- Open Access
Non-canonical regulation of phosphatidylinositol 3-kinase gamma isoform activity in retinal rod photoreceptor cells
© Gupta et al.; licensee BioMed Central. 2015
Received: 23 November 2014
Accepted: 20 January 2015
Published: 3 February 2015
Phosphatidylinositol 3-Kinases (PI3Ks) are a family of lipid kinases that phosphorylate the D3-hydroxyls of the inositol ring of phosphoinositides, and are responsible for coordinating a diverse range of cellular functions. A canonical pathway of activation of PI3Ks through the interaction of RA-domain with Ras proteins has been well established. In retinal photoreceptors, we have identified a non-canonical pathway of PI3Kγ activation through the interaction of its RA-domain with a putative Ras-like domain (RLD) in alpha subunit of cyclic nucleotide-gated channel (CNGA1) in retinal rod photoreceptors.
The interaction between PI3Kγ and CNGA1 does not appear to play a role in regulation of CNG channel activity, but PI3Kγ uses CNGA1 as an anchoring module to achieve close proximity to its substrate to generate D3-phosphoinositides.
Our studies suggest a functional non-canonical PI3Kγ activation in retinal rod photoreceptor cells.
Phosphatidylinositol 3-Kinases (PI3Ks) are a family of lipid kinases that catalyze the phosphorylation of D3-hydroxyls in the inositol head group and generate several phosphorylated phosphoinositides . The D3-phosphorylated products further act as membrane tethers to several phospholipid binding proteins which regulate a diverse range of cell functions, including proliferation, cell survival, degranulation, vesicular trafficking, and cell migration . There is a highly conserved Ras-associating (RA) domain present in different isoforms of class I and class II PI3Ks . A canonical pathway of the activation of PI3K through interaction of its RA-domain with Ras protein has been well established . In retinal photoreceptors, we found a non-canonical pathway of PI3Kγ activation through interaction of its RA-domain with a putative Ras-like domain (RLD) in the cyclic nucleotide-gated channel alpha subunit (CNGA1) in retinal rod photoreceptors.
Photoreceptor cyclic nucleotide-gated (CNG) channels are critical elements in phototransduction and light adaptation [3,4]. They are responsible for generating the light response in photoreceptors and are directly and co-operatively gated by the availability of cGMP . Interestingly, PI3K-generated phosphoinositides have also been shown to modulate the olfactory [5,6] and cone photoreceptor CNG  channels. However, the interaction between PI3Kγ and CNGA1 does not appear to play a role in regulation of CNG channel activity.
This first report of PI3Kγ interaction with a Ras-like domain in the nucleotide binding region of CNGA1 provides a framework within which detailed kinetic and structural studies can be carried out to establish the detailed mechanism and functional relevance of this interaction. The functional regulation of PI3Kγ by G-protein coupled receptors , Ras , and CNG channels through its RA domain establishes the important position of PI3Ks in signal transduction and provides further evidence of how closely the cell signaling network is integrated with phototransduction.
PI3K interacts with CNGA1
We previously reported that the C-terminal region of CNGA1 displays 50-70% tertiary structural similarity towards Ras proteins . We named this region the Ras-like (RLD) domain. Our studies also suggest that growth factor receptor-bound protein 14 (Grb14), a Ras-associating (RA) domain-containing protein, binds to CNGA1 and modulates channel activity . Our membrane yeast two-hybrid screens also identified PI3K as a CNGA1 interaction partner (data not shown).
Expression of PI3K isoforms in retinal rod photoreceptors
CNGA1 is expressed mainly in the plasma membranes of the rod photoreceptor outer segments . To determine whether class I PI3K isoforms are expressed in rod outer segments (ROS), we immunoblotted ROS from dark- and light-adapted rats with anti-PI3Kα, anti-PI3Kβ, anti-PI3Kγ, anti-CNGA1, and anti-opsin antibodies. We found expression of all class I PI3K isoforms in ROS, irrespective of dark- or light-adaption (Figure 1D-F). The ROS specific proteins, CNGA1 and opsin, were used as internal controls (Figure 1C, G). The expression profile of PI3Kδ in ROS was not evaluated, as its expression is restricted exclusively to leukocytes .
