Phosphoinositide 3-kinase β mediates microvascular endothelial repair of thrombotic microangiopathy

Endothelial damage causing thrombotic microangiopathy (TMA) commonly results from exposure to Shiga-like toxin associated with epidemic Escherichia coli infections, mutations of complement regulatory proteins, drug toxicity, and acute and chronic immune responses against allogeneic vascular endothelium in the context of bone marrow or solid-organ transplantation.^1-3  These diseases often follow a fulminant course leading to rapid failure of affected organs.⁴  In particular, thrombosis involving the kidney glomerular capillaries results in acute kidney failure and is a major cause of morbidity.^2,5,6  Nevertheless, despite catastrophic glomerular endothelial damage requiring kidney-replacement therapy, many individuals with acute injury to a healthy microvasculature recover kidney function.^7,8 

The mechanisms underlying repair of the microvasculature after TMA injury of microvessel endothelial cells (ECs) and resolution of fibrin clot occlusion of the capillaries leading to recovery of organ function are poorly defined. The cues to embryonic vascular development are lead candidates to initiate repair of the established microcirculation. These include factors secreted by parenchymal cells of the tissue such as vascular endothelial growth factor (VEGF), angiopoetin-1, or sphingosine 1-phosphate, which signal to the endothelium through receptor tyrosine kinases (RTKs) and G protein–coupled receptors (GPCRs).^9-11 

In the endothelium, phosphoinositide 3-kinases (PI3Ks) integrate many growth factor signals in development and hence are candidates to mediate reparative cues.¹²  However, the function of individual PI3K isoforms and their role in repair of the established microvasculature are unclear. This family of lipid kinases includes 4 catalytic isoforms expressed in ECs (p110α, p110β, p110δ, and p110γ) that associate with a regulatory subunit to couple PI3K activity to receptor activation.¹³  PI3Ks that associate with the p85 regulatory subunit are grouped into class I_A, to couple to RTKs, whereas p110γ, the single class I_B isoform, is regulated by p101/p84 and is coupled to GPCRs. PI3Kβ uniquely signals downstream of both RTKs and GPCRs^14,15  and is uniquely regulated by rho family guanosine triphosphate–binding proteins.¹⁶  The PI3Kα isoform (encoded by the Pik3ca gene) is critically required for embryonic development, because both global and EC-restricted knockout of Pik3ca results in embryonic lethality with features of disordered vascular development.^17,18  Conversely, both global¹⁹  and endothelial-restricted PI3Kβ loss is tolerated in unstressed mice.¹⁸  Consequently, the function of PI3Kβ in the endothelium is unknown.

We observe that postnatal endothelial deletion of PI3Kβ sensitizes mice to lethal acute kidney failure after TMA injury. In vitro, PI3Kβ inactivation inhibits EC invasion of fibrin matrix. This correlates with decreased expression of the angiogenic tip cell markers apelin and Dll4 and is associated with a reduction in EC migration and proliferation. In vivo, PI3Kβ-knockdown ECs are defective in assembly of microvessel-like structures. These data identify an important role for endothelial PI3Kβ in microvascular repair following injury.

The following antibodies were purchased: anti-human DLL4 (R&D Systems), polyclonal anti-human apelin and anti-human ESM1 (Abcam), anti-human CD31 and anti-mouse CD31 (BD Pharmingen), anti-mouse fibrinogen/fibrin (Genway), and anti-apelin-36 (Phoenix Pharmaceuticals). Secondary fluorophore- or horseradish peroxidase (HRP)-conjugated species-specific antibodies were purchased from Jackson Immunoresearch. Marasmius oreades lectin A (EY Laboratories) was conjugated to saporin (Sigma-Aldrich) using a sulfo-LC-SPDP kit as described by the manufacturer (Thermo Scientific), then purified by high-performance liquid chromatography. The PI3Kβ-selective small-molecule inhibitor TGX-221 was obtained from Cayman Chemical and used in vitro at 100 nM. CellTracker green and red were from Invitrogen. VEGF-A and stromal cell–derived growth factor-1 were obtained from R&D Systems.

