Variable phosphorylation states of pigment-epithelium–derived factor differentially regulate its function

The pigment epithelium–derived factor (PEDF) is a member of the serine protease inhibitors (serpin) superfamily, but as of today was not found to exhibit inhibitory activity against any proteases.^1,2  It was first isolated based on its ability to convert dividing retinoblastoma cells into differentiated neurons, and thus was characterized as a neurotrophic factor.^1,3  Later, it was shown that besides its neurotrophic functions, PEDF is a potent natural inhibitor of angiogenesis in the eye,⁴  where it inhibits stimulatory activity of several strong proangiogenic factors. This antiangiogenic potency has also been shown in several animal models in which PEDF was demonstrated as the factor responsible for the reduction of blood-vessel growth in the eye.^5-9  Although originally discovered in the culture medium of pigment epithelial cells obtained from the fetal human retina,¹⁰  it is clear today that PEDF is expressed not only in the retina, but also at multiple sites in the adult eye,^11-13  as well as in the adult human brain, the spinal cord,^14,15  and human plasma.¹⁶  Therefore, it is possible that PEDF has the potential to inhibit angiogenesis throughout the body.

It is well established that protein phosphorylation plays a key role in the regulation of most intracellular processes. However, it is becoming increasingly evident that protein kinases can also regulate extracellular processes, as the kinases are present extracellularly as either ectoprotein or exoprotein kinases.^17,18  The ectoprotein kinases are membrane-bound enzymes whose catalytic activities are localized on the extracellular cell surface of a wide variety of cells. The exoprotein kinases are secreted, soluble enzymes whose catalytic activities are present in the extracellular environment without being directly associated with cells.¹⁷  These protein kinases were shown to phosphorylate both extracellular soluble substrates as well as cell-surface proteins, thereby playing a regulatory role in many physiologic processes including cell-cell interaction, differentiation, proliferation, and ion fluxes.¹⁷ 

Plasma circulating proteins were shown to be candidate substrates for phosphorylation by exoprotein kinases, and phosphorylation appeared to modulate the function of some of these proteins.^18-20  Indeed, our recent work has shown that PEDF purified from human plasma (plPEDF) is a phosphoprotein, which is phosphorylated in the circulation by casein kinase CK2 (CK2) and protein kinase A (PKA).²¹  We have shown that CK2 phosphorylates PEDF on 2 main residues, Ser24 and Ser114, while PKA was shown to phosphorylate PEDF on Ser227. Using several phosphorylation site mutants that mimic either the phospho (Ser to Glu) or nonphospho (Ser to Ala) forms of PEDF, we found that both CK2 and PKA phosphorylations of PEDF markedly affect its physiologic function. The CK2-phosphorylated PEDF had a reduced neurotrophic activity, while its antiangiogenic activity was significantly increased. On the other hand, PKA phosphorylation reduced PEDF antiangiogenic activity but had very little effect on its neurotrophic activity.²¹ 

Since we found that both CK2 and PKA sites on PEDF can be phosphorylated,²¹  we then questioned the interplay between the 2 phosphorylation events and their functional consequences. We report here that a mutant mimicking PKA-phosphorylation of PEDF served as a good substrate for CK2, whereas the mutant mimicking the CK2-phosphorylated PEDF could not be phosphorylated by PKA. By generating triple mutants imitating the differential phosphorylation state of PEDF, we found that the triple mutant that mimics phosphorylation on the CK2 together with the PKA sites exhibits a significantly elevated antiangiogenic activity. This activity was stronger than the antiangiogenic activity of PEDF with phosphomimetic mutations in the CK2 but not PKA sites. Moreover, the triple phosphomimetic mutant was able to induce neuronal differentiation of Y-79 retinoblastoma cells, while the neurotrophic activity of the mutant mimicking only CK2 phosphorylation was abolished. Thus, we conclude that differential phosphorylation of PEDF induces at least 4 functional states, and that combined PKA and CK2 phosphorylation induces its maximal antiangiogenic as well as neurotrophic activities.

