Separation of Fast from Slow Anabolism by Site-specific PEGylation of Insulin-like Growth Factor I (IGF-I)*

Insulin-like growth factor I (IGF-I) has important anabolic and homeostatic functions in tissues like skeletal muscle, and a decline in circulating levels is linked with catabolic conditions. Whereas IGF-I therapies for musculoskeletal disorders have been postulated, dosing issues and disruptions of the homeostasis have so far precluded clinical application. We have developed a novel IGF-I variant by site-specific addition of polyethylene glycol (PEG) to lysine 68 (PEG-IGF-I). In vitro, this modification decreased the affinity for the IGF-I and insulin receptors, presumably through decreased association rates, and slowed down the association to IGF-I-binding proteins, selectively limiting fast but maintaining sustained anabolic activity. Desirable in vivo effects of PEG-IGF-I included increased half-life and recruitment of IGF-binding proteins, thereby reducing risk of hypoglycemia. PEG-IGF-I was equipotent to IGF-I in ameliorating contraction-induced muscle injury in vivo without affecting muscle metabolism as IGF-I did. The data provide an important step in understanding the differences of IGF-I and insulin receptor contribution to the in vivo activity of IGF-I. In addition, PEG-IGF-I presents an innovative concept for IGF-I therapy in diseases with indicated muscle dysfunction.


Insulin-like growth factor I (IGF-I) has important anabolic and homeostatic functions in tissues like skeletal muscle, and a decline in circulating levels is linked with catabolic conditions. Whereas IGF-I therapies for musculoskeletal disorders have been postulated, dosing issues and disruptions of the homeostasis have so far precluded clinical application. We have developed a novel IGF-I variant by site-specific addition of polyethylene glycol (PEG) to lysine 68 (PEG-IGF-I).
In vitro, this modification decreased the affinity for the IGF-I and insulin receptors, presumably through decreased association rates, and slowed down the association to IGF-I-binding proteins, selectively limiting fast but maintaining sustained anabolic activity. Desirable in vivo effects of PEG-IGF-I included increased half-life and recruitment of IGF-binding proteins, thereby reducing risk of hypoglycemia. PEG-IGF-I was equipotent to IGF-I in ameliorating contraction-induced muscle injury in vivo without affecting muscle metabolism as IGF-I did. The data provide an important step in understanding the differences of IGF-I and insulin receptor contribution to the in vivo activity of IGF-I. In addition, PEG-IGF-I presents an innovative concept for IGF-I therapy in diseases with indicated muscle dysfunction.
Insulin-like growth factor (IGF-I) 2 is a circulating anabolic hormone that unlike insulin is complexed to high affinity binding proteins (IGFBP) that exert homeostatic effects by regulating availability to most cells in the body (1,2). The most important endogenous source of IGF-I is the liver (3), although many other tissues produce IGF-I in a paracrine manner. IGF-I has nonselective trophic activity on many cell types including neurons (4,5) and skeletal muscle fibers (6).
Multiple therapeutic applications have been envisaged and numerous trials were performed with recombinant human IGF-I (rhIGF-I). These trials included treatment of insulin resistance, diabetes, and growth disorders (7), cardiovascular diseases (8), and stroke (9). For skeletal muscle conditions, similar approaches have been proposed (10) and preclinical proof of concept has been shown for attenuation of muscular deterioration after subchronic rhIGF-I therapy (11). However, dosing limitations have precluded its clinical development in these indications. Acute side effects such as hypoglycemia and suppression of growth hormone (GH) release after subcutaneous (s.c.) application of rhIGF-I (12,13) led to the development of IPLEX (mecasermin rinfabate, Ref. 14), a 1:1 molecular complex of rhIGF-I and IGFBP-3. Yet clinical trials with this drug did not improve efficacy and showed similar side effect profiles including GH suppression (15)(16)(17), suggesting that acute disruption of the homeostatic system due to inappropriate delivery remains a major limitation of rhIGF-I therapy.