Immunodeficient 129S6/SvEvTac-Rag2^tm1Fwa (Rag2^−/−γc^−/−) mice were purchased from Taconic. The generation of p110β^flox/flox mice was previously described.¹⁵  B6.D2-Tg(Tek-Cre/ERT2)^1Arnd (Tie2-CreER^T2) mice were supplied by Dr Bernd Arnold²⁰  and bred against p110β^flox/flox mice to produce Tie2- CreER^T2+/−/p110β^flox/flox mice used in these experiments. Activation of the Cre recombinase deletes exons 21 and 22 from Pik3cb, resulting in expression of a messenger RNA (mRNA) that encodes an internally truncated protein that lacks lipid kinase activity.¹⁵  However, this protein is absent from tissues for reasons that are unclear at the moment and thus creates the equivalent of a p110β knockout.¹⁹  Twelve-week-old mice were treated with 2 mg tamoxifen (Sigma-Aldrich) or vehicle (oil) daily for 5 days via an intraperitoneal injection. Littermate animals were used as controls. The mice were rested for 1 week before experimentation. Expression of the mutant Pik3cb^δ21,22 mRNA was determined in kidney cortex RNA via reverse-transcription polymerase chain reaction (PCR). Kidney glomerular TMA microvascular injury was induced as described in detail elsewhere.²¹  Briefly, mice were treated with 50 μg/kg lipopolysaccharide and 200 μg/kg M oreades lectin A conjugated to the cytotoxin saporin by intra-arterial injection (toxin), retrograde via the left carotid artery. The animals were maintained according to the Canadian Council for Animal Care guidelines under a protocol approved by the Animal Care and Use Committee of the University of Alberta.

Late blood-outgrowth human primary endothelial progenitor cells (ECs) were isolated as described previously,²²  with modifications. Mononuclear cells were isolated from 120 mL peripheral blood from healthy adult donors or from single-donor leukapheresis. CD34⁺ cells were isolated using the EasySep human CD34 selection kit (STEMCELL Technologies) according to the manufacturer’s instructions. The CD34⁺ cells were seeded on fibronectin-coated 6-well tissue culture plates in endothelial growth medium 2 (EGM2; endothelial basal medium-2 [EBM2] + SingleQuots growth supplement + 10% fetal bovine serum; Lonza). After a 24-hour incubation at 37°C in 5% CO₂, nonadherent cells were discarded, and the adherent cells were fed every other day with EGM2 until late-outgrowth EC colonies appeared in culture 7 to 21 days after seeding. Mouse cardiac microvascular ECs were purchased from Cedarlane (Burlington, ON, Canada).

The serum creatinine concentration was determined by the mouse creatinine kit (Crystal Chem) as per the manufacturer’s instructions. The creatinine concentration from samples was interpolated using a calibration standard curve. The serum urea concentration was determined using a urea assay kit (BioAssay Systems) as per the manufacturer’s instructions.

Real-time PCR was performed using a model 7500 thermal cycler (Applied Biosystems). Total RNA was isolated from kidney cortex using the RNeasy mini kit (Qiagen) or from fibrin gels in vitro using TRIzol/chloroform. A total of 1 μg of total RNA was reversed transcribed into complementary DNA using qScript cDNA supermix (Quanta). The PCR primers were designed using PrimerExpress 3.0 (Applied Biosystems) software and purchased from Integrated DNA Technology (supplemental Table 1, available at the Blood Web site).

Whole-cell lysates were prepared using RIPA buffer, then proteins were resolved on sodium dodecyl sulfate (SDS) gels and transferred onto a nitrocellulose membrane (Bio-Rad). Membranes were immunoblotted using the primary antibodies, followed by HRP-conjugated secondary antibody as described previously.²³  The membranes were developed using the ECL prime kit (GE Healthcare) and visualized using MultiImage 2 and FluorChem software (Alpha Innotech).

Cell proliferation was evaluated using the Click-iT 5-ethynyl-2′-deoxyuridine flow cytometry assay kit (Invitrogen), as per the manufacturer’s instructions, and quantitated using flow cytometry. In vitro apoptosis was evaluated as described previously.²³ 

Three different small interfering RNA (siRNA) sequences against human p110β were used, and all produced similar results: 2 sequences targeted the coding region, and 1 sequence targeted the 3′ untranslated region (SI04436327, SI00085862, and SI02622214; Qiagen). Transfection of 50 nM siRNA was performed using HiPerFect (Qiagen) as described previoulsy.²³  The efficiency of gene knockdown was evaluated by western blot and/or real-time PCR after 72 hours (supplemental Figure 3).