Recombinant human CK2 (expressed in E coli) was from Calbiochem (Darmstadt, Germany). The catalytic subunit of PKA was purified as described.²²  Endothelial mitogen (ECGS) was from Biomedical Technologies (Stoughton, MA). Recombinant human bFGF (expressed in E coli), poly-l-lysine (70-150 kDa), and porcine intestinal heparin were from Sigma (St Louis, MO). Matrigel was from BD Biosciences (Bedford, MA). Restriction enzymes were from Roche (Mannheim, Germany). Pfu DNA polymerase was from Promega (Madison, WI). Polyclonal antibody against PEDF was developed by the Antibody Unit of the Weizmann Institute of Science (Rehovot, Israel). Full-length human PEDF cDNA was provided by Dr N. Bouck (Northwestern University, Chicago, IL). Recombinant human VEGF121 was from CytoLab (Rehovot, Israel).

Human Y-79 retinoblastoma cells were grown in MEM supplemented with 2 mM l-glutamine and 15% fetal calf serum (FCS). HEK-293T cells were cultured in DMEM F-12 supplemented with 10% FCS. Human umbilical vein endothelial cells (HUVECs) were grown in M-199 supplemented with 20% FCS, 25 μg/mL endothelial-cell growth supplement (ECGS) as mitogen, and 5 U/mL heparin.

The triple mutants were generated by replacing a HindIII and KpnI digestion fragment (1-362 bp), containing the region of Ser24 and Ser114 and their mutations,²¹  with the same fragment in plasmids containing various mutations in Ser227. This yielded the triple mutants as follows: EEE mutant, S24,114E insert ligated with digested pcDNA3-S227E vector; EEA mutant, S24,114E insert ligated with digested pcDNA3-S227A vector; AAE mutant, S24,114A insert ligated with digested pcDNA3-S227E vector; AAA mutant, S24,114A insert ligated with digested pcDNA3-S227A vector.

The various plasmids carrying rPEDF or the mutants were introduced into HEK-293T cells using LipofectAMINE reagent (Invitrogen, Carlsbad, CA), and the secreted proteins were purified on Ni⁺² columns as described.²¹ 

plPEDF was purified from human citrated plasma (1 L) by a 9% to 20% PEG cut followed by DEAE-Sephacel column (2.9 × 40 cm) and heparin agarose column as previously described.²¹ 

The phosphorylation assay (40 μL) contained either rPEDF, plPEDF, or rPEDF mutants (50 μg/mL). For CK2, the constituents were CK2 (4 μg/mL), glycerol (2%), NaCl (20 mM), β-mercaptoethanol (0.1 mM), MgCl₂ (20 mM), [γ³²P]-ATP (10 μM), poly-l-lysine (200 nM), and Tris-HCL (50 mM, pH 7.4). For PKA, the constituents were pure catalytic subunit of PKA (2.5 μg/mL), MgCl₂ (10 mM), heparin (50 μg/mL), [γ³²P]-ATP (10 μM), and Tris-HCL (50 mM, pH 6.5). Reactions were for 45 minutes at 30°C. Then, boiled sample buffer was added, and the samples were subjected to 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE).

Following stimulation, ERK and phosphorylated ERK were detected by Western blotting using the appropriate antibodies as described.²³ 

Proliferation was determined by the methylene blue assay as previously described.²⁴  Briefly, HUVECs were seeded in gelatin-coated 24-well tissue-culture plates (20 × 10³ cells/well) in M-199 supplemented with 2.5% FCS or 5% FCS with or without bFGF or VEGF (10 ng/mL; 0.5 mL/well). The various PEDFs were added immediately following seeding in quadruplicate (all at 10 nM), and plates were incubated in a humidified incubator for 48 hours. Then cells were fixed in 4% buffered formaldehyde solution for 2 hours, washed twice with 0.1 M sodium borate buffer, pH 8.5, and stained with 1% methylene blue dissolved in 0.1 M borate buffer for 20 minutes. Excess dye was washed out and cell-bound dye was eluted with 200 μL/well of 0.1 M HCl. The optical density was read at 595 nm in a Wallac 1420 multilabel counter (Turku, Finland).

