. 4 A). T cell activation, as measured by CD25 up-regulation, was normal after stimulation of PBMCs with anti-CD3 and anti-CD28 (unpublished data). We next evaluated glucose uptake in T cell blasts using a fluorescent glucose analogue and found, as observed for PASLI patients with PIK3CD mutations, an increase in glucose uptake by T cell blasts from PIK3R1 patients compared with controls (Fig. 4 B). Moreover, the majority of patient CD8 T cells were CD57⁺, indicating terminal differentiation and senescence of the CD8 T cell population (Fig. 4 C). To extend these cellular findings, we assessed telomere length in the lymphocyte population by flow cytometry–based fluorescence in situ hybridization (flow-FISH) and observed severely shortened telomere length in PIK3R1 patients (Fig. 4 D). Thus, in PIK3R1-driven disease as in p110δ PASLI disease, hyperactive PI3K signaling in lymphocytes causes expansion, terminal differentiation, and senescence of CD8 T cells.

To further validate our hypothesis that the PIK3R1 mutation underlies the disease pathogenesis in these patients and causes hyperactivation of PI3K and mTOR in lymphocytes, we next tested the effect of inhibiting these kinases in patient T cell cultures. As expected, inhibition of mTOR with rapamycin clearly reduced S6 phosphorylation but not AKT phosphorylation, whereas treatment of patient cells with GS1101, an FDA-approved p110δ-specific inhibitor, or Wortmannin, a pan-PI3K inhibitor, efficiently reduced both AKT and S6 phosphorylation (Fig. 5). The observation that the hyperactive PI3K pathway in patient T cell blasts was inhibited equally well by p110δ-specific inhibition and broader inhibition of all class IA PI3K catalytic subunits suggests that p110δ accounts for most or all of the hyperactive PI3K signaling in T cell blasts, which also express p110α and p110β. This finding raises the interesting possibility that delE11 p85α has unique effects on p110δ relative to other catalytic subunits and may help to explain why the clinical phenotype of PIK3R1 patients involves effects mainly in the immune system. Moreover, these findings suggest treatment options for these patients using rapamycin or p110δ-specific inhibitors.

The PI3K signaling pathway plays critical roles in virtually all cell types and has been intensively studied because of its prominent role in promoting tumorigenesis. It has become clear that either too little or too much PI3K signaling can be pathological. Mice completely deficient in the PIK3R1 gene products die in the perinatal period and exhibit liver necrosis and hypoglycemia, whereas mice lacking just the p85α isoform are viable but suffer from hypoglycemia and increased insulin sensitivity linked to up-regulation of the p55α and p50α isoforms (Fruman et al., 2000). Hyperactivation of PI3K signaling has been identified as a cause of human diseases including megalencephaly syndromes (Lee et al., 2012; Rivière et al., 2012) and overgrowth syndromes (Lindhurst et al., 2012; Keppler-Noreuil et al., 2014). A vast literature also places PI3K among the top pathways driving cancer cell growth and survival through gain-of-function mutation of PI3K genes, most commonly PIK3CA encoding the p110α catalytic subunit, or mutations activating upstream initiators of PI3K signaling (Wong et al., 2010).

Although most studies of PI3K signaling in disease states focus on the catalytic subunit, recent reports have clearly highlighted the important role that the regulatory subunit can play in modulating PI3K activity. Recently, several independent groups have identified heterozygous mutations in PIK3R1 in patients with SHORT syndrome, a rare disorder characterized by short stature, hyperextensible joints, ocular depression, Rieger anomaly of the eye, teething delays, and characteristic facial gestalt (Chudasama et al., 2013; Dyment et al., 2013; Thauvin-Robinet et al., 2013; Bárcena et al., 2014). The PIK3R1 mutations in these patients cluster predominantly in the cSH2 domain and the biochemical effect of these heterozygous mutations appears to be a reduction in insulin-stimulated PI3K activity (Chudasama et al., 2013; Dyment et al., 2013; Thauvin-Robinet et al., 2013; Bárcena et al., 2014). No severe immune abnormalities have been reported in patients with SHORT syndrome, and the clinical features of SHORT syndrome are not found in our PIK3R1 patients with PASLI-like immune disease. These observations raise the possibility that the C-terminal SH2 domain of p85α plays a unique role in binding tyrosine-phosphorylated insulin receptor substrate-1 (IRS-1) during insulin signaling, leading to SHORT syndrome when mutated. In contrast, the N-terminal SH2 and inter-SH2 domains of p85α appear to be particularly important in inhibiting catalytic activity of p110δ because their alteration in our heterozygous PIK3R1 patients causes PASLI-like immune disease via hyperactivation of p110δ.