Cell migration was assayed using a 24-well Transwell plate with 8 μm pores (Corning). The ECs (treated with p110β siRNA or control siRNA or left untreated) were seeded at a density of 10⁵ cells in the insert in the presence of EBM2. EGM2 was placed in the receiver chamber. To evaluate chemokinesis, EGM2 was placed in both chambers. The plate was incubated at 37°C for 24 hours. Migrated cells in the lower chamber were labeled with Calcein AM diluted in cell-dissociation solution (Sigma-Aldrich) for 1 hour at 37°C and 5% CO₂. The inserts were removed and the plate was read at 485 to 527 nm using a fluorescent plate reader (Fluoroskan; Thermo Scientific).

Angiogenic sprouting of ECs (treated with p110β siRNA or control siRNA or left untreated) in vitro was determined as described previously,²⁴  with some modifications. Human ECs were mixed with collagen-coated Cytodex 3 microcarrier beads (Sigma-Aldrich) and incubated at 37°C in 5% CO₂ for 4 hours with shaking every 20 minutes. The coated beads were then transferred to T25 tissue culture flasks and incubated for a further 2 hours. The EC-coated beads were washed with phosphate-buffered saline (PBS) X2 and mixed with 2 mg/mL fibrinogen (Sigma-Aldrich) and 0.15 U/mL aprotinin (Sigma-Aldrich). A total of 500 μL of the EC-coated beads and fibrinogen solution was added to 0.625 U/mL thrombin (T4648; Sigma-Aldrich) in a 24-well tissue culture plate and incubated for ∼15 min until the fibrin gelled, EGM2 was added, and sprout formation was analyzed at 18 hours. At least 30 beads from each treatment in each experiment were analyzed.

Vasculogenesis was analyzed in vivo as described previously.²⁵  Briefly, human ECs (pretreated with p110β siRNA or control siRNA or left untreated) and human aortic smooth muscle cells (Cell Applications) were mixed at a ratio 75:25 and suspended in growth factor–reduced Matrigel (BD Biosciences). The cell suspension was injected subcutaneously into Rag 2^−/−γc^−/− mice and retrieved after 1 week. The Matrigel plugs were placed in optimal cutting temperature medium, snap frozen in liquid nitrogen, and then processed for immunofluorescence microscopy.

Kidneys were collected from each killed mouse and either placed in optimal cutting temperature and snap frozen in liquid nitrogen or placed in IHC zinc fixative (BD Pharmingen). Zinc-fixed tissues were stained with hematoxylin and eosin. For immunofluorescence staining, 5 μm frozen sections were stained with the primary antibody or a nonspecific immunoglobulin G at 4°C overnight and detected with a fluorescein-labeled or HRP-labeled secondary antibody. The fluorescence signal of at least 10 glomeruli in an equatorial section per mouse was visualized by fluorescence microscopy (Quorum Wave FX-X1; Guelph, ON, Canada), and quantitated using open-source ImageJ software. Microvessels in Matrigel plugs were stained with species-specific anti-human or anti-mouse CD31. The number of microvessel-like tubes formed by human ECs was counted in 20 ×40 fields of view in a blinded fashion.

Kidney pathology and glomerular injury was evaluated as described elsewhere²⁶  by a renal pathologist (D.C.R.) blinded to the treatment group. Briefly, injury was scored if thrombosed capillary lumens, apoptotic cells, or capillary microaneurysms were detected. Similarly, acute tubular necrosis was scored on a scale per field of view (1 ≤ 25%, 2 ≤ 50%, 3 ≤ 75%, and 4 ≥ 75% of injured tubules) sampling at least 2 fields per glomerulus.

Statistical analyses were performed using PRISM software (GraphPad). The data are expressed as mean ± standard error of the mean (SEM) and analyzed using 1-way analysis of variance (ANOVA) followed by Tukey’s post hoc analysis or 2-way ANOVA where applicable. Nonparametric results were analyzed using the Mann-Whitney U test. Kaplan-Meier data were analyzed by log-rank test for trend. A P value <.05 was considered significant.