A pharmaceutically safe formulation of rhIGF-I should have reduced acute side effects like hypoglycemia and GH suppression without loss of long-term beneficial activity. We show here that IGF-I, exclusively PEGylated at lysine 68 (Lys-68) fulfills these criteria. Most likely, a slower rate of receptor association reduced rapid activation of both IGF-I and insulin receptors (IGF-IR and InsR) but preserved steady-state activity in vitro. PEG-IGF-I was equipotent to rhIGF-I in a muscle injury paradigm without altering glucose metabolism, thus providing a safety differentiation versus rhIGF-I. We propose PEG-IGF-I as a novel therapy for a variety of clinical applications including musculoskeletal disorders or catabolic pathologies.

EXPERIMENTAL PROCEDURES
Generation of PEG-IGF-I-Details on the process of sitespecific PEGylation of IGF-I are described in patent US20100210547. All IGF-I variants were stored in 20 mM sodium acetate, 140 mM NaCl (pH 5.5) at 1 mg/ml protein concentration.
Functional Assays-Proliferation was assessed for L6, MDCK, and ASPC-1 cell lines by serum starvation for 4 h, and subsequent incubation with rhIGF-I or PEG-IGF-I for 16 h (L6, ASPC-1) or 40 h (MDCK). After addition of 10 M bromodeoxyuridine (BrdU, Roche) cells were incubated 4 h at 37°C and lysed in FixDenat solution for 30 min. BrdU labeling was assessed by ELISA (Roche) according to supplier protocols.
The ability and time course for induction of negative cooperativity in IGF-IRs by PEG-IGF-I compared with rhIGF-I was investigated using human embryonic renal cells (293 EBNA) transfected with cDNA encoding full-length human IGF-IR (293 EBNA IGF-IR) (gift from Jonathan Whittaker) using previously described methods (20). The cells were preincubated with 125 I-IGF-I for 120 min at 15°C. The cells were then resuspended in binding buffer at the initial volume. Duplicate aliquots were diluted 40-fold in the absence or presence of 3, 10, 30, 100, or 300 nM unlabeled ligand and incubated at 15°C. The cells were centrifuged at various time points and the bound activity counted.
Species-specific ELISAs were used to detect human (R&D Systems) or mouse IGF-I (Mediagnost) in serum or plasma according to supplier protocols. PEG-IGF-I plasma concentrations were determined after an acid/neutralization step to dissociate PEG-IGF-I/protein complexes, using a biotinylated mouse-anti-PEG antibody on streptavidin-coated microtiter wells and digoxigenylated rhIGFBP-4 along with sheep antidigoxigenin Fab fragments conjugated to horseradish peroxidase for detection. ABTS was used as substrate and absorbance measured at 405/490 nm.
Surface plasmon resonance was assayed on a Biacore 3000 instrument in 10 mM HEPES, 150 mM NaCl, 0.005% polysurfactant, pH 7.4 at room temperature. IGFBPs were amine-coupled at 5 g/ml on a CM5 chip. PEG-IGF-I or rhIGF-I were injected at 5 concentrations (1-100 nM) for 5 min (association phase) and washed for 5 min at a flow rate of 50 l/min. Data evaluation was done by using a 1:1 Langmuir binding model after subtraction of buffer versus sample curves.
Serum samples were analyzed by Western ligand blot analysis as previously described (22). Using recombinant human IGFBP-2 and -3 as standards on each ligand blot, concentrations of IGFBP-2 and -3 were assessed in ng/ml.
In Vivo Studies-All procedures were approved by the Animal Experimentation Ethics Committee of the University of Melbourne and the Swiss federal regulation of animal protection. C57Bl/6 mice (Charles River) received single s.c. injections of PEG-IGF-I or rhIGF-I for pharmacokinetic investigations. Male C57BL/10ScSnmdx/J (dystrophic) mdx mice (Jackson) at 3 weeks of age were treated with PEG IGF-I (0.3 mg/kg s.c. twice-a-week) or rhIGF-I (1 mg/kg/d, s.c. minipumps) for up to 6 weeks, C57BL/10ScSn mice (Jackson) were used as wild-type controls. Blood glucose was assessed from tail punctures using Accu-check devices (Roche). At times indicated, animals were sacrificed and serum generated by clotting blood for 10 min followed by centrifugation.
For assessment of contraction-induced injury, the tibialis anterior (TA) muscle was surgically exposed under anesthesia and attached to the lever arm of a servomotor for measurement of contractile properties in situ (11). TA muscles were stimu-lated to produce maximum isometric force and then subjected to a series of lengthening contractions at progressively increasing magnitudes from 2 to 55% stretch.