Human Y-79 retinoblastoma cells (obtained from ATCC, Manassas, VA) were assayed for neurite outgrowth as previously described.²⁵  Y-79 cell suspension (1 mL; 2.5 × 10⁵ cells/mL) was incubated with rPEDF, or the various rPEDF mutants (20 nM) in MEM supplemented with 1 mM sodium pyruvate, 10 mM HEPES, 0.1 mM nonessential amino acids, 2 mM l-glutamine, antibiotics, and 0.1% insulin, transferrin, selenium (ITS) mix. After 7 days in culture, the cells were transferred to poly-d-lysine–coated plates, and their neurite outgrowth was monitored by light microscopy at various periods of time.

Matrigel (0.5 mL/mouse) kept on ice was mixed with bFGF at the indicated concentration with or without PEDF or its mutants (20 nM), and was injected subcutaneously into the flank of 8-week-old nude mice as described.²⁶  After injection, the Matrigel rapidly formed a plug. On day 7, mice were killed and their skin was carefully pulled back to expose the intact plugs. The plugs were removed, fixed (4% formaldehyde), paraffin embedded, and sectioned. Sections were stained using hematoxylin-eosin (H&E). Endothelial cells/microvessels infiltrating the Matrigel were confirmed by Masson trichrome staining.

We have previously shown that phosphorylation of PEDF by each of the exoprotein kinases CK2 and PKA affects its physiologic function. Here we undertook to study the interplay between the phosphorylations by the 2 kinases. As a first step, we examined whether each phosphorylation changes the ability of PEDF to serve as a substrate for the other protein kinase. To this end, we subjected the CK2 phosphorylation mutant S24,114E (EE) to PKA phosphorylation by incubating it with the pure catalytic subunit of PKA and [γ³²P]-ATP. Since PKA phosphorylation is pH dependent, we repeated the experiment in different pHs.

To our surprise, we found that while rPEDF served as a good substrate for PKA, the phosphorylated mutant, S24,114E, was not phosphorylated by PKA under the conditions used (Figure 1A). However, a brief heat treatment, which was designed to mildly abrogate the native structure of the S24,114E, restored its phosphorylation by PKA (Figure 1B). This result indicates that the lack of phosphorylation is due to a conformation-dependent masking of the PKA phosphorylation site by the phosphorylated CK2 sites, and this is despite the fact that the 2 sites are located in separate regions of the PEDF molecule.²⁷  Furthermore, the mutant S24,114A (AA), which cannot be phosphorylated by CK2 at all, was readily phosphorylated by PKA under the different reaction conditions tested (Figure 1C), indicating that the lack of phosphorylation is indeed due to the negative charge on Ser24 and Ser114. In contrast to the prevention of PKA phosphorylation by the negative charges at the CK2 sites, negative charge at the PKA site (Ser227) had no significant effect on CK2 phosphorylation. This was true for both PKA-phosphorylated and nonphosphorylatable site mutants (S227E and S227A), which were both readily phosphorylated by CK2 (Figure 1D). The hyperphosphorylation of the denatured PEDF by CK2 was previously reported,²¹  and indicates that additional CK2 sites are exposed by PEDF denaturation. Moreover, as previously shown, the phosphorylation of the plPEDF by CK2 was significantly lower than CK2 phosphorylation of the rPEDF (Figure 1D), and this probably occurs due to prephosphorylation of the circulating protein on these sites.²¹  Thus, we concluded that CK2 phosphorylation of Ser24 and Ser114 induces a conformational change in PEDF that makes Ser227 inaccessible to PKA phosphorylation. However, PKA phosphorylation of Ser227 does not affect the ability of PEDF to be phosphorylated by CK2.