PI3Kγ is associated with the CNGA1 in light-adapted ROS
CNGA1-associated PI3Kγ activity is independent of dark and light conditions, as well as insulin receptor impairment in rods
PI3Kγ interacts with CNGA1 through its RA domain
Effect of RA-PI3Kγ on channel activity
We observed that the RA domain of CNGA1 exhibits considerable tertiary structural homology with the RA domain of Grb14, which is known to inhibit CNGA1 activity, as shown in Figure 6C. RA-Grb14 inhibits CNGA1 through the involvement of Glu,180, 181 which is present at the β-turn with a unique orientation . Our in silico studies identified that, in RA-PI3Kγ, these Glu residues are present towards the β-sheet extension and their orientation does not allow them to interact adequately with the CNGA1 in a manner that would facilitate modulation of its activity (Figure 6D). These in silico predictions were verified by site-directed insertional mutagenesis, where two Glu residues were introduced at 300,301 positions between an Asn and a Gly residue (Figure 6E). This modification imparts PI3Kγ with the capability to inhibit CNGA1 significantly, as predicted from the model and as shown experimentally in the Figure 6F. To confirm the expression of proteins, immunoblotting was performed with anti-CNGA1 and anti-Myc antibodies. The results show comparable levels of protein expression under experimental conditions (Figure 6G). Together, these experiments suggest that, unlike Grb14, PI3Kγ may use CNGA1 as an adapter for activation, rather than playing a role in the modulation of CNG channel activity.
Identification of the interaction constraints of CTR-CNGA1 and RA-PI3Kγ
Examination of the RA domains in PI3Kγ, PI3Kα, and Grb14 suggests a molecular divergence among the RA domains in intracellular signaling proteins. We attributed this difference between RA-PI3Kγ and RA-Grb14’s inhibitory effects on the channel to the orientation of the Glu residues in the β-turn. We confirmed this hypothesis by site-directed insertional mutagenesis in which two Glu residues were introduced at the 300,301 positions between an Asn and a Gly residue, and found a significant inhibition of channel activity by RA-PI3Kγ.
The fraction of the total PI3Kγ bound to CNGA1 in photoreceptors is unknown. Nevertheless, it must constitute an important proportion of the total pool, as the alternative Ras-mediated pathway is insufficient to drive membrane translocation of PI3Kγ, a necessity for its activation . Alternatively, this binding might complement the Ras-PI3Kγ interaction pathway by accepting the allosterically activated PI3K from Ras and enabling it to stay bound at the membrane. The interaction of CNGA1 with PI3Kγ also indicates that CNGA1 may contribute towards the existence of an alternative pathway to Ras activation of PI3Kγ, strengthened by the fact that both CNGA1 and Ras proteins are nucleotide binding proteins showing similar structural folds. The relative contribution of these controlling mechanisms, as well as the fraction of the PI3Kγ participating in each of the regulatory interactions, is yet to be determined in vivo.
The crystal structure of PI3Kγ in complex with Ras has revealed that K251, 254, 255, and 256 are critical residues for establishing an interaction with Ras . Lys140 of Grb10  and Grb14  have been shown to be involved in the interaction of the RA domain with Ras proteins and the Ras-like domain of CNGA1. Lys140 of Grb14 corresponds to the Lys251and Lys255 of the RA-PI3Kγ, but is present amongst a highly positively charged cloud along the β-turn, imparting it with more leverage to interact and associate with proteins carrying negatively charged surface pockets. These types of electrostatic interactions comprise the basis of PI3Kγ’s interaction with Ras proteins . These interactions may also be a good model for the weak or potentially transient interaction of RA-PI3Kα with CNGA1. In addition, these interactions might contribute to conformational changes in PI3Kγ, similar to those caused by Ras protein binding  and leading to altered PI3Kγ activity and membrane recruitment. The structural folds of the Ras-associating domain of PI3Kγ are also similar to Raf and RalGDS, two other effectors of Ras proteins [18,19].
The phospholipids have been shown to play significant roles in the negative modulation of CNG channels [20-23]. Even though PI3Kγ may not have a direct effect on CNG channel modulation, PI3Kγ-generated PI-3,4,5-P3 may indirectly influence the channel activity. Accordingly, PI-3,4,5-P3 has been previously shown to inhibit the action of olfactory  and cone CNG  channels. Our studies thus suggest a functional non-canonical PI3Kγ activation in retinal rod photoreceptor cells.
