Endothelial PKA targets ATG16L1 to regulate angiogenesis by limiting autophagy

The cAMP-dependent protein kinase A (PKA) regulates a plethora of cellular functions in health and disease. During angiogenesis, PKA activity in endothelial cells controls the transition from sprouting to vessel maturation and limits tip cell formation independently of Notch signaling. The molecular PKA targets mediating these effects remain unknown. We report a chemical genetics screen identifying endothelial-specific substrates of PKA in human umbilical vein endothelial cells (HUVEC). We identified ATG16L1, a regulator of autophagy, as novel target of PKA. Biochemical validation, mass spectrometry and peptide spot arrays revealed that PKA phosphorylates ATG16L1α at Ser268 and ATG16L1β at Ser269. The phosphorylations drive degradation of ATG16L1 protein. Knocking down PKA or inhibiting its activity increased ATG16L1 protein levels and endothelial autophagy. In vivo genetics and pharmacological experiments demonstrated that autophagy inhibition partially rescues vascular hypersprouting caused by PKA deficiency. We propose that endothelial PKA activity restricts active sprouting by reducing endothelial autophagy through phosphorylation of ATG16L1.


Introduction
Angiogenesis is the process of new blood vessels formation from pre-existing vessels via sprouting and remodeling. Blood vessels are crucial for tissue growth and physiology in vertebrates since they are the pipelines for oxygen and nutrients supply and for distribution of cells, such as of the immune system. Inadequate vessel formation and maintenance as well as abnormal vascular remodeling cause, accompany or aggravate many disease processes including myocardial infarction, stroke, cancer, and inflammatory disorders (Geudens & Gerhardt, 2011;Potente, Gerhardt, & Carmeliet, 2011).
Recently, we identified endothelial cAMP-dependent protein kinase A (PKA) as a critical regulator of angiogenesis during vascular development in vivo. Inhibition of endothelial PKA results in hypersprouting and increased numbers of tip cells, indicating that PKA regulates the transition from sprouting to quiescent vessels (Nedvetsky et al., 2016). However, the PKA targets mediating these effects remained unknown.
Here, we established a chemical genetics approach based on the mutation of a ''gatekeeper residue'' of the ATP-binding pocket of a protein kinase to identify endothelial-specific substrates of PKA. We verified autophagy-related-protein-16-like 1(ATG16L1) as a substrate of endothelial PKA and revealed that phosphorylation of ATG16L1 facilitates its degradation. In vivo experiments showed that autophagy inhibition partially rescued the hypersprouting and increased tip cell numbers caused by PKA deficiency.