We have previously shown that PEDF purified from the circulating blood (plPEDF) contains phosphates on its CK2 sites and to a lesser extent also on its PKA site.²¹  In addition, the results in Figure 1 show that the 3 PEDF sites can be phosphorylated at the same time, first by PKA and then by CK2. In view of these findings, it became important to characterize the effects of the simultaneous phosphorylation of PEDF by both these kinases. Therefore, we constructed a set of triple site mutants by replacing the CK2 and PKA phosphorylation sites, Ser24, Ser114, and Ser227, with Ala or Glu as follows: S24E114E227E (EEE) mimics phosphorylation on the CK2 and PKA sites, S24E114E227A (EEA) mimics phosphorylation on CK2 but not PKA sites, S24A114A227E (AAE) mimics phosphorylation on PKA but not CK2 sites, and S24A114A227A (AAA) mimics the non-CK2– or PKA-phosphorylated PEDF (Figure 2A). The use of the Glu mutants, rather than the partially phosphorylated plPEDF, was important because it provides a homogenous population of molecules with a negative charge in the relevant sites, and thereby enables the accurate detection of the phosphorylation effect. The mutations of the phosphorylated Ser to Ala residues were important as well, because they supply a homogenous population of unphosphorylated molecules. This is unlike the small amount of phosphorylation that occurs on the PKA site in the S24,114E or S24,114A, as well as on the CK2 site in the S227E and S227A mutants (Figure 1 in Maik-Rachline et al²¹  and data not shown), which can partially affect the properties of these molecules. These mutants were expressed by transfecting HEK 293T cells with the full-length human PEDF cDNA or the mutants, and were then purified on Ni⁺² columns. High-performance liquid chromatography (HPLC) and SDS-PAGE confirmed that all the mutated proteins were more than 80% pure, and were similar in their expression and solubility levels (data not shown).

To characterize the mutants and confirm their biochemical integrity, we subjected each of them to an in vitro phosphorylation by CK2 or PKA. As expected from our previous study with the S24,114A mutant, the AAE and the AAA mutants almost completely abolished CK2 phosphorylation (Figure 2B). On the other hand, the triple mutants EEE and EEA were readily phosphorylated by CK2, indicating that these mutants expose additional phosphorylation sites in accordance with the hyperphosphorylation of the previously described S24,114E (EE) mutant.²¹  Therefore, the addition of the mutations of S227A or S227E to the CK2-mutated sites did not significantly affect their CK2 phosphorylation level compared with the phosphorylation of S24,114E alone (Figure 2B and data not shown). In addition, none of the triple mutants was phosphorylated by PKA (Figure 2C), indicating that mutation of the CK2 sites either to Glu or to Ala within the triple mutants did not affect the conformational structure of the PKA site.

Although the receptor for PEDF has not been identified yet, it was shown that this factor can stimulate various intracellular signaling processes,^28-30  including the extracellular signal-regulated kinase (ERK)/mitogen-activated protein kinase (MAPK) cascade.²¹  To further characterize the various triple mutants of PEDF, we examined their effect on ERK activation in HUVECs. Thus, the various mutants were added to serum-starved HUVECs and the activity of ERK was determined using anti–phospho-ERK and anti–general-ERK antibodies. In our previous report, we found that phosphorylation of PEDF by CK2 markedly elevates the effect of rPEDF on ERK activation, while PKA phosphorylation alone had essentially no effect.²¹  Indeed, in the current experiment, the S24,114E mutant induced a significant ERK phosphorylation (Figure 3A-B), which was higher than the phosphorylation induced by rPEDF (data not shown). However, the phosphorylation of ERK induced by the triple phosphorylation mutant EEE resulted in an even stronger phosphorylation, which was about 1.5-fold higher than that of the S24,114E mutant (Figure 3B). The EEA and the AAE mutants induced ERK phosphorylation to a slightly higher level than the S24,114E mutant, while the AAA mutant significantly reduced the ability of PEDF to activate ERK. These results suggest that phosphorylation of PEDF by either PKA or CK2 is necessary for the PEDF-induced activation of the ERK cascade and that accumulative phosphorylation of the 3 sites of these kinases significantly elevates this induction.