Materials and methods
Cell lines and culture conditions
HEK-293 T cells were maintained at 37°C in DMEM medium containing 10% (v/v) FBS. Approximately 2.5 × 105 cells were seeded in each 60-mm culture dish 12–18 h before transfection. Calcium phosphate-mediated DNA transfection was performed using each of the plasmids containing the cDNA of interest . Cells were harvested for experiments ~48 h post-transfection.
Plasmids and DNA
The regions encompassing the RA domain of PI3Kα (Uniprot id. P42336, residues 132–314) and PI3Kγ isoform (P48736, residues 141–311; Open Biosystems, Huntsville, AL, USA) were amplified from their respective full-length PI3Ks and cloned into Myc-tagged pCDNA3 vector with Bam H1 and Xho1 sites. The CNGA1 channel antibodies were a kind gift from Dr. Robert Molday, University of British Columbia (Canada). The Myc-tagged CNGA1 construct was generated from the amplification of pCDNA3-CNGA1 with sense (GAA TTC ACC ATG GAG CAA AAA CTC ATC TCA GAA GAG GAG GAT CTG ATG AAG AAA GTG ATT ATC AAT ACA TGG CAC ) and antisense (CTC GAG TCA GTC CTG TGT AGA GTC TGT GGG CCC ACT TTC) primers. The NH2-terminal FLAG-tagged C-terminal region of CNGA1 (CTR-CNGA1,residues 483–690) was cloned into pCDNA3 using sense (GAA TTC ACC ATG GAT TAC AAG GAT GAC GAC GAT AAG GCT TGG TCT GTT GGT GGAG) and antisense (CTC GAG TCA GTC CTG TGT AGA GTC TGT GGG CCC ACT TTC) primers. The cDNA encoding the full-length CNGA1 was cloned into the NcoI site of pDHB1 membrane-based yeast two-hybrid bait vector. Full-length Grb14, RA-PI3Kγ, and RA-PI3Kα isoforms were cloned as BamH1/Sal1 into pDL2-Nx prey vector. The positive (pMBV-Alg5 and pAlg5-NubI) and negative (pMBV-Alg5 and pAlg5-NubG) plasmids were obtained from Dualsystems Biotech AG (Switzerland). The membrane yeast two-hybrid assay was carried out using the Dual hybrid kit (Dualsystems Biotech AG, Switzerland). The membrane yeast two-hybrid screen was used in the S. cerevisiae strains DSY-1. Site-directed mutagenesis (SDM) was carried out according to the method described earlier . The primers used for SDM were PI3Kγ-RA (E 300, 301; sense: TGC CTC AAG AAC GAA GAA GGA GAA GAG ATT; antisense AAT CTC TTC TCC TTC TTC GTT CTT GAG GCA), CTR-CNGA1(DDE 577, 578, 581 NNQ; sense: TCA AAA AAT ACC CTC ATG CAA GCT CTA ACT; antisense: AGT TAG AGC TTG CAT GAG GTT ATT TTT TGA), CTR-CNGA1 (D577N; sense: CTC TCA AAA AAT GAC CTC ATG GAA GCT CTA; antisense: TAG AGC TTC CAT GAG GTC ATT TTT TGA GAG), CTR-CNGA1 (D577,578 N; sense: CTC TCA AAA AAT AAC CTC ATG GAA GCT CTA; antisense: TAG AGC TTC CAT GAG GTT ATT TTT TGA GAG), CTR-CNGA1 (E581Q; sense: AAA GAT GAC CTC ATG CAA GCT CTA ACT GAG; antisense: CTC AGT TAG AGC TTG CAT GAG GTC ATC TTT). After sequencing, the mutant cDNAs were excised from the sequencing vector and cloned into pCDNA3 mammalian expression vector.
All animal work was performed in strict accordance with the NIH Guide for the Care Use of Laboratory Animals, and the Association for Research in Vision and Ophthalmology on the Use of Animals in Vision Research. All protocols were approved by the IACUC of the University of Oklahoma Health Sciences Center and the Dean McGee Eye Institute. The generation of rod-specific conditional IR knockout mice has been reported previously . A breeding colony of Albino Sprague–Dawley rats is maintained in our vivarium in cyclic light (12 h on/ off; ~300 lux). Experiments were carried out on male and female rats and mice.