Other PIK3R1 findings particularly relevant for the studies described here are reports describing patients with homozygous loss of p85α (Conley et al., 2012) and tumors with heterozygous gain-of-function PIK3R1 variants that augment PI3K signaling (Jaiswal et al., 2009; Sun et al., 2010; Urick et al., 2011). Conley et al. (2012) described a patient with colitis and absent B cells with normal T cells who harbored a homozygous loss-of-function mutation in PIK3R1 resulting in absence of p85α protein with normal expression of the p50α and p55α splice variants. The mutation results in a premature stop codon in the BH domain that is present in the p85α but not the p55α or p50α isoforms of PIK3R1. Due to the lack of the regulatory p85α protein, a loss in expression of the catalytic p110δ subunit was observed, emphasizing the requirement for p85α to stabilize p110δ even when p55α and p50α are present. Interestingly, loss of p85α led to a dramatic decrease in p110δ with a more subtle decrease in expression of p110α and p110β (Conley et al., 2012). Thus, p85α/p110δ are required for B cell development, and the N-terminal region of p85α plays a particularly important role in stability of p110δ (more so than the other class IA catalytic subunits). The heterozygous parents of this patient were not reported to be clinically affected, providing further support that haploinsufficiency of p85α is not a driver of PI3K hyperactivation in the immune system. In contrast to p85α loss of function, heterozygous mutations in PIK3R1 that drive hyperactivation of PI3K have been identified in human cancers and found to have no effect on p110 protein expression levels (Jaiswal et al., 2009; Sun et al., 2010; Urick et al., 2011). In endometrial cancer, these mutations cluster in the inter-SH2 domain of p85α around amino acid residues 434–475, which are precisely the residues deleted by skipping of exon 11 in our patients, and the PIK3R1 mutants drive hyperactivation of PI3K signaling (Urick et al., 2011).

Here, we describe patients who suffer from immunodeficiency with lymphoproliferation, antibody defects, and terminal differentiation and senescence of CD8 T cells. We identified a heterozygous splice site mutation that leads to in-frame deletion of exon 11 of PIK3R1 and hyperactivation of PI3K and AKT. Similar patients with the same splice site mutation have now been independently reported (Deau et al., 2014). As in Deau et al. (2014), we describe a similar patient phenotype, mutation site, exon skipping, hyperphosphorylation of AKT and S6, and ability of overexpressed delE11 p85α to induce hyperphosphorylation of AKT. We further show that other AKT substrates including FOXO and GSK3 proteins are hyperphosphorylated in patient T cells. Most of the clinical features of the disease observed in our patients and those reported by Deau et al. (2014) are consistent, with lymphoproliferation and autoimmune/inflammatory disease being the two exceptions. We observed marked lymphoproliferation and reduced CD4/CD8 ratios in our patients, whereas only one of four patients reported by Deau et al. (2014) is reported to have evidence of lymphoproliferation (enlarged tonsils) and two of four patients have reduced CD4/CD8 ratios. Additionally, two of our four patients had arthritis, whereas none was reported by Deau et al. (2014). In both patient cohorts, recurrent sinopulmonary infections, immunoglobulin defects, and reduction in naive T cells are key characteristics of the disease. The molecular basis of PI3K hyperactivation and metabolic effects of cellular phenotype were not elaborated in Deau et al. (2014). We now report the mechanism by which deletion of amino acid residues 434–475 of p85α augments PI3K signaling and alters T cell phenotype and function, and our major conclusions are discussed below.