PEDF was previously shown to inhibit cell proliferation of endometrial carcinoma cells³¹  and to induce apoptosis of endothelial cells.⁵  We were therefore interested in determining the effect of PEDF and its phosphorylation on HUVEC proliferation. Thus, HUVECs were seeded in a 24-well tissue-culture plate, and measured for their PEDF-dependent proliferation rate either in 2.5% FCS for nonstimulated conditions or in 5% FCS with either bFGF or VEGF (10 ng/mL each) for stimulated growth. After 48 hours, the cells were analyzed for proliferation rate by the methylene blue assay.²⁴  Under all of these conditions, rPEDF inhibited HUVEC proliferation to a moderate extent (Figure 3C), while the inhibitory effect of PEDF purified from plasma (plPEDF) was more pronounced. The inhibition of HUVEC proliferation observed when cells were treated with the S24,114E mutant was very similar to the inhibition observed with plPEDF, indicating again that PEDF in the circulating plasma is largely phosphorylated on its CK2 sites. Of interest, when cells were treated with the EEE mutant, the level of inhibition was highly elevated (57%), whereas cells treated with the EEA mutant behaved very similarly to cells treated with the S24,114E mutant (27% and 33% inhibition, respectively). On the other hand, the PKA phosphorylation mutants, S227E and AAE, had almost no effect on HUVEC proliferation. In addition, cells treated with the nonphosphorylated mutants, S24,114A and S227A, showed similar levels of proliferation as cells treated with rPEDF, whereas the AAA mutant barely affected HUVEC proliferation. Thus, these results show that the inhibition of HUVEC proliferation is dependent on the phosphorylation state of PEDF. In particular, cell proliferation is inhibited by PEDF phosphorylated on its CK2 sites, and this inhibition is highly increased upon addition of a negative charge in the PKA sites to that of the CK2 sites. It should also be noted that despite the partial correlation between the effect of PEDF on proliferation, and its effect on ERK, it is unlikely that ERK mediates the PEDF effects on proliferation (data not shown). This may suggest that the reduced proliferation or the elevated ERK activation is secondary to other processes that are directly affected by PEDF.

We have previously shown that CK2 phosphorylation significantly reduces the neurotrophic effect of PEDF, while PKA phosphorylation of PEDF has no, or very little, influence on the neurotrophic effect of PEDF.²¹  We were further interested to determine whether simultaneous phosphate incorporation to the PKA and CK2 sites of PEDF modulate the ability of the latter to induce differentiation in human retinoblastoma Y-79 cells in culture. Surprisingly, cells treated with the EEE mutant exhibit neurite outgrowth and formed large aggregates, while the EEA mutant formed small corona-like structures, which were very compact, without any sprouts projecting from these cells, in a similar fashion to cells treated with the S24,114E mutant (Figure 4). Mutation of the PKA phosphorylation site S227E and AAE mutant revealed a different phenotype, where colonies were smaller, although their processes were clearly observed. Therefore, we concluded that while CK2 phosphorylation significantly reduces the PEDF neurotrophic effect, the addition of phosphate to the PKA site, on top of the phosphorylation by CK2, preserves the neurotrophic activity of PEDF. This occurs despite the minimal effect of PEDF phosphorylated on its PKA site alone.