For muscle histology, TA muscles were blotted, weighed and snap frozen. Muscle sections were reacted for succinate dehydrogenase activity to assess oxidative capacity and fiber type proportions as described (11).
Data Analyses and Statistics-Individual in vitro experiments were repeated 3-5 times, data are expressed as means Ϯ S.E. Specific binding or dose-response curves were fitted to one-site binding association or competition functions using GraphPad Prism software. For statistical comparison of two groups, Student's t test was used. For comparison of more than two groups, one-way ANOVA was used followed by Newman-Keul post-hoc test.

Site-specific MonoPEGylation of IGF-I-Human
IGF-I has four putative PEGylation sites, the N terminus and lysines (Lys) 27, 65, and 68. Isomer purification from mixtures of randomly monoPEGylated IGF-I identified the Lys-68 derivative as the most active isomer, and we used site-directed PEGylation to generate a Lys-68 PEG-IGF-I by a combination of site-directed mutations and proteolytic cleavage. The manufacturing process is schematically described in supplemental Fig. S1. With this process nonPEGylated IGF-I mutant could not be generated. The resulting molecule with amino acid exchanges at positions 27 and 65 (lysines to arginines) is shown in Fig. 1A. A three-dimensional model of PEG-IGF-I in the site 1 IGF-I-interacting domain of the IGF-IR illustrates that PEGylation occurs outside of the binding site near the C terminus (Fig. 1B).
Receptor Binding, Negative Cooperativity, and Phosphorylation-Changes of receptor binding were investigated by competition of 125 I-rhIGF-I in L6(WT) cells (expressing endogenous IGF-IR but no insulin receptors, data not shown) or 125 Table S1). The time course of induction of negative cooperativity in IGF-IRs by PEG-IGF-I compared with rhIGF-I was studied by measuring the dose-dependence of accelerated dissociation of 125 I-rhIGF-I by either unlabeled ligand at various time points at 15°C in EBNA cells transfected with IGF-IR. As shown in Fig. 2D, the curves for unlabeled rhIGF-I move leftwards as a function of time, reflecting the time progression of unlabeled ligand binding during the dissociation phase. Fig. 2E demonstrates that this progression is much slower for unlabeled PEG-IGF-I, indicative for a much slower association rate. However, the maximal effect is similar for both ligands, indicating no loss of negative cooperativity by PEGylation. This became also evident from dissociation experiments with various concentrations of cold ligands as a function of time  NT and CT indicate N and C termini, respectively. The arrowhead points to the PEGylation site Lys 68 , gray spheres indicate mutations at amino acids 27 and 65 to arginine (R), green spheres indicate confirmed residues required for IGF-IR binding, red labels indicate three cysteine bridges as formed in the native and functional IGF-I structure. B, three-dimensional model demonstrating the exposed position of Lys 68 where PEG is attached (arrowhead). Blue, IGF-IR binding domain; yellow, IGF-I; green, amino acids 60 -70; red, Lys 68 . (Fig. 2, F and G), suggesting moderately reduced affinity of PEG-IGF-I for IGF-IR, and negligible affinity for InsR. The non-PEGylated K27R/K65R mutant showed similar affinity with rhIGF-I (supplemental Table S2).
Kinetic properties of the interaction of rhIGF-I and PEG-IGF-I with IGFBP were directly analyzed by surface plasmon resonance at 24°C. On average, PEG-IGF-I showed ϳ10-fold lower K D values because of lower on-rates (k on ) and unchanged off-rates (k off , Fig. 2H, supplemental Table S3). This resulted in extended PEG-IGF-I availability for 15 min IGF-IR activation in the presence of serum IGFBP compared with rhIGF-I (supplemental Fig. S2). Correlation of pK values (Ϫlog EC 50 ) of all binding and receptor activation experiments revealed an average 10 -15-fold lower activity for PEG-IGF-I compared with rhIGF-I for IGF-IR, InsR, and IGFBPs (Fig. 2I).