Normal T cells robustly express the p85α and p50α isoforms but not the p55α isoform of PIK3R1. Analysis of patient T cell lysates by immunoblot for PIK3R1 protein products using an antibody against the nSH2 domain revealed, as expected, that full-length WT p85α and p50α were reduced in protein abundance. A truncated delE11 p85α band was observed at the expected size of 78 kD; however, the delE11 p50α mutant protein could not be detected despite clear mRNA expression (unpublished mRNA data). These observations left open the possibilities of gain-of-function effects on PI3K signaling mediated by low-level expression of delE11 p85α or a haploinsufficiency effect due to the reduction in expression of full-length WT p85α and p50α. Although haploinsufficiency of PIK3R1 has been shown previously to increase PI3K signaling in response to insulin (Mauvais-Jarvis et al., 2002), we ruled out a major contribution of PIK3R1 haploinsufficiency in this immune cell phenotype using siRNA-mediated knockdown of PIK3R1 in T cells from healthy control subjects. These experiments showed no effect on levels of basal AKT phosphorylation, which is consistent with previous reports that insulin signaling is unique in its hyperresponsiveness caused by PIK3R1 haploinsufficiency (Vanhaesebroeck et al., 2005). In contrast, when delE11 p85α was overexpressed in control T cells, we observed a clear increase in basal AKT phosphorylation relative to the WT p85α transfectants. This finding supports the conclusion that expression of the delE11 variant of p85α in patient immune cells is a dominant driver of PI3K hyperactivation.

To understand how delE11 p85α drives exuberant PI3K signaling, we examined the ability of this mutant protein to associate with the p110δ catalytic subunit because the deleted exon encodes residues adjacent to the p110 binding site. These experiments demonstrated a detectable but reduced association between endogenous p110δ and overexpressed delE11 p85α, whereas the association with overexpressed WT p85α was robust. This reduced association offers an explanation for the low-level expression of delE11 p85α in patient lysates because p85α is unstable if not bound to p110 (Brachmann et al., 2005; Zhao et al., 2006). How can delE11 p85α promote p110δ hyperactivation despite less stable interaction with p110δ? We evaluated three major possibilities: (1) structural effects of the deletion result in increased recruitment of the PI3K complex to phospho-tyrosine; (2) delE11 p85α that is free from p110 interacts with other signaling molecules via its N-terminal SH3 domain, proline-rich regions, and/or BH domain to promote PI3K activity; and/or (3) the ability of delE11 p85α to form inhibitory contacts with p110δ is impaired. To address (1), we evaluated association of myc-tagged WT or delE11 p85α with LAT, which recruits p85α once it is tyrosine phosphorylated upon TCR engagement, and found that WT and delE11 p85α were similarly associated with LAT (unpublished data). We next addressed possibilities (2) and (3) by cloning and overexpressing the C-terminal fragment of WT and delE11 p85α containing only the nSH2, inter-SH2, and cSH2 domains (i.e., p50α) that normally form inhibitory contacts with p110. In T cells from healthy subjects, we found that overexpression of delE11 p50α was sufficient to cause hyperphosphorylation of AKT at serine 473. This result rules out a requirement for the N-terminal SH3, PRR, and/or BH domains of “free” delE11 p85α in promoting PI3K activity. Instead, our findings are consistent with a failure of the SH2 and inter-SH2 domains of delE11 p85α to properly inhibit p110δ due to structural derangements that interfere with formation of inhibitory contacts.