To further explore the effect of both CK2 and PKA phosphorylations on PEDF function, we examined the influence of the triple mutants on the antiangiogenic activity of PEDF using the Matrigel plug assay.²⁶  Thus, liquid Matrigel supplemented with the examined factors was injected subcutaneously into CD-1 nude mice. The Matrigel polymerized to form a plug, which was removed after a week and analyzed for the growth and infiltration of microvessels or endothelial cells by histology staining. Matrigel plugs were treated with rPEDF, plPEDF, or the various mutants in combination with bFGF (300 ng/mL). As expected,^21,26  control plugs treated with PBS showed a very minute angiogenic response, bFGF-treated plugs showed robust angiogenic activity, and the rPEDF, plPEGF, as well as the S24,114E and S24,114A mutants exhibited the previously decribed³¹  antiangiogenic activity in the system (Figure 5). The triple mutants EEE or EEA reduced bFGF-induced angiogenesis to a degree that was roughly similar to the effect of S24,114E, while AAA and AAE mutants had only a minor effect that was roughly similar to that of S24,114A. Of interest, the antiangiogenic effect of the EEE mutant was somewhat higher than that of the other mutants, but this was not statistically significant under the conditions used in the current experiment. It is noteworthy to mention that a few proliferating cells that did not form mature vessels were observed in the periphery of all plugs, independent of the type of PEDF mutants, suggesting that the antiangiogenic activity is mediated in part by mechanisms that do not involve inhibition of proliferation.

In view of the slightly better inhibition of blood-vessel sprouting by the EEE mutant, we undertook to examine whether this mutant is indeed a better antiangiogenic factor than the other phosphorylation mutants. To this end, we challenged the inhibitory effect of these mutants on Matrigel plugs treated with a higher level of bFGF (500 ng/mL). Overall, plugs that were treated with 500 ng/mL bFGF exhibited a higher angiogenic response than plugs treated with 300 ng/mL (Figure 6). This was evident by a significant elevation in the number of cells infiltrating the plug (Figure 6) and the clear staining of actual mature vessels in these plugs (data not shown), which were not apparent in plugs treated with the lower concentration of bFGF. Under these higher bFGF levels, the EEE mutant significantly increased PEDF antiangiogenic activity, as no vessels were observed in the plug (Figure 6). This inhibitory activity was significantly more pronounced compared with the antiangiogenic activity observed with the S24,114E or EEA mutants (P = .04, and P = .01, respectively). Therefore, these results indicate that PKA phosphorylation of PEDF is essential for its robust antiangiogenic activity when combined with CK2 phosphorylation, but not when presented by itself.²¹  We conclude that the simultaneous phosphorylation of PEDF by CK2 and PKA results in its highest level of antiangiogenic activity, more so than its antiangiogenic activity when PEDF is phosphorylated by CK2 alone.

Phosphorylation of PEDF plays an important role in the determination of its physiologic activity. In a previous paper,²¹  we showed that extracellular phosphorylation of PEDF by CK2 abolishes PEDF neurotrophic activity, but enhances its antiangiogenic activity, while PKA phosphorylation reduces PEDF antiangiogenic activity without affecting its neurotrophic activity. In the current report, we demonstrate that combined replacement of the Ser at the PKA and CK2 phosphorylation sites to Glu turns PEDF into its most potent antiangiogenic form, while retaining its neurotrophic activity. Considering all of these observations, we concluded that the extracellular phosphorylation of PEDF by PKA, CK2, or both kinases together could result in distinct PEDF activities. Thus, the nonphosphorylated protein has weak antiangiogenic as well as neurotrophic activities; the PKA-phosphorylated protein exhibits a strong neurotrophic activity but not antiangiogenic activity; the PEDF phosphorylated on the 2 CK2 sites exhibits an antiangiogenic activity without having any neurotrophic effect; and finally, the triply phosphorylated protein regains both activities. Furthermore, the antiangiogenic activity of the triply phosphorylated protein is much higher than that of the nonphosphorylated or mono-PKA–phosphorylated PEDF, and is also higher than that of the doubly CK2-phosphorylated protein (Figure 7). Therefore, these extracellular phosphorylations contribute a higher level of complexity by allowing the existence of 4 distinct forms of the PEDF proteins, which modulate its activity.