Downstream Signaling and Function-To assess cellular function, we analyzed downstream signaling of rhIGF-I or PEG-IGF-I in L6 myoblasts after a 30-min treatment period at 37°C, when signaling is considered to be in steady state. We obtained dose relationships with EC 50 values for rhIGF-I or PEG-IGF-I of 4.5 and 1.7 nM for phosphorylated Akt (Fig. 3A), and of 7.8 and 7.9 nM for phosphorylated MAPK (Fig. 3B), respectively. As similar results were found using different cell lines (ASPC-1, MDCK, supplemental Fig. S3), these data showed that PEGylation of IGF-I did not change activity after 30 min of incubation.
To investigate downstream function as a result of InsR or IGF-IR activation, we analyzed glucose uptake in myotubes, elicited mainly by IGF-IR, and in differentiated adipocytes, where glucose uptake is InsR-mediated (see "Discussion"). In contrast, adipocyte stimulation strongly induced uptake by all ligands reaching plateau values within Յ20 min. Initially, the potency of PEG-IGF-I was 10-fold lower compared with rhIGF-I (EC 50 values: 0.30, 3.3, 0.02 nM for rhIGF-I, PEG-IGF-I, rhInsulin, respectively), but this difference was reduced to ϳ3-fold after stimulation for 2 h (EC 50 values: 0.23, 0.74, 0.03 nM for rhIGF-I, PEG-IGF-I, and rhInsulin, respectively; Fig. 3, E and F). The data suggest that in differentiated adipocytes, PEG-IGF-I is markedly less potent than rhIGF-I on effector functions driven by rapid InsR binding, whereas the difference decreases toward longer incubation (see also supplemental Table S4).
We next analyzed rhIGF-I and PEG-IGF-I effects on myoblast differentiation and hypertrophy, steady-state functions that involve IGF-I signaling (23). Under growth conditions, only minor differentiation was observed over 4 days (Fig. 4A), whereas differentiating conditions resulted in ϳ40% coverage of the total dish area by troponin-T positive myotubes (Fig. 4B). Fiber area was further increased after treatment with rhIGF-I (Fig. 4C) or PEG-IGF-I (Fig. 4D) for 4 days by ϳ30% with EC 50 values of 63 and 53 pM, respectively (Fig. 4E). Concomitantly, the fusion index as marker for hypertrophy was increased by ϳ50% with EC 50 values of 86 and 111 pM, respectively (Fig. 4F). Finally, in L6 myoblasts known to proliferate in response to IGF-I treatment (24), 18 h treatment with rhIGF-I or PEG-IGF-I induced BrdU uptake with EC 50 values of 106 and 113 pM, respectively (Fig. 4G). Again, equipotency was confirmed in ASPC-1 and MDCK cells (supplemental Fig. S3 and Table S5). Extrapolation of pK values for rhIGF-I versus PEG-IGF-I from all experiments monitoring steady state activity (Fig. 4H, supplemental Table S6) yielded a linear correlation line close to identity (r 2 ϭ 0.939), indicating that rhIGF-I and PEG-IGF-I were equipotent for IGF-IR-mediated functions in different cell types, in contrast with higher potency for rhIGF-I on InsRmediated glucose uptake in adipocytes (Fig. 4H).
In Vivo Pharmacokinetic and Pharmacodynamic Properties-To assess acute side effects in mice, we investigated hypoglycemia elicited by PEG-IGF-I and rhIGF-I in relation to their pharmacokinetic and pharmacodynamic properties. After s.c. injection of PEG-IGF-I up to 10 mg/kg, serum levels increased to 43.3 g/ml, reaching t max at ϳ24 h after injection, with ϳ50% of C max remaining after 48 h. The same treatment with rhIGF-I produced a rapid transient rise in serum to 7.2 g/ml at 10 mg/kg (t max at ϳ1 h), and minimal rhIGF-I left after 6 h (Fig.  5A). PEG-IGF-I produced a marked and sustained reduction of endogenous mouse IGF-I levels by 85% whereas rhIGF-I elicited transient minor effects (49% reduction, Fig. 5B). PEG-IGF-I did not cause hypoglycemic responses at any dose up to an exposure of 20 -30 g/ml. In contrast, rhIGF-I induced significant transient hypoglycemia at all doses with recovery over 2 h only at the 1 mg/kg dose ( Fig. 5C; supplemental Fig. S4) and a threshold for hypoglycemia induction at 0.3-0.5 g/ml, i.e. Ͼ40 times lower than that of PEG-IGF-I. These data showed that mice can be exposed to much higher doses of PEG-IGF-I than rhIGF-I without risk of hypoglycemia. Changes in blood levels of bioactive IGFBP were evaluated by Western Ligand Blot and no changes were found after rhIGF-I treatment, whereas PEG-IGF-I treatment increased IGFBP-2 (Fig. 5, D and E) and IGFBP-3 (Fig. 5, D and F) up to a serum concentration of 30 g/ml (assay validation in supplemental Fig. S5).