We now recognize two classes of PASLI disease distinguished by the gene underlying hyperactive PI3K signaling. We term these classes PASLI-CD and PASLI-R1 for the forms caused by mutations in PIK3CD and PIK3R1, respectively. The clinical and cellular consequences of hyperactive PI3K signaling in immune cells were revealed by studies of PASLI-CD patients harboring heterozygous, gain-of-function mutations in PIK3CD encoding p110δ (Angulo et al., 2013; Lucas et al., 2014). The clinical presentations of the PASLI-R1 patients described here and in a recently published report (Deau et al., 2014) are remarkably similar to PASLI-CD patients, with recurrent sinopulmonary infections, poor antibody responses requiring IVIG, susceptibility to EBV and CMV, lymphoproliferation, and increased incidence of lymphoma. At the cellular level, we now show that, as in PASLI-CD patients, this new cohort of PASLI-R1 patients have increased glucose uptake, expanded CD8 T cells with significant reduction in the frequency of naive cells, and concomitant increase in the frequency of terminally differentiated effector-type T cells (Lucas et al., 2014). The CD57 senescence marker is also highly expressed on CD8 T cells in PIK3R1 patients, and we extended this finding by determining that telomeres in lymphocytes are markedly shorter than expected for the age of the patient. Although presently these diseases appear to be very similar, following the natural history of these patients may reveal subtle clinical and/or molecular differences.

Given the molecular etiology of this disease, treatment with drugs inhibiting the PI3K signaling pathway would offer the most ideal therapeutic option for these patients. Indeed, we found that inhibition of mTOR using rapamycin in vitro robustly inhibited the hyperphosphorylation of S6 but had no effect on AKT hyperphosphorylation in PASLI-R1 T cell blasts. Because other targets of AKT besides mTOR (e.g., FOXO and GSK3α/β) are also hyperphosphorylated, a more ideal approach would be to specifically target select PI3K isoforms while avoiding negative consequences of global PI3K inhibition. Although T cells express all three class IA catalytic subunits, we were intrigued to find that inhibition of the leukocyte-restricted p110δ catalytic subunit alone was sufficient to reduce the hyperphosphorylation of AKT and S6 in PIK3R1 patient T cell blasts. This has the interesting biological implication of p85α-mediated regulation of p110δ being unique and selectively affected by loss of PIK3R1 exon 11. Given these findings and the predominantly immunological clinical disease in these patients, specific targeting of p110δ using inhibitors such as those currently in clinical trials for hematological malignancy (Lannutti et al., 2011; Brown et al., 2014; Flinn et al., 2014; Gilbert, 2014; Kahl et al., 2014) offers a potentially safe and effective treatment option to restore immune homeostasis while limiting unwanted effects in nonimmune cells that do not express p110δ.

All human subjects (or their legal guardians) in this study provided written informed consent in accordance with Helsinki principles for enrollment in research protocols that were approved by the Institutional Review Board of the National Institute of Allergy and Infectious Diseases, National Institutes of Health (NIH). Blood from healthy donors was obtained at the NIH Clinical Center under approved protocols. Mutations will be automatically archived by Online Mendelian inheritance in Man (OMIM) once the paper is published, and whole-exome data will be submitted in dbGaP. We will also create a webpage through National Center for Biotechnology Information (NCBI) to accumulate patient mutation data in the format of the Leiden Online Variant Database (LOVD) as patients are identified.

Genomic DNA was isolated from PBMC for proband A.1 and family members. SureSelect Human All Exon 50 Mb kit (Agilent Technologies) coupled with massively parallel sequencing by Illumina HiSeq Sequencing System was performed using the collected DNA. For individual samples, WES produced ∼50–100× sequence coverage for targeted regions. Similar WES was also performed for patient B.II.1 and C.1. For confirmation of mutations in the patients, genomic DNA was PCR amplified and Sanger sequencing of purified PCR amplified products was performed. For cDNA preparation, total RNA was extracted from T cell blasts and cDNA was produced using an oligo-dT primer. The PIK3R1 gene was amplified from the cDNA and the full-length and truncated bands were excised from an agarose gel for Sanger sequencing to detect joining of exon 10 with exon 12.