Screen for novel substrates of endothelial PKA
In order to identify direct substrates of endothelial PKA, we employed a chemical genetics approach (Allen, Lazerwith, & Shokat, 2005). The ATP-binding pocket of kinases contains a conserved "gatekeeper residue", which in wild type (WT) kinases is usually a methionine or phenylalanine. Engineered analogue specific (AS) kinases replace this "gatekeeper residue" with a smaller amino acid (glycine or alanine), enabling the AS-kinases to accept ATP analogues (or ATPγS analogues) that are modified at the N 6 position with bulky groups as (thio-)phosphodonors. In contrast, WT-kinases poorly use these analogues. Once the substrates are thiophosphorylated by AS-kinases, they can be further alkylated and therefore recognized by a thiophosphate ester-specific antibody (Alaimo, Shogren-Knaak, & Shokat, 2001;Allen et al., 2007Allen et al., , 2005Banko et al., 2011). To generate AS-PKACα, we mutated the methionine 120 to a glycine residue. Testing seven different variants of N 6 -substituted bulky ATPγS analogues, we identified 6-cHe-ATPγS as the best thiophosphodonor for AS-PKACα substrates in HUVEC lysates ( Figure S1).
To identify endothelial substrates of PKACα, HUVECs expressing WT-PKACα or AS-PKACα were lysed in kinase lysis buffer(KLB) and the thiophosphorylation reaction with 6-cHe-ATPγS was performed. After alkylation with p-Nitrobenzyl mesylate (PNBM), thiophosphorylated proteins were immunoprecipitation with rProtein G Agarose beads coupled to the thioP antibody ( Figure 1A). For quality control, one thirtieth of the protein on agarose beads was eluted for Western blot analysis, and the same amount was used for gel silver staining ( Figure 1B). The rest was subjected to mass spectrometry analysis. Two independent experiments were performed. Candidate endothelial PKA targets were identified as peptides that were at least 2-fold (log ratio(AS/WT)>1) enriched in the AS-PKACα samples compared to WT-PKACα samples in both experiments.
Thirty proteins with at least 8-fold enrichment (log ratio(AS/WT)>3) are listed according to their log ratio(AS/WT) value in Table 1, presenting the most likely direct substrates of PKA in this screen.
To validate novel candidate PKACα substrates, we overexpressed WT-PKACα or AS-PKACα together with flagor GFP-tagged candidate proteins in 293T cells, and used the 6-cHe-ATPγS as thiophosphate donor to thiophosphorylate substrates in lysates as described above. Lysate immunoprecipitation was carried out with M2 anti-flag beads or anti-GFP antibody coupled agarose beads, and the immune complexes were probed by western blot using thiophosphate antibody. The known PKA substrate NFATC1 served as a positive control. Five selected new candidate proteins (PPP1R12C, ATG16L1α, DDX17, ANKRD40 and ATG5) out of ninety-seven proteins were tested; four of these five proteins were confirmed to be thiophosphorylated by AS-PKACα, indicating that they are indeed direct substrates of AS-PKACα (figure 1C). Only ATG5 was not thiophosphorylated by AS-PKACα in the validation of the screen ( figure 1C). Bioinformatic analysis of ATG5 amino acids sequence also failed to identify a consensus PKA substrate motif (R-R/K-X-S/T;K/R-X 1-2 -S/T) (Kennelly & Krebs, 1991). Since ATG5 directly binds to ATG16L1 (Matsushita et al., 2007;Noboru Mizushima, Noda, & Ohsumi, 1999), it likely co-precipitated with ATG16L1 in our screen.
ATG5 and ATG16L1 are conserved core components of the autophagy process, and PKA activity has been shown to negatively regulate autophagy in S. cervisiae and mammalian cells through phosphorylation of ATG1/ULK1(Noboru Mizushima, 2010). ATG16L1 however has not previously been identified as a PKA target, prompting us to further investigate this interaction and the potential regulatory role of PKA and autophagy in endothelial sprouting.

PKACα phosphorylates ATG16L1α at S268 and ATG16L1β at S269
To identify the PKACα phosphorylation sites in ATG16L1, we spot-synthesized 25mer overlapping peptides that cover the entire ATG16L1 protein.

PKA-deficiency
ATG16L1 is an important component of the ATG16L1-ATG5-ATG12 protein complex, required for LC3 lipidation and autophagosome formation. Both LC3 lipidation and autophagosome formation represent essential steps in autophagy (Kuma, Mizushima, Ishihara, & Ohsumi, 2002;Levine & Kroemer, 2008;Matsushita et al., 2007;N. Mizushima, 2003;Noboru Mizushima et al., 1999). Accumulation of ATG16L1 upon PKA knock down resulted in increased levels of the positive autophagy marker LC3II whilst reducing the negative autophagy marker p62 in HUVECs (figure 3G). Since ATG16L1 protein levels in endothelial cells isolated from dnPKA mice were also increased, we hypothesized that increased autophagy in endothelial cells may contribute to the vascular phenotype in these mice.