It was previously shown that physiologic stimulation of platelets with thrombin causes them to release the exoprotein kinases PKA¹⁸  and CK2³²  in their active forms. Additionally, it was shown that the plasma concentration of ATP following platelet activation by thrombin in vitro is approximately 12 μM,³³  which is sufficient to support extracellular kinase activity. Moreover, using partially purified plasma samples from individual CD-1 nude mice for immunoblot analysis with anti-CK2, PKA, or PEDF antibodies, we found the existence of CK2 protein in 60% of the mice, whereas PKA protein was identified in the plasma of 20% of the mice (data not shown). Thus, CK2 and PKA protein do exist in the circulating blood, and thereby can affect the phosphorylation state of PEDF. Indeed, in our previous report,²¹  we showed that PEDF purified from plasma (plPEDF) is partially phosphorylated on its CK2 sites and to some extent also on the PKA site. Since CK2 and PKA can phosphorylate PEDF individually, and the phosphorylated residues are located at opposite sides of the molecule, as judged by locating these residues within the 3D structure of PEDF,²⁷  it is possible that the PKA and CK2 sites on PEDF can be simultaneously phosphorylated.

Surprisingly, the in vitro phosphorylation assay (Figure 1) revealed that the double mutation in the CK2 phosphorylation sites Ser24 and Ser114 to Glu prevented further phosphorylation of PEDF by PKA, and partial denaturation of this mutant restored its sensitivity to PKA phosphorylation. We concluded that CK2 phosphorylation is followed by a conformational change in the PEDF molecule, thereby making Ser227 inaccessible to PKA phosphorylation. On the other hand, Glu mutation in the PKA phosphorylation site of PEDF does not affect the conformational state of the CK2 phosphorylation sites and allows their further phosphorylation. The nature of this conformational change and the mechanism that allows the masking of the PKA site in the center of the molecule by the N-terminal phosphorylation are not clear, but may be related to the exposure of functional epitopes of PEDF that are related to its activities.³⁴  The inhibitory effect of the CK2-phosphorylated PEDF on its PKA phosphorylation may serve as a regulatory mechanism of PEDF function such as under conditions where the highly antiangiogenic activity of PEDF should be preserved but its neurotrophic activity eliminated. On the other hand, the PKA phosphorylation of PEDF that was previously shown to reduce PEDF antiangiogenic function²¹  can be further phosphorylated by CK2, and by doing so, converts it from a poor to a very potent antiangiogenic factor that maintains its neurotrophic activity.

As in many other studies (eg, Seger et al³⁵ ), we used Glu residue to mimic phosphorylated Ser, while the substitution with Ala residue mimicked the nonphosphorylated state. Thus, the triple mutant EEE served to mimic the fully PKA- and CK2-phosphorylated molecule, with the advantage that unlike the partial phosphorylation stoichiometry in the phosphorylated PEDF, all the mutated molecules were indeed fully pseudophosphorylated.As mentioned in “Results,” the Ala replacement mutants, EEA, AAE, and AAA, served to prevent any phosphorylation on the corresponding Ser residues, and indeed showed some differences from the double mutants. These results suggest that rPEDF and its mutants are purified from the medium of HEK293T cells as partially phosphorylated proteins. Thus, the use of the Ala-mutated PEDF protein was important to determine the full magnitude of the phosphorylation effects that could otherwise be masked by the partial phosphorylation of rPEDF.