All sequenced DNA reads were mapped to the hg19 human genome reference by Burrows-Wheeler Aligner with default parameters. Single nucleotide variant and indel calling were performed using the Genome Analysis Toolkit (the Broad Institute). All SNVs/indels were annotated by SeattleSeq Annotation and an in-house custom analysis pipeline was used to filter and prioritize for autosomal recessive or de novo disease-causal variants based on the clinical pedigree for patient A.1. For patient B.II.1 and C.1, the mutations were identified by targeted gene screening of the WES databased on the similarity of clinical phenotype in the cohort.

The structure of the delE11 p85α N-SH2 and inter-SH2 domains was generated by manual editing of p85α coordinates obtained from the complex of p110α and p85α (PDB ID: 3HHM), which were regularized (i.e., disordered amino acids were interpolated) using SWISS-MODEL. The p85α N-SH2 domain and extant region of the coiled-coil domain were treated as rigid bodies, and were fused together by manual transposition of the SH2 domain to the new terminal residue of the coiled-coil. The relative orientations of the domains were then modified to resolve steric clashes with the cognate p110δ subunit of PI3K, modeled using I-TASSER and ModRefiner. All manual modeling and subsequent rendering was performed using PyMOL.

Human PBMCs were isolated by Ficoll-Paque PLUS (GE Healthcare) density gradient centrifugation, washed twice in PBS, and resuspended at 10⁶ cells/ml complete RPMI 1640 (cRPMI) medium (Lonza) containing 10% FBS, 2 mM glutamine, and 100 U/ml each of penicillin and streptomycin (Invitrogen). Cells were activated with 1 µg/ml anti-CD3 (clone HIT3α; BD) and 1 µg/ml anti-CD28 (clone CD28.2; BD). After 3 d, activated T cells were washed and then cultured in complete RPMI-1640 medium supplemented with 100 U/ml recombinant human IL-2 (rhIL-2; R&D Systems). Transfection was performed with Amaxa Nucleofection kits (Lonza) for primary cells, using C-terminally 5×-myc-tagged constructs in pcDNA3.1.

For standard surface staining, PBMCs (10⁶ cells/sample), sorted cells, expanded T cells, or cell lines were washed with PBS and incubated for 30 min at 4°C (dark) in 100 µl 5% FBS in PBS with indicated fluorochrome-labeled monoclonal antibodies or their isotype controls. After washing with PBS two times, 10,000-50,000 live cells were analyzed by flow cytometry. For phosflow staining, unless otherwise indicated, cells were kept in cRPMI while alive, fixed directly in cRPMI using Lyse/Fix (BD), and then permeabilized with perm buffer III (BD) according to manufacturer’s instructions. The indicated antibodies were purchased from BD, eBioscience, BioLegend, or Cell Signaling Technology. For phosflow analyses, anti-pAKT S473 Alexa Fluor 647 (D9E; Cell Signaling Technology), anti-pAKT T308 Alexa Fluor 488 (C31E5E; Cell Signaling Technology), anti-pS6 S235/236 PE (N7-548; BD), and/or anti-pS6 S240/244 Alexa Fluor 647 (D68F8; Cell Signaling Technology) were used. Rabbit isotype control staining was performed for each experiment (Fig. 2, B, left, with Fig. 2 C, top; Fig. 2 B, right, with Fig. 2 C, bottom; and Fig. 5).