Discussion
Our chemical genetics screen and biochemical analysis identified ATG16L1 as novel target of PKA activity in endothelial cells. The combined results demonstrate that PKA activity inhibits autophagy in culture human vein endothelial cells (HUVEC) via the phosphorylation of ATG16L1, which accelerates its degradation. In cultured bovine aortic endothelial cells, induction of autophagy by overexpression of ATG5 has been shown to promote in vitro vascular tubulogenesis, whereas ATG5 silencing suppressed this morphogenic behaviour (Du et al., 2012). In mice, inhibition of autophagy by bafilomycin, or genetic beclin heterozygosity as well as ATG5 knockout impairs angiogenesis post myocardial infarction, whereas the angiogenic factor AGGF1 enhances therapeutic angiogenesis through JNK -mediated stimulation of endothelial autophagy (Lu et al., 2016). Angiogenesis during tissue regeneration in a burn wound model also relied on induction of endothelial autophagy, by driving AMPK/AKT/mTOR signaling (Liang et al., 2018), together suggesting that autophagy regulation may represent a critical determinant of the extent of vascular sprouting. The identification of ATG16L1 as a direct target of PKA therefore raises the hypothesis that the dramatic hypersprouting phenotype in dnPKA iEC mouse retinas deficient in endothelial PKA activity may result from exuberant endothelial activation of autophagy. Both chemical inhibition of autophagy and genetic endothelial inactivation of ATG5 partially normalized the hypersprouting phenotype in dnPKA iEC mice, suggesting that indeed the activation of autophagy in dnPKA iEC mice contributes to vascular hypersprouting. However, the failure to fully normalize vascular patterning by autophagy inhibition indicates that additional PKA targets and mechanisms may be involved. Our mass spectrometry analysis identified a list of presumptive endothelial PKA substrates, which will potentially also be involved in angiogenesis. For example, we identified RAPGEF2 as a candidate target, deficiency of which causes embryonic lethality at E11.5 due to yolk sac vascular defects (Satyanarayana et al., 2010), very similar the yolk sac phenotype in dnPKA iEC embryos (Nedvetsky et al., 2016). Similarly the potential target Rock2 has a well known role in regulating endothelial functions in angiogenesis Montalvo et al., 2013;Seto, Chang, Jenkins, Bensoussan, & Kiat, 2016;Shimizu et al., 2013). Further studies will need to validate all the listed targets and establish which of these exert critical endothelial functions, and under what conditions.
An alternative explanation for the partial rescue of the dnPKAiEC phenotype by inhibition of autophagy could lie in additional functions of ATG16L1 itself. Although ATG16L1 plays an essential role in autophagy, and is part of a larger protein complex ATG16L1-ATG5-ATG12 that is necessary for autophagy, ATG16L1 is also involved in the production of inflammatory cytokines IL-1β and IL-18 and exerts antiinflammatory functions during intestinal inflammation (Cadwell et al., 2008;Diamanti et al., 2017;Saitoh et al., 2008;Sorbara et al., 2013). IL-1β promotes angiogenesis by activating VEGF production during tumor progression (Carmi et al., 2013;Voronov et al., 2003), while IL-18 suppresses angiogenesis in cancer (Cao, Farnebo, Kurimoto, & Cao, 1999;Xing et al., 2016;Yang, Chen, Lu, Li, & Lin, 2010). How ATG16L1 regulates angiogenesis through inflammatory cytokines and whether this regulation operates downstream of PKA activity in vivo requires further investigation.
Intriguingly, our rescue experiments show that in wild type mice, inhibition of autophagy has no significant effect on developmental retinal angiogenesis. This could indicate that autophagy in developmental angiogenesis, unlike in pathological angiogenesis and post-ischemic tissue responses, is not very active.
Although, to our knowledge, this is the first identification of ATG16L1 as a PKA target and the first indication that PKA regulates autophagy in endothelial cells, PKA has previously been identified as regulator of autophagy. For example, PKA reduces autophagy through phosphorylation of ATG13 in Saccharomyces cerevisiae (Hundsrucker et al., 2006), and through phosphorylation of LC3 in neurons (Cherra et al., 2010). In our research, ATG16L1 was identified as a novel direct PKA substrate in endothelial cells, but not ATG13 or LC3. Mechanistically, the phosphorylation of ATG16L1 by PKA accelerates its degradation, and consequently decreases autophagy levels in endothelial cells. The finding of different components of the autophagy pathway as targets of PKA identified in yeast and various vertebrate cell populations raises the intriguing possibility that although the principle regulatory logic of PKA in autophagy is conserved, different protein targets mediate this effect in different cells or organisms. In addition, or alternatively, this regulation carries multiple levels of redundancy, and the individual studies simply identify the most prevalent targets within the respective cell types. The fact that also ATG16L1 comes in two splice variants that are both targeted by PKA in endothelial cells lends some strength to this idea.
ATG16L1 can itself be regulated by multiple phosphorylation events by distinct kinases, with opposing effects on protein stability and autophagy. ATG16L1 can be phosphorylated at Ser139 by CSNK2 and this phosphorylation enhances its interaction with the ATG12-ATG5 conjugate (Song et al., 2015). IKKα promotes ATG16L1 stabilization by phosphorylation at Ser278 (Diamanti et al., 2017). In addition, phospho-Ser278 has similar functions as phospho-Thr300, since both phospho-mutants ATG16L1 S278A and ATG16L1 T300A accelerate ATG16L1 degradation by enhancing caspase 3 mediated ATG16L1 cleavage (Diamanti et al., 2017;Murthy et al., 2014). In contrast, our finding suggest that the PKA target sites Ser268 in ATG16L1α (or Ser269 in ATG16L1β) work in the opposite way of Ser278 and Thr300; ATG16L1α S268A (and ATG16L1β S269A ) are more stable than ATG16L1 WT .
Furthermore, PKA deficiency also stabilizes ATG16L1 in endothelial cells in vivo and in vitro. Taken together, it appears that the different phosphorylation sites of ATG16L1 play different roles in fine tuning protein stability under the influence of alternative upstream kinases, and thereby adapt autophagy levels. Given the increasing insights into the role of autophagy in cell and tissue homeostasis and in disease, it will be of great interest to investigate whether the newly identified regulation by PKA extends beyond developmental angiogenesis into pathomechanisms associated with endothelial dysfunction.
Finally, on a technical note, the chemical genetics approach developed by Shoktat and colleagues (Alaimo et al., 2001;Allen et al., 2007Allen et al., , 2005 has successfully been used in other cell types, but to our knowledge, this is the first report on direct endothelial PKA targets. Our initial attempts using published cell lysate conditions based on RIPA buffer however failed to identify differences in thiophosphorylation when comparing AS-PKA expressing cells to WT-PKA expressing cells. Our buffer optimization revealed that RIPA buffer limits the activity of PKA kinase, whereas our new kinase lysis buffer (see methods) allows effective substrate phosphorylation, thus giving rise to strong signals in AS-PKA samples. This optimization will hopefully be valuable for researchers aiming to utilize this approach for additional chemical genetic kinase substrate screens in the future.  Research.