To examine the role of PI3Kβ in the mature microvasculature, we induced Cre-recombinase–mediated postnatal deletion of Pik3cb selectively in ECs (hereafter β^ECKO). Similar to the impact of embryonic Pik3cb deletion, adult mice tolerated PI3Kβ loss (supplemental Figure 1). Kidney function was comparable in the uninjured β^ECKO and p110β-wild type (β^fl/fl) littermates. A challenge with a sublethal dose of lectin-cytotoxin conjugate to selectively injure the glomerular capillary microvascular endothelium²¹  was well tolerated by β^fl/fl mice. In contrast, the β^ECKO mice became morbid on day 3 after challenge, and none survived more than 4 days after cytotoxin exposure (Figure 1A).

Histologic features of TMA injury of glomerular capillaries persisted on day 3 after cytotoxin injury among β^ECKO mice (supplemental Figure 1b-c). Glomerular injury was more frequent among β^ECKO mice than controls (Figure 1B), and consistent with the anticipated sequelae of the damaged glomerular microcirculation, histologic features of acute tubular necrosis were evident among both β^ECKO and control, cytotoxin-treated groups (Figure 1C). Notably, the kidney function of the β^ECKO mice was markedly worse compared with β^fl/fl mice at day 3 after TMA injury (Figure 1D-E).

Within the kidney cortex, increased fibrin staining of the glomerulus was evident among both β^ECKO and β^fl/fl mice after cytotoxin exposure vs PBS controls, confirming similar localization of microvascular TMA injury among these groups (Figure 2A-D). To better quantitate the extent of microvascular endothelial injury, the abundance of CD31 and Tie2 mRNAs in kidney cortex was determined (Figure 2E). Expression of each of these endothelial-specific, constitutively expressed genes was decreased in kidney cortex in cytotoxin-treated animals, reflecting EC loss from the kidney microcirculation (Figure 2E). Expression of the tubular epithelial stress genes, KIM-1 and lipocalin 2, was upregulated (data not shown), consistent with the histologic findings of tubular injury. Conversely, no injury was evident in the heart, and heart function was normal in toxin-treated β^ECKO mice (supplemental Figure 2). Taken together, these data indicate that endothelial p110β loss in the quiescent mature vasculature was tolerated. However, endothelial PI3Kβ was required to maintain organ function and survival after renal microvascular TMA injury.

We next evaluated the mechanistic role of endothelial PI3Kβ in vitro. Primary human ECs express all 3 class I_A PI3K catalytic subunit isoforms (supplemental Figure 3a-b). Selective knockdown of the p110β catalytic isoform did not alter the expression of the p110α or p110δ isoforms or of the counterregulatory phosphatase, PTEN (supplemental Figure 3a-b).

We first tested the effect of p110β loss on apoptosis of ECs. We observed no difference in caspase-3 activation in unstressed mouse or human ECs (Figure 3A-B). Tumor necrosis factor/cycloheximide challenge induced apoptosis in a similar fraction of wild-type, p110β-deficient, and p110β-inhibited (TGX-221–treated) ECs (Figure 3A-B). Moreover, VEGF stimulation elicited similar phosphorylation of Akt and downstream effectors including FoxO1/3 in TGX-221-treated or p110β-deficient ECs vs vehicle-treated ECs (Figure 4). However, GPCR-dependent responses to stromal cell–derived growth factor-1 are blocked, confirming effective disruption of PI3Kβ signaling (supplemental Figure 4). Thus, disrupted p110β signaling does not sensitize ECs to apoptosis or interrupt VEGF-mediated signaling to this important prosurvival pathway.