In spite of the extensive work on PEDF, its receptor(s) has not been identified yet.³⁶  Therefore, the exact mechanisms by which PEDF and its mutants exert their different activities are not clear. We have found that rPEDF binds in a similar affinity to both HUVEC and Y-79 cells (kDa, ∼5 nM), and this binding is displaced equally well by the different phosphorylation mutants (data not shown). Therefore, the differences between the mutants and the 2 cell lines cannot be explained just by different binding properties. Rather, it is possible that 2 different receptors with similar binding properties transmit the PEDF signals. Alternatively, PEDF receptor may be composed of 2 distinct subunits that cooperate in transmitting the PEDF signals. Such a receptor should contain a binding component that is similar in different cells and a signaling component that may vary in different cells. This type of receptor may be influenced differently by distinct PEDF mutants, and therefore such a mechanism may explain the various signaling networks that are induced by PEDF, including the use of the unrelated CREB and NFkB signaling to secure survival in cerebral granula cells.³⁷  This is in contrast to the antiangiogenic effect demonstrated here.

The concomitant CK2 and PKA phosphorylation of PEDF seems to increase ERK activation (Figure 3A), but did not change JNK or p38MAPK activities (data not shown). Of interest, these phosphorylation mutants also caused inhibition of proliferation (Figure 3B), which is normally correlated with enhanced JNK/p38 activities and not ERK.³⁸  Since it is unlikely that the activation of ERK leads to inhibition of proliferation in HUVECs, the 2 processes are probably differentially regulated by PEDF. Additionally, we found that the antiangiogenic activity of PEDF is mostly independent from the inhibition of proliferation, but is fairly well correlated to the increased ERK activation. Therefore, it is possible that ERK activation is involved in a proliferation-independent mechanism that leads to the antiangiogenic activity of PEDF.

It is well established that PEDF can be successfully delivered to the eye where it inhibits choroidal neovascularization.^7-9,39  The potential use of PEDF gene therapy as an antiangiogenic agent in the treatment of ocular diseases also targeted it recently as a therapeutic candidate for the treatment of cancer. It was shown that PEDF overexpression could suppress glioma growth and invasion both in vitro and in vivo.^40,41  Another recent study showed that PEDF overexpression greatly inhibited subcutaneous tumor formation and completely prevented metastasis in human melanoma xenografts.⁴²  In view of our findings, we suggest that phosphorylation of PEDF by CK2 significantly increases its antiangiogenic activity and decreases its neurotrophic function. Therefore, the CK2-phosphorylated mutant S24,114E, and more so the CK2 and PKA phosphomimetic mutant EEE, which exhibits the highest level of antiangiogenic activity, can serve as a specific therapeutic agent targeting vessel growth in malignancies as well as in retinopathy-associated diseases.

In summary, we studied here the effect of combined phosphorylation of PKA and CK2 upon various PEDF activities. To distinguish between the different phosphorylated states of PEDF and to receive a homogenous population of modified PEDF molecules, we replaced phosphorylation site Ser to either Glu to mimic the incorporated phosphate or to Ala to mimic the nonphosphorylated form. Our results with the different mutants lead to the conclusion that upon extracellular phosphorylations, PEDF can exist in 4 distinct forms including a triply PKA- and CK2-phosphorylated protein that exhibits both antiangiogenic and neurotrophic activity, a doubly CK2–phosphorylated protein that has antiangiogenic but not neurotrophic activity, a mono-PKA–phosphorylated protein that exhibits neurotrophic but not antiangiogenic activities, and a nonphosphorylated protein that exhibits strong neurotrophic and reduced antiangiogenic activities. Therefore, extracellular phosphorylation regulates the fate of the PEDF signals dependent on cell type and growing conditions. Since the antiangiogenic activity of the EEE mutant of PEDF was much stronger than the other forms of this protein, it is suggested that this mutant can be used to develop an efficient antiangiogenic drug for use in retinopathies and malignancies.

We thank Drs Tamara Berkutzki, Aharon Rabinkov, Irina Shin, Bose S. Kochupurakkal, and Slava Klachenko as well as Mr Raanan Margalit for their contribution to this study.