Chemical Genetic Screen and validation of PKA substrates
HUVECs ( For validation of the identified substrates, 293T cells (1.5x10 6 on a 6cm dish) were transfected with 0.5g pRRL PKACa WT or pRRL PKACaM120G and 1.5 g of indicated candidate substrate using X-tremeGENE HP DNA transfection reagent (Roche). 30 hours after transfection, cells were stimulated with Sp-8-CPT-cAMPS (Biolog) for 10 minutes, lysated and treated as described above.

Silver staining
Silver staining was performed using Pierce™ Silver Stain Kit (Thermo Fisher,24600) according to the manufacturer's protocol. Briefly, the SDS-page gel was washed in ultrapure water and fixed by fixing solution(30% ethanol,10% acetic acid) for 30 minutes. After incubating the gel in sensitizer working solution(provided in the kit) for 1 minutes, silver stain enhancer(provided in the kit) was added for another 5 minutes.
Subsequently, the gel was incubated with developer working solution(provided in the kit) for 2-3 minutes, before stopping the reaction with stop solution(provided in the kit).

Mass spectrometry to identify new PKA substrates and phosphorylation sites
For mass spectrometric analysis to identify new PKA substrates, samples were prepared by chemical genetical approach as described above, each sample was run on a stacking SDS-PAGE collecting all proteins in a single band. After coomassie blue staining, the minced gel pieces were digested with trypsin based on Shevchenko et al. (Shevchenko, Tomas, Havliš, Olsen, & Mann, 2007)  The minimum score for modified peptides was set to 40. The S-lens RF level was set at 50 and we excluded precursor ions with single, unassigned and charge states above five from fragmentation selection.
Fmoc-protected amino acids (Merck Millipore) and amino-modified acid-stable cellulose membranes with PEG-spacers (Intavis) were used for peptide spots synthesis on an Intavis ResPep-SL device.