We directed our attention to investigate features of ECs that contribute to microvascular repair. Loss of p110β markedly impaired endothelial invasion into 3-dimensional (3D) fibrin matrices in vitro (Figure 5). The p110β-deficient ECs consistently produced both fewer sprouts and shorter sprout extensions compared with ECs treated with nonspecific siRNA or ECs without siRNA transfection (Figure 5A-B). Similarly, pharmacologic inhibition of PI3Kβ activity using TGX-221 inhibited EC sprouting (Figure 5C). Further, p110β wild-type ECs failed to rescue p110β-deficient EC sprouting in mixed cocultures (Figure 5D and supplemental Figure 5). Similar to p110β-deficient EC cultures, p110β-deficient ECs initiated few sprouts in coculture with wild-type cells. Although the p110β wild-type cells preferentially occupied the tip cell position in the mixed-EC sprouts, p110β-deficient ECs were able to contribute to either tip or stalk position. Interestingly, the angiogenic sprouting defect was rescued by loss of the counterregulatory lipid phosphatase, PTEN, in p110β/ PTEN double-deficient ECs in vitro (Figure 5E). Conversely, combined overexpression of p110β and p85α failed to promote additional EC sprouting (supplemental Figure 6). Hence, endothelial p110β loss impairs EC responses to reparative cues in a cell-autonomous fashion, but PI3 kinase-β activity is insufficient to augment growth factor-stimulated angiogenesis.

Next, we evaluated the effect of p110β knockdown on VEGF-stimulated EC proliferation or migration. p110β knockdown results in a modest reduction in the fraction of proliferating ECs (supplemental Figure 7a). In addition, the p110β-deficient ECs showed decreased chemotactic (supplemental Figure 7b) and chemokinetic (supplemental Figure 7c) migration compared with control cells. Similarly, treatment of ECs with TGX-221 decreased migration (supplemental Figure 7B-C). The loss of p110β function therefore modestly decreases EC proliferation and movement that support sprout extension in addition to the defect in sprout initiation.

To evaluate the effect of loss of p110β function on EC differentiation that enables invasion of the fibrin matrix, we evaluated the expression of the angiogenic tip cell markers APLN, DLL4, and ESM1.²⁷  Growth factor stimulation of ECs in 3D fibrin gels induced expression of the characteristic tip cell morphology and genes (Figure 6A-B). In contrast, p110β knockdown resulted in a significant decrease in Dll4 and apelin mRNA and protein expression compared with controls (Figure 6C-D). However, induced expression of the tip cell gene, ESM1, was not reduced after p110β knockdown. To determine if EC apelin expression in β^ECKO mice is altered, we immunostained tissue sections from uninjured and injured mice for apelin expression. The unstressed glomerulus exhibits minimal apelin staining, but apelin expression is induced after microvascular injury among wild-type, but not β^ECKO, mice (Figure 7). Taken together with defective sprout initiation, these data suggest that loss of p110β impairs full EC differentiation to acquire the tip cell phenotype associated with sprout invasion of fibrin matrices in vitro and in vivo.

Finally, we investigated the impact of PI3Kβ loss in human EC function and in vivo vasculogenesis. Human ECs were transfected with nonsilencing or p110β siRNA were suspended in Matrigel and injected subcutaneously into immunodeficient Rag2^−/−γc^−/− mice. Control siRNA-treated ECs in the Matrigel plugs were observed to form microvessel-like structures and to inosculate with the invading host vasculature to form chimeric blood vessels (supplemental Figure 8a). Some of these were perfused with mouse blood. However, p110β-deficient ECs were generally restricted to the center of the Matrigel plug and formed fewer microvessel-like structures, and no chimeric blood vessels were observed (supplemental Figure 8b-c). Similarly, VEGF-stimulated neoangiogenesis into subcutaneous Matrigel plugs is impaired in the β^ECKO mice (supplemental Figure 9). Taken together, these results indicate that PI3Kβ-deficient ECs, like endogenous mouse glomerular microvascular ECs, are defective in vascular cell remodeling associated with microvascular repair.

In this study, we identify an important function of PI3Kβ in the vasculature. This lipid kinase is dispensable for vascular development in the embryo and vascular homeostasis in the adult quiescent vasculature. However, endothelial PI3Kβ confers substantial protection to the adult microvasculature under stress and facilitates recovery of organ function after microvascular injury. Mechanistic studies in vitro characterize defects in repair functions of PI3Kβ-deficient ECs, including invasion of fibrin matrix, motility, and gene transcription associated with differentiation to the invasive angiogenic tip cell phenotype. Human ECs deficient in PI3Kβ also failed to remodel to form vascular structures in vivo.