Membrane-based yeast two-hybrid assay
Yeast transformations were performed using lithium acetate . We used the split ubiquitin system to genetically investigate interactions between membrane proteins [27,28] to probe the possible intercommunications between full-length CNGA1 and either of the full-length Grb14, RA-PI3Kα, or RA-PI3Kγ, using the transcription factor protein-LexA-VP16 (PLV) as the reporter molecule. The filter paper β-galactosidase assay for the detection of interactions and liquid culture assay for their quantification was performed as described previously .
PI3K activity assays
Mouse and rat photoreceptor outer segments (ROS) were prepared as described previously . Enzyme assays were performed on solubilized ROS, essentially as described previously . Briefly, assays were performed directly on anti-CNGA1 immunoprecipitates in 50 μl of the reaction mixture containing 0.2 mg/ml PI-4, 5-P2, 50 μM ATP, 0.2 μCi [γ32P] ATP (Perkin Elmer, MA, USA), 5 mM MgCl2, and 10 mM HEPES buffer (pH 7.5). The reaction was performed for 30 min at 25°C and was stopped by the addition of 100 μl of 1 N HCl followed by 200 μl chloroform-methanol (1:1 v/v). Lipids were extracted and resolved on TLC plates (silica gel 60) with a solvent system of 2-propanol/2 M acetic acid (65/35, v/v). The plates were coated in 1% (w/v) potassium oxalate in 50% (v/v) methanol and then baked in the oven at 100°C for 1 h before use. TLC plates were exposed to X-ray film overnight at −70°C. Radioactive lipids were scraped and quantified by liquid scintillation counting.
Assessment of the channel activity by ratiometric measurement of [Ca2+]i
The fluorescent indicator Indo-1/AM was used to monitor Ca2+ influx through the CNGA1 channels in cell suspensions. The assays were performed as described [10,15] using a spectrofluorometer (Fluostar Omega, BMG lab tech GmBH, Offenburg, Germany). This assay was designed to determine CNG channel activity in cell populations (2 × 106) in response to 8-pCPT-cGMP stimulation. Briefly, cells (36–48 h post-transfection) were harvested with cell dissociation medium (Invitrogen, Carlsbad, CA), washed with the extracellular solution (ECS; 140 mM NaCl, 5 mM KCl, 1 mM MgCl2, 1.8 mM CaCl2, 10 mM glucose, 15 mM HEPES, pH 7.4), and incubated with 2 μM Indo-1/AM (Sigma-Aldrich) in ECS in the presence of 0.05% Pluronic F-127 (Invitrogen, Carlsbad, CA) for 40 min at room temperature. Then, the cells were washed three times with ECS and resuspended in ECS (1 × 106/mL). Ca2+ influx in response to 8-pCPT-cGMP was determined by ratiometric measurement, which represents the free intracellular Ca2+ concentration. Changes of intracellular Ca2+ concentration were expressed as a ∆405/485 ratio.
Molecular modeling studies
The protein tertiary structural modeling was done using the MODELLER program . Structural manipulations, surface topology studies, and graphical representations were performed using Arguslab (Thompson, Planaria software), VMD , and DaliLite v3 . The extensive energy minimization was done using Deep View . Structural constraints and prediction quality of the modeled structures were evaluated using WHATIF  and PROCHECK .
Data were analyzed and graphed using Graphpad Prism software (GraphPad Software, San Diego, CA). The data were expressed as the mean ± SD and compared by Student’s t test for unpaired data. The critical level of significance was set at p < 0.05.
This work was supported by grants from the NIH/NEI (EY016507; EY00871; EY021725) and an unrestricted grant from Research to Prevent Blindness, Inc. to the Department of Ophthalmology. The authors acknowledge Ms. Kathy J. Kyler, Staff Editor, Office of Research Administration, University of Oklahoma Health Sciences for editing this manuscript.