Animal procedures
All animal experimental procedures were approved by the Institutional Animal Care and Research Advisory Committee of the University of Leuven and performed according to the European guidelines. Following mouse strains were used: Prkar1a tm2Gsm (Willis, Niswender, Su, Amieux, & McKnight, 2011), Tg(Cdh5cre/ERT2) 1Rha (Wang et al., 2010) and ATG5 flox/flox (Hara et al., 2006). All animals used in the experiments were of mixed N/FVB x C57/Bl6 background. Comparisons have been done between littermates only. For chloroquine rescue retinal angiogenesis experiment, pups were intraperitoneally injected with 50 l of 1mg/ml tamoxifen from postnatal day one (P1) to P3, and 100ul l of 1.25mg/ml chloroquine or PBS from P1 to P5. Mice were euthanized at P6, and dissection and staining of the retinas were performed as described below. For endothelial cells isolation, pups were intraperitoneally injected with 75 l of 1mg/ml tamoxifen daily from P7 until P10 and the mice were euthanized at 8 weeks and endothelial cells were isolated as described below.

Retinal angiogenesis assay
To analyse retinal angiogenesis , the procedures of isolation and staining of the retinas were performed as published (Pitulescu, Schmidt, Benedito, & Adams, 2010).
Briefly, retinas were dissected in PBS and blocked/permeabilized in retina blocking buffer (1% BSA and 0.3% Triton X-100 in PBS) for 1-2 hours at room temperature.

Endothelial cell isolation from liver or lung
Livers or lung lobes were collected in dry 10cm dishes and minced finely with blades for one minute, and then incubated in 25ml of pre-warmed Dulbecco modified Eagle medium (4.5 g/L glucose with L-glutamine) containing 2 mg/mL collagenase (Invitrogen) in 50 ml tubes, gently shaking for 45 minutes at 37°C. Suspensions were passed through a 70μm cell strainer (VWR) and cells were spun down at 400g for 8 minutes at 4°C. Pellets were resuspended in 10 ml Dulbecco modified Eagle medium containing 10% FBS, 50U/ml penicillin and 50μg/ml streptomycin, passed through 40μm Nylon cell strainer (BD Falcon, Cat. No. 352340) and centrifuged at 400g for 8 minutes at 4°C. Cells were resuspended in cold DPBS (1ml/lung and 2ml/liver), added to sheep anti-Rat IgG-coupled Dynabeads (Invitrogen) preincubated with purified Rat Anti-Mouse CD31 (BD Pharmingen) and incubated at 4°C. for 20 min The beads were separated using a magnetic particle concentrator (Dynal MPC-S; Invitrogen)and washed with cold DPBS with 0.1% BSA. This washing step was repeated 5 times after which cells were lysed in RIPA buffer for Western Blotting.