The study of microvascular development during embryogenesis and tumor neoangiogenesis has discovered specialization of the EC at the tip of an angiogenic sprout.¹¹  In addition to characteristic filopodia extensions, expression of PDGFB and several other endothelial-specific genes have been identified as components of the tip cell phenotype.^27-29 

This same program may be exploited in repair of the endothelium of the established vasculature. Mechanical injury to the adult endothelium, such as the carotid artery, is repaired by migration of ECs from the edges of the wound. In vitro and in vivo, the lead ECs at the edge of a wound in the monolayer share features of the angiogenic tip cell, such as induced expression of PDGFB, but the lead ECs lack filopodia.³⁰  The role of this EC specialization in repair of injury to the established microvasculature has been reported in a model of vascular regression in response to pharmacologic inhibition of endothelial growth factor receptors. McDonald et al identify both endothelial tip cell morphologic differentiation and invasion of remnant microvessel basement membrane sheaths to pattern microvascular regrowth upon withdrawal of the growth factor receptor inhibitor.^31,32  Taken together with these observations, our data suggest a common model for microvascular repair dependent on EC specialization to the tip cell phenotype to re-endothelialize the glomerular microvasculature.

In several models of angiogenesis, VEGF stimulates differentiation of the ECs leading angiogenic sprouts and induces expression of apelin, Dll4, and ESM1.^11,27  Reduced expression of both apelin and Dll4, but not ESM1, in sprouting PI3Kβ-deficient ECs stimulated with VEGF indicates that PI3Kβ activity regulates a subset of these genes. Because apelin deficiency in mice confers impaired angiogenesis, apelin loss likely contributes to the defective sprouting and repair among PI3Kβ-deficient ECs.^33,34  On the other hand, Dll4 loss is anticipated to have the opposite effect.^35-37  However, because sprout initiation by PI3Kβ-deficient ECs could not be rescued by paracrine effects of apelin elaborated by wild-type cells, the dominant effect of PI3Kβ inactivation may also involve other events. For example, we document defective responses to GPCR signaling.

The downstream events controlled by PI3Kβ activity are largely unknown. A subtle defect in ex vivo angiogenesis in response to CXCL12, but not VEGF, was demonstrated in aortic rings after embryonic deletion of PI3Kβ.¹⁸  Defective integrin-mediated adhesion contributes to poor platelet aggregation, aggregate stability, and attenuation of thrombosis in vivo after pharmacologic inhibition of PI3Kβ in wild-type animals.^38-40  A similar defect in adhesion complex remodeling may explain modest impairment of EC migration and sprout extension in vitro and account, in part, for impaired vascular assembly in vivo. Moreover, rescue of the sprouting defect we observed in PI3Kβ-deficient ECs by concomitant PTEN loss suggests alternate PI3K isoforms may compensate for PI3Kβ loss and may explain the mild phenotype of embryonic PI3Kβ deletion. This points to PTEN inactivation as a means to overcome defective signaling for angiogenesis and repair along this pathway.

Taken together, our data identify endothelial signal transduction integrated through PI3Kβ as a novel pathway in vascular repair. Several growth factors are implicated in the defense of the glomerular endothelium, including mediators secreted by endogenous glomerular cells and both paracrine and autocrine VEGF signaling to the endothelium.^26,41-43  Collectively, our data identify the PI3Kβ signaling node as an integration point for these signals.

The authors acknowledge technical assistance from Qiu-Xia Zhang, Steve Kulak, and Yue Huang.

This work was funded in part by Canadian Institutes for Health Research operating grants MOP 53319 (A.G.M.) and MOP 58998 (G.Y.O.). G.H. was supported by a University of Alberta 75^th Anniversary studentship award. P.Z. was supported by a fellowship award from the Alberta Institute of Transplant Sciences.

Contribution: G.H. and A.G.M. designed the research; G.H., P.Z., M.F., and L.F.Z. performed the research; B.V., Z.K., and G.Y.O. contributed materials; G.H., P.Z., M.F., D.C.R., and A.G.M. analyzed the data; and G.H., B.V., G.Y.O., and A.G.M. edited the manuscript.

Conflict-of-interest disclosure: B.V. is a consultant to Karus Therapeutics (Oxford, United Kingdom) and Activiomics (London, United Kingdom). The remaining authors declare no competing financial interests.

Correspondence: Allan G. Murray, Room 275 HMRC, University of Alberta, Edmonton, AB T6G 2S2, Canada; e-mail: allan.murray@ualberta.ca.