Cells were washed in PBS or RPMI with no FCS and immediately lysed on ice in 1% Triton X-100, 50 mM Tris-Cl, pH 8, 150 mM NaCl, 2 mM EDTA, 10% glycerol, complete protease inhibitor cocktail (Roche), and phosphatase inhibitor cocktails (Sigma-Aldrich). Protein was quantitated by BCA assay (Thermo Fisher Scientific). The lysates were then clarified by centrifugation at 15,000 g at 4°C for 10 min. Supernatants were transferred to a separate tube and used for subsequent experimentation. Approximately 20 µg of total protein was separated by SDS-PAGE and transferred to a nitrocellulose membrane (Bio-Rad Laboratories). Membranes were blocked with 5% nonfat dry milk in Tris-buffered saline (TBS) with 0.01% Tween-20 (TBST) for 1 h at room temperature before incubating with primary antibody overnight at 4°C. After washing with TBST for 1 h at room temperature, HRP-conjugated secondary antibody was added for an additional hour at room temperature. After a final 1-h wash step, HRP substrate (Luminata Forte; Millipore) was added to the membranes, which were then subjected to chemiluminescent imaging. Validated primary antibodies were purchased from Cell Signaling Technology, Santa Cruz Biotechnology, Inc., or Millipore, and secondary antibodies were from SouthernBiotech. For immunoprecipitation (IP) of p110δ, cells overexpressing empty vector (EV), WT p85α-myc, or delE11 p85α-myc were lysed as described above, precleared of proteins that nonspecifically bind anti-rabbit IgG IP beads (Rockland Immunochemicals), and incubated overnight with rabbit anti-p110δ IgG (Santa Cruz Biotechnology, Inc.). Anti-rabbit IgG IP beads were then added to capture the immune complexes, and samples were rotated for 1 h at 4°C. Captured immune complexes were pelleted by centrifugation, washed three times with lysis buffer, and eluted by boiling after addition of SDS-containing sample buffer with reducing agent.

Cells were starved of glucose by incubation in PBS for 1 h before incubation with 100 µg/ml 2-deoxy-2-[(7-nitro-2,1,3-benzoxadiazol-4-yl)amino]-d-glucose (2-NBDG) for 20 min. Uptake was measured by flow cytometric evaluation of signal in the fluorescein channel.

Fresh blood from patient A.1 was shipped directly to Repeat Diagnostics for measurement of telomere length by flow cytometry, gating on either lymphocytes or granulocytes. A fluorescently labeled nucleic acid probe was used to hybridize with the TTAGGG repeats in telomeres, and fluorescence intensity provided a measure of telomere length. Telomere length measurements for patient lymphocytes and granulocytes were plotted relative to telomere lengths established for healthy control subjects with percentiles indicated.

The authors thank the referring physicians as well as our patients and families who have gone to extraordinary efforts to join this study and provide samples. We also thank Deborah Rawson, Elaine Smoot, Julie Niemela, Jennifer Stoddard, and Dr. Sergio Rosenzweig for their important contributions to the clinical management of and laboratory medicine findings in these patients.

This research was supported by the Division of Intramural Research, National Institute of Allergy and Infectious Diseases and National Institutes of Health (NIH) Clinical Center, NIH. The work was also supported by Merck Inc., and we would like to thank Robert Plenge, Rupert Vessey, Dennis Zoller, Debbie Law, Huseyin Mehmet, and Ravi Ramadas for their support and collaboration. C.L. Lucas was supported by the Postdoctoral Research Associate (PRAT) Fellowship through the National Institute of General Medical Sciences. X. Wang and Y. Wang were supported by the National Natural Science Foundation of China (81172877, 81373221). The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.

The authors declare no competing financial interests.

Contributions: C.L. Lucas performed experiments, analyzed data, and developed and wrote the manuscript. Y. Zhang analyzed WES data and discovered PIK3R1 mutations. A. Venida assisted with experiments and analyzed data. Y. Wang provided clinical care for and analysis of patient C.1. J. Hughes performed/organized WES studies and analyzed data. J. McElwee performed/organized WES studies and analyzed data. M. Butrick assisted with clinical management of patient A.1. H. Matthews assisted with clinical management of patients A.1, B.1, and B.II.1. S. Price assisted with clinical care of patient A.1. M. Biancalana performed structural modeling analyses. X. Wang provided clinical care for and supervised analysis of patient C.1. M. Richards provided clinical care for and evaluation of patient B.II.1. T. Pozos provided clinical care for and evaluation of patient B.II.1. I. Barlan provided clinical care for and evaluation of patient A.1. A. Ozen provided clinical care for and evaluation of patient A.1. V.K. Rao provided clinical care for and evaluation of patient A.1. H.C. Su provided clinical and research advice and supervised research and data analysis. M.J. Lenardo supervised research and data analysis, provided advice, and edited the manuscript. All authors discussed and reviewed the manuscript.