- Fruman DA, Meyers RE, Cantley LC. Phosphoinositide kinases. Annu Rev Biochem. 1998;67:481–507.View ArticlePubMedGoogle Scholar
- Pacold ME, Suire S, Perisic O, Lara-Gonzalez S, Davis CT, Walker EH, et al. Crystal structure and functional analysis of Ras binding to its effector phosphoinositide 3-kinase gamma. Cell. 2000;103:931–43.View ArticlePubMedGoogle Scholar
- Kaupp UB, Seifert R. Cyclic nucleotide-gated ion channels. Physiol Rev. 2002;82:769–824.PubMedGoogle Scholar
- Yau KW, Hardie RC. Phototransduction motifs and variations. Cell. 2009;139:246–64.View ArticlePubMed CentralPubMedGoogle Scholar
- Brady JD, Rich ED, Martens JR, Karpen JW, Varnum MD, Brown RL. Interplay between Pip3 and calmodulin regulation of olfactory cyclic nucleotide-gated channels. Proc Natl Acad Sci U S A. 2006;103:15635–40.View ArticlePubMed CentralPubMedGoogle Scholar
- Brunert D, Klasen K, Corey EA, Ache BW. Pi3kgamma-dependent signaling in mouse olfactory receptor neurons. Chem Senses. 2010;35:301–8.View ArticlePubMed CentralPubMedGoogle Scholar
- Bright SR, Rich ED, Varnum MD. Regulation of human cone cyclic nucleotide-gated channels by endogenous phospholipids and exogenously applied phosphatidylinositol 3,4,5-trisphosphate. Mol Pharmacol. 2007;71:176–83.View ArticlePubMedGoogle Scholar
- Suire S, Condliffe AM, Ferguson GJ, Ellson CD, Guillou H, Davidson K, et al. Gbetagammas and the Ras binding domain of P110gamma are both important regulators of Pi(3)Kgamma signalling in neutrophils. Nat Cell Biol. 2006;8:1303–9.View ArticlePubMedGoogle Scholar
- Rajala RV, Rajala A, Gupta VK. Conservation and divergence of Grb7 family of Ras-binding domains. Protein Cell. 2012;3:60–70.View ArticlePubMed CentralPubMedGoogle Scholar
- Gupta VK, Rajala A, Daly RJ, Rajala RV. Growth factor receptor-bound protein 14: a new modulator of photoreceptor-specific cyclic-nucleotide-gated channel. EMBO Rep. 2010;11:861–7.View ArticlePubMed CentralPubMedGoogle Scholar
- Rajala RV. Phosphoinositide 3-kinase signaling in the vertebrate retina. J Lipid Res. 2010;51:4–22.View ArticlePubMed CentralPubMedGoogle Scholar
- Fung-Leung WP. Phosphoinositide 3-kinase delta (Pi3kdelta) in leukocyte signaling and function. Cell Signal. 2011;23:603–8.View ArticlePubMedGoogle Scholar
- Rajala A, Anderson RE, JX M, Lem J, Al Ubaidi MR, Rajala RV. G-protein-coupled receptor rhodopsin regulates the phosphorylation of retinal insulin receptor. J Biol Chem. 2007;282:9865–73.View ArticlePubMedGoogle Scholar
- Rajala RV, Mcclellan ME, Ash JD, Anderson RE. In vivo regulation of phosphoinositide 3-kinase in retina through light-induced tyrosine phosphorylation of the insulin receptor beta-subunit. J Biol Chem. 2002;277:43319–26.View ArticlePubMedGoogle Scholar
- Gupta V, Rajala A, Rodgers K, Rajala RV. Mechanism involved in the modulation of photoreceptor-specific cyclic nucleotide-gated channel by The tyrosine kinase adapter protein Grb14. Protein Cell. 2011;2:906–17.View ArticlePubMed CentralPubMedGoogle Scholar
- Suire S, Hawkins P, Stephens L. Activation of phosphoinositide 3-kinase gamma by Ras. Curr Biol. 2002;12:1068–75.View ArticlePubMedGoogle Scholar
- Depetris RS, Wu J, Hubbard SR. Structural and functional studies of the Ras-associating and pleckstrin-homology domains of Grb10 and Grb14. Nat Struct Mol Biol. 2009;16:833–9.View ArticlePubMed CentralPubMedGoogle Scholar
- Nassar N, Horn G, Herrmann C, Scherer A, Mccormick F, Wittinghofer A. The 2.2 A crystal structure of the Ras-binding domain of the serine/threonine kinase C-raf1 in complex with Rap1a and a Gtp analogue. Nature. 1995;375:554–60.View ArticlePubMedGoogle Scholar