In Vivo Effects of Chronic PEG-IGF-I Treatment-To determine the efficacy of PEG-IGF-I, 3-week-old mdx dystrophic mice (25) received 0.3 mg/kg PEG-IGF-I s.c. twice-a-week for 6 weeks. The ability of this treatment to ameliorate use-dependent muscle injury was compared with s.c. minipump treatment with 1 mg/kg/d rhIGF-I (11). Serum PEG-IGF-I levels peaked at 1.1 g/ml after 24 h (dashed line in Fig. 6A). This peak value matched to the exposure of 1.1 g/ml rhIGF-I in mice receiving a constant s.c. minipump infusion, which was lower than estimated earlier (26). Treatment with rhIGF-I resulted in slight but significant hypoglycemia (Fig. 6B, inset) and sensitized the treated mdx mice against glucose challenge, as evident by the rapid removal of glucose from blood (Fig. 6B). In con-trast, PEG-IGF-I had no effect on fasted glucose or glucose tolerance (Fig. 6B). The susceptibility to contraction-induced injury of the tibialis anterior (TA) muscle, which was strongly increased in mdx mice compared with healthy controls was similarly attenuated after 6 weeks s.c. treatment with both PEG-IGF-I and rhIGF-I (Fig. 6C). In vivo effects of rhIGF-I on muscle protection in mdx mice have been reported to involve metabolic changes toward higher catabolic physiology with increased oxidative capacity and the shift to slower fiber phenotypes (11,27). As these changes have been considered of relevance for the therapeutic efficacy, we analyzed physiological and metabolic properties of TA muscles from 18-week-old mdx mice after a treatment period of 6 weeks. Interestingly, minipump rhIGF-I but not PEG-IGF-I treatment prolonged time-to-peak (Fig. 6D) and half-relaxation time (Fig. 6E) of the isometric twitch contraction. In addition, PEG-IGF-I increased succinate dehydrogenase (SDH) activity, a marker of oxidative capacity, to a lesser extent as rhIGF-I (Fig. 6F).

DISCUSSION
PEGylation improves the efficacy of therapeutic proteins by prolonging half-lives and reducing concentration fluctuations (28,29). However, random PEGylation generates monoPEGylated isomer mixtures of the protein (30) with difficulties of purification on a large scale. Therapeutic applications with PEGylated isomer mixtures, however, face challenges of potential single isomer toxicity and reproducibility of isomer ratios between different batches. We circumvented this issue by developing a process that exclusively generated the desired monomer.
The large PEG residue generally inhibits association of the ligand to its different binding targets, which we directly showed for IGFBP using kinetic Biacore assays. We were unable to obtain radiolabeled PEG-IGF-I and thus could not directly assess IGF-IR and InsR binding kinetics, but we suggest similar mechanisms are involved. PEG-IGF-I activity for eliciting fast receptor and downstream signaling (Յ15 min) was reduced, whereas activity differences disappeared toward steady-state conditions, where both rhIGF-I were equipotent. Compared with rhIGF-I, this resulted in reduced fast InsR-mediated glucose uptake in adipocytes but equipotency in steady-state IGF-IR-mediated glucose uptake in myotubes. Since maximum responses at steady state were identical, differences in association kinetics for both InsR and IGF-IR are the likely explanation.