Statistical analysis
Statistical analyses were performed using GraphPad Prism 7. The one-way ANOVA was used to compare more than two experimental groups.  thiophosp hate with the el ectrophilic tag p-nitroben zylmesylat e (PNBM) is concomitant with format ion of undesired al kylation products resulting from derivatization of cysteine thiols. To specifically detect and purify thiophosphate reaction products, we developed polyclonal and monoclonal antibodies that discriminate thiophosphateestersfrom cystei nealkylation products(thioethers). In our first rep ort 15 , we had used this approach to detect substrates of the ser ine-threonine kinase, Cdk1. To further devel op this strategy and ex pand the portion of the kinome tract able with this approach, we analyzed each of the reaction st ep s (Fi g. 1a) for speci fici ty and ap plicability to many kinase-substrate pairs. We scr eened diver se kinases for the ability to thiophosp horylate thei r protei n su bstratesan d found that the vast majority of kinasesused ATPgS as a phosphodonor. We synthesized orthogonal A*TPgS analogs and found them to be pref er red su bst rates for AS kinases, permitting delivery of thiophosphat e to individual kinase su bstrates. The resulting labeled kinase su bst rates, which contained modified ser ine, threo nine and tyrosine residues in the context of diverse kinase consensu s motifs, wer e all recognized by new thiophosp hate ester -speci fic antibodies. Our initial hapten design presen ted the thiophosphat e ester modification on a threo nine backbone 15 ; the el ici ted hapten -speci fic antibodies, however , suffered from ei ther low yield of sp ecificantibody or low af finity, and wer e not ex pect ed to recognize mod-ified tyrosine resi due contai ning only the m the immune respons el icited rabbit polyclo phate ester-specific, nopreci pitation. Appl approach allowed us (Erk2) , which was ex RESULTS Ge nerat ion of thiop Weel icited rabbit a-h hapten conjugate, con with a three-carbon Fi g. 1b). Affinity-pu c-Jun-GST contai nin modifications (Fi g. nate thiophosphate alkylat ion products. (a-hap ten -I gY) recog not they had been osp hate with the el ectrophilic tag p-nitroben zylmesylat e ) is concomitant with format ion of undesired al kylation cts resulting from derivatization of cysteine thiols. To specidetect and purify thiophosphate reaction products, we ped polyclonal and monoclonal antibodies that discriminate osphateestersfrom cystei nealkylation products(thioethers). our first rep ort 15 , we had used this approach to detect ates of the ser ine-threonine kinase, Cdk1. To further devel op rategy and ex pand the portion of the kinome tract able with proach, we analyzed each of the reaction st ep s (Fi g. 1a) for ci ty and ap plicability to many kinase-substrate pairs. We ed diver se kinases for the ability to thiophosp horylate thei r n su bstratesan d found that the vast majority of kinasesused as a phosphodonor. We synthesized orthogonal A*TPgS s and found them to be pref er red su bst rates for AS kinases, tting delivery of thiophosphat e to individual kinase su b-. The resulting labeled kinase su bst rates, which contained ed ser ine, threo nine and tyrosine residues in the context of e kinase consensu s motifs, wer e all recognized by new osp hate ester -speci fic antibodies. Our initial hapten design ted the thiophosphat e ester modification on a threo nine one 15 ; the el ici ted hapten -speci fic dies, however , suffered from ei ther eld of sp ecificantibody or low af finity, er e not ex pect ed to recognize mod-ified tyrosine resi dues. Her e we designed a new hapten (Fi g. 1b) contai ning only the minimal desi red epitope in an ef fort to focus the immune response on the thiophosp hate ester moiety. The el icited rabbit polyclonal an d monoclonal antibodiesar ethiosphosphate ester-specific, contex t-indep en dent and capable of immunopreci pitation. Application of thissem isynthet ic immunoaffinity approach allowed us to purify the direct substrates of an ASkinase (Erk2) , which was ex pressed at en dogenous levels.

RESULTS
Ge nerat ion of thiophosphat e ester-specificIgG Weel icited rabbit a-hapten -immunoglobulin gamma (IgG) with a hapten conjugate, consi st ing of a p-nitroben zylthiophosphateester with a three-carbon linker to key hole limpet hemocyanin (KLH; Fi g. 1b). Affinity-purified polycl onal an tibodies only recognized c-Jun-GST contai ning both thiophosp horylation and PN BM modifications (Fi g. 1c), ver ifying that the an tibodies discr iminate thiophosphate alkylat ion products from all other possible alkylat ion products. In contrast , an tibodies raised in chicken (a-hapten -I gY) recognized PN BM al kylated protei ns whether or not they had been thiophosphorylated (Supplem entary Fi g.     AS-PKACα and purified twice using M2 beads and thioP antibody coupled beads, followed by mass spectrometric analysis. LC-MS/MS spectra of the PKA-phosphorylated ATG16L1α tryptic peptide pSVSSFPVPQDNVDTHPGSGK and ATG16L1β tryptic peptide RLpSQPAGGLLDSITNIFGR. The results demonstrate that PKA phosphorylated ATG16L1α at S268 and phosphorylated ATG16L1β at S269.  Figure 1C, then treated with or without 1% TFA at 37°C for 4 hours.