- Walker EH, Perisic O, Ried C, Stephens L, Williams RL. Structural insights into phosphoinositide 3-kinase catalysis and signalling. Nature. 1999;402:313–20.View ArticlePubMedGoogle Scholar
- Womack KB, Gordon SE, He F, Wensel TG, Lu CC, Hilgemann DW. Do phosphatidylinositides modulate vertebrate phototransduction? J Neurosci. 2000;20:2792–9.PubMedGoogle Scholar
- Nilius B, Mahieu F, Prenen J, Janssens A, Owsianik G, Vennekens R, et al. The Ca2 + −activated cation channel Trpm4 is regulated by phosphatidylinositol 4,5-biphosphate. EMBO J. 2006;25:467–78.View ArticlePubMed CentralPubMedGoogle Scholar
- Suh BC, Hille B. Regulation of ion channels by phosphatidylinositol 4,5-bisphosphate. Curr Opin Neurobiol. 2005;15:370–8.View ArticlePubMedGoogle Scholar
- He F, Mao M, Wensel TG. Enhancement of phototransduction G protein-effector interactions by phosphoinositides. J Biol Chem. 2004;279:8986–90.View ArticlePubMedGoogle Scholar
- Wigler M, Pellicer A, Silverstein S, Axel R. Biochemical transfer of single-copy eucaryotic genes using total cellular DNA as donor. Cell. 1978;14:725–31.View ArticlePubMedGoogle Scholar
- Rajala RV, Mcclellan ME, Chan MD, Tsiokas L, Anderson RE. Interaction of the retinal insulin receptor beta-subunit with the P85 subunit of phosphoinositide 3-kinase. Biochemistry. 2004;43:5637–50.View ArticlePubMedGoogle Scholar
- Rajala A, Tanito M, Le YZ, Kahn CR, Rajala RV. Loss of neuroprotective survival signal in mice lacking insulin receptor gene in rod photoreceptor cells. J Biol Chem. 2008;283:19781–92.View ArticlePubMed CentralPubMedGoogle Scholar
- Johnsson N, Varshavsky A. Split ubiquitin as a sensor of protein interactions in vivo. Proc Natl Acad Sci U S A. 1994;91:10340–4.View ArticlePubMed CentralPubMedGoogle Scholar
- Stagljar I, Korostensky C, Johnsson N, Te HS. A genetic system based on split-ubiquitin for the analysis of interactions between membrane proteins in vivo. Proc Natl Acad Sci U S A. 1998;95:5187–92.View ArticlePubMed CentralPubMedGoogle Scholar
- Breeden L, Nasmyth K. Cell cycle control of the yeast Ho gene: cis- and trans-acting regulators. Cell. 1987;48:389–97.View ArticlePubMedGoogle Scholar
- Kaplan DR, Whitman M, Schaffhausen B, Pallas DC, White M, Cantley L, et al. Common elements in growth factor stimulation and oncogenic transformation: 85 kd phosphoprotein and phosphatidylinositol kinase activity. Cell. 1987;50:1021–9.View ArticlePubMedGoogle Scholar
- Eswar N, Eramian D, Webb B, Shen MY, Sali A. Protein structure modeling with modeller. Methods Mol Biol. 2008;426:145–59.View ArticlePubMedGoogle Scholar
- Humphrey W, Dalke A, Schulten K. Vmd: visual molecular dynamics. J Mol Graph. 1996;14:33–8.View ArticlePubMedGoogle Scholar
- Holm L, Kaariainen S, Rosenstrom P, Schenkel A. Searching protein structure databases with Dalilite V. 3. Bioinformatics. 2008;24:2780–1.View ArticlePubMed CentralPubMedGoogle Scholar
- Guex N, Peitsch MC. Swiss-Model and the Swiss-Pdb viewer: an environment for comparative protein modeling. Electrophoresis. 1997;18:2714–23.View ArticlePubMedGoogle Scholar
- Vriend G. What If: a molecular modeling and drug design program. J Mol Graph. 1990;8:52–6. 29.View ArticlePubMedGoogle Scholar
- Morris AL, Macarthur MW, Hutchinson EG, Thornton JM. Stereochemical quality of protein structure coordinates. Proteins. 1992;12:345–64.View ArticlePubMedGoogle Scholar
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.