Further argument for this hypothesis can be derived from what we know about the binding mechanism of insulin and IGF-I. InsR and IGF-IR belong to the receptor tyrosine kinase family (Ref. 31, for review) and are covalent transmembrane dimers. Insulin and IGF-I bind to their cognate receptors with affinities of 100 -200 pM (32) and to the non-cognate receptor with affinities 100 -500 times lower (30). It appears that the binding of insulin and IGF-I to their cognate receptors occurs when two binding sites on opposite surfaces of the ligand alternatively crosslink two binding sites on each receptor halve (31)(32)(33). Sites 1 and 2 on insulin have been mapped by alaninescanning mutagenesis (Ref. 34, for review). For the IGF-I binding sites, mutagenesis has shown the involvement of the amino acids 23, 24, 44, 60, and 62, equivalent to site 1 of insulin, and the amino acids 8,9,12,16,54,58, 59, equivalent to site 2 of insulin (34 -36). In addition, site 1 of IGF-I is extended by amino acids 31, 36, and 37 from the C-peptide that are not present in insulin (Ref. 34, for review). The latter residues are thought to bind to an extension of the IGF-IR equivalent to InsR site 1 into the cysteine-rich region.
This would suggest that the Lys-68 residue is outside the receptor binding sites 1 and 2 of IGF-I in the D-domain. Substitution of both Lys-65 and Lys-68 by alanine, however, was reported to cause a 10-fold decrease in IGF-IR affinity (37), a finding that contrasts with data showing that deletion of the D-domain has a negligible effect (38). There is little information on the role of the D-domain in IGFBP binding (35). As evident FIGURE 6. In vivo effects of PEG-IGF-I and rhIGF-I in mdx mice. A, serum drug exposure after single s.c. PEG-IGF-I injection at 0.3 mg/kg, or minipump delivery of rhIGF-I at 1 mg/kg/d in mdx mice (analyzed after 2 weeks at steady state) Ͻn ϭ 3-5Ͼ. B, glucose tolerance test in wild-type (Bl/10), and mdx mice treated with vehicle or rhIGF-I (1 mg/kg/d minipump) or PEG-IGF-I (0.3 mg/kg s.c.) for 2 weeks. Blood was sampled and analyzed for glucose before (inset), and at 15, 30, 60, 90, and 120 min after s.c. injection of glucose Ͻn ϭ 4 -6Ͼ. C, cumulative deficit in force producing capacity of TA muscles (% initial maximum force) in response to lengthening contractions of increasing magnitudes of strain (2-55%) Ͻn ϭ 4 -7Ͼ. D, time-to-peak twitch contraction response of TA muscles after treatment with rhIGF-I or PEG-IGF-I for 6 weeks Ͻn ϭ 5-9Ͼ. E, half-relaxation time of isometric twitch contraction of TA muscles after treatment with rhIGF-I or PEG-IGF-I for 6 weeks Ͻn ϭ 5-9Ͼ. F, succinate dehydrogenase activity in TA muscle after treatment with rhIGF-I or PEG-IGF-I for 6 weeks Ͻn ϭ 3Ͼ.
from the three-dimensional model of IGF-I computationally docked through its site 1 on the L1-CR-L2 domains of the IGF-IR ( Fig. 2A), the only area of IGF-IR for which a crystallographic structure has been determined (36,39), Lys-68 is not located in this interaction site 1. PEGylation at this site is likely to sterically hinder both receptor association and closing of the receptor dimer through site 2. Both of these effects most likely explain the lower in vitro affinity for both IGF-IR and InsR and its impact on fast but not sustained IGF-I effects. The involvement of Lys-68 in site 2 receptor binding is further supported by recent molecular dynamic studies of an atomic structural model of an IGF-IR ectodomain complex suggesting that Lys-68 makes partial contact with Lys-306 in site 2 of the IGF-IR (40). It is known that in the case of insulin binding to the InsR, some site 2 substitutions can reduce the association rate of 125 I-insulin up to 20-fold, while the dissociation rate is also slowed down. The prototype of a natural insulin variant with such behavior is hagfish insulin (20). Such insulin analogues typically have a low metabolic potency due to the slow on-rate affecting fast responses, while the increased residence time on the receptor enhances growth-promoting signaling; thus, hagfish insulin exhibits a mitogenic/metabolic potency ratio of 3.8 (20) to 8.0 3 depending on the assay. Such a profile is definitely therapeutically undesirable for an insulin analog, but to the contrary beneficial for an IGF-I analog like ours. Given the similarity of the InsR and IGF-IR binding mechanisms (32,36), it is not unreasonable to extrapolate the consequences of impaired site 2 interaction to this analog.
Although the association rate of PEG-IGF-I could not be determined directly since it cannot be radiolabeled, it was demonstrated to be much slower. Determining the time required for induction of negative cooperativity by the unlabeled ligand revealed that it took much longer for PEG-IGF-I to induce the same acceleration of dissociation of tracer compared with rhIGF-I. However, the maximal extent of negative cooperativity was not altered by PEGylation (Fig. 2, D and E and supplemental Fig. S2, A-E).
It was surprising to see equipotency for 2 h glucose uptake in myotubes, but this result is consistent with the equipotent downstream signaling within the time frame analyzed. Interestingly, despite the postulated close interaction of the two insulin binding sites within the InsR, a fusion protein between insulin and albumin (albulin) is also equipotent to wild-type insulin both in vitro and in vivo (41).
Both IGF-I and PEG-IGF-I were more potent than insulin in stimulating glucose uptake in differentiated skeletal myotubes while in adipocytes, insulin was the most potent agonist. Muscle cells in vitro (despite their widespread use as a model to study insulin action) respond poorly to insulin compared with IGF-I, because of a lack or paucity of InsR but high abundance of IGF-IR (42). Primary cultures of human myoblasts and myotubes have a Ͼ10-fold excess of IGF-IR over InsR, and microarray studies show that IGF-I is a more potent regulator of gene expression than insulin in these cultures (42), which might explain why muscle-specific InsR knock-out mice are normoglycemic and develop only mild insulin resistance (43). This indicates that the IGF-I system in vivo can compensate for the loss of InsR as it does in vitro (44), and suggests an underestimation of the role of the IGF-I/IGF-IR system as a metabolic regulator in muscle. Results of studies of "insulin signaling" in such models using 100 nM insulin where at least 50% of the IGF-IR are occupied are thus likely to reflect IGF-IR signaling and have to be taken with care. In differentiated adipocytes, the data are consistent with a dominant role of insulin in stimulating glucose transport.
As shown, s.c. application of PEG-IGF-I to mice at high doses markedly increased endogenous IGFBP's, replaced endogenous IGF-I, and hypoglycemia occurred with PEG-IGF-I only at ϳ50 times higher blood levels than with rhIGF-I. Because of their high, subnanomolar affinities, IGFBP are the primary IGF binding sites in the circulation, and it is assumed that their primary function is to protect IGF from degradation or excretion thus assuring a pool of accessible IGF (2). With respect to side effects, the absence of hypoglycemia associated with PEG-IGF-I can be explained by both receptor kinetics and the pharmacodynamic profile that compared with rhIGF-I are expected to selectively eliminate the short-term complications of IGF-I dosing. Differences in short-term versus long-lasting effects of elevated IGFBP's in mouse models (45) support a role for IGFBP's in preserving IGF-I function while decreasing the impact of IGF fluctuations on signaling. Consistent with this concept, elevated concentrations of IGFBP-2 in the circulation protect against type II diabetes and diet-induced obesity (46), and similar functions are considered to apply to IGFBP-3 and -4 (47,48). Elevated IGFBPs presumably lower free and increase complexed, "controllable" IGF, thus favoring mechanisms driven by IGF/IGFBP complexes rather than by free, rapidly acting IGF. In this context it becomes clear why PEG-IGF-I reduced the hypoglycemic risk.
Finally, the potential therapeutic benefit of s.c. PEG-IGF-I treatment for muscular injuries and related muscle disorders was reflected in its ability to ameliorate contraction-induced muscle injury at least to the same degree as rhIGF-I minipump treatment at similar serum levels. This beneficial effect of PEG-IGF-I happened without the catabolic changes in glucose homeostasis or muscle metabolism that were observed for rhIGF-I, strongly suggesting that these effects are not required for amelioration of muscle injury. PEG-IGF-I represents an IGF-I variant with purely homeostatic anabolic properties apparently devoid of short-term side effects. These beneficial properties result presumably from its ability to alter IGF homeostasis and its slower nature of downstream signaling cascade activation. These properties are propitious for safely providing benefit in catabolic conditions and enhancing muscle function. In view of the multiple unsuccessful clinical trials performed with rhIGF-I or IPLEX so far, PEG-IGF-I represents a novel therapeutic approach for those conditions, where short-term IGF-I side effects have up to now precluded the achievement of therapeutic levels. Moreover, site-specific PEGylation might represent a general approach for drugs that presently cannot be used because of their side effects at efficacious doses.