The History of the Insulin-Like Growth Factor System

The growth hormone (GH)–insulin-like growth factor (IGF) cascade is central to the regulation of growth and metabolism. We describe key historical events related to the components of the IGF system, with a focus on the peptide hormones, IGF-I and -II, their cell surface receptors, and the IGF binding proteins (IGFBPs) and IGFBP proteases that regulate the availability of the peptide hormones to interact with their receptors in the relevant target tissues. We describe landmark events, including the development of the original somatomedin hypothesis and its modification as new evidence has become available from experiments at the molecular and cellular levels, whole animal and tissue-specific gene knockouts, studies of cancer epidemiology, identification of prismatic human cases, and short- and long-term clinical trials of IGF-I therapy in humans. In addition, this new evidence has expanded our clinical definition of GH insensitivity (GHI) beyond GH receptor mutations (classic Laron syndrome) to include conditions that cause primary IGF deficiency by impacting post-receptor signal transduction, IGF production, IGF availability to interact with the IGF-I receptor (IGF-1R), and defects in the IGF-1R itself. We also discuss the clinical aspects of IGFs from their description as insulin-like activity, the use of IGF-I levels in the diagnosis and treatment of GH deficiency, and the use of recombinant human IGF-I for therapy of children with GHI.

Before the development of radioimmunoassays (RIAs), insulin was typically bioassayed by measuring glucose uptake by rat epididymal fat or rat diaphragm. Many of these bioassays lacked both specificity and sensitivity, and had very large inter-assay variability, especially for imprecise end-points such as convulsive threshold in experimental animals. Increased sensitivity with more reproducible results was determined by Feldman et al. [1], who were able to detect a hypoglycemic effect of various preparations in the blood of adrenalectomized rats following anoxia or metrazole, an antimicrobial which lowers the seizure threshold [1]. Several of the same investigators increased the sensitivity of the assay for insulin by using hypophysectomized, adreno-demedullated mice and rats [2]. When specific RIAs became available, one could quantitate the amount of insulin rather than its activity [3] and it became apparent that immunoactive insulin represented only about 10% of bioassayable insulin [4]. Fasting plasma insulin levels were 2–5 times higher in the rat diaphragm assay and serum insulin levels 10-fold higher in the fat pad assay [4]. What then was the insulin-like activity (ILA) measured in the bioassays?

With addition of ¹³¹I- insulin followed by extraction with acid-alcohol, only 10 to 15% of the original ILA was associated with the radioactive insulin [5]. More than 50% of the activity remained after pancreatectomy [5], suggesting that ILA might originate from sources other than the pancreas. Furthermore, some patients with large (but not necessarily rapidly growing) non-endocrine tumors had hypoglycemia in the absence of elevated immunoassayable insulin concentrations. However, the mechanism of the hypoglycemia was uncertain and likely intersected with the great majority of ILA in the blood being non-suppressible with antibodies to insulin (NSILA). Among the theories summarized by Unger [6], work soon focused on the secretion of an “insulinoid” substance from the tumor.

Froesch and colleagues showed that antibody suppressible ILA (SILA) was presumably equivalent to insulin, itself, and was independent from NSILA. The former comprised only 7% of the total ILA, but was much higher in some patients with hypoglycemic islet-cell tumors and rose in the plasma of subjects following the ingestion of glucose [7]. ILA persisted in the serum of pancreatectomized animals. Schoeffling and colleagues [8] showed that part of the ILA depended on a pituitary factor. Following hypophysectomy, the ILA decreased by about half but did not decrease further after subsequent pancreatectomy. Importantly, no serum insulin was detected by RIA, nor was there pancreatic tissue on autopsy. Thus NSILA was not insulin, and did not come from the pancreas [4].

In 1957, Salmon and Daughaday presented evidence for a GH-dependent molecule, “sulfation factor” (SF) [9]. Its properties included increasing the uptake and incorporation of ³⁵SO₄ into chondroitin sulfate of rat costochondral cartilage incubated in vitro. Plasma from normal rats increased this incorporation ∼2.5-fold, compared to plasma from hypophysectomized rats. When the hypophysectomized rats were treated with exogenous GH, their plasma was able to increase the incorporation of ³⁵SO₄. However, when GH was added directly to the incubation medium, it did not increase the rate of incorporation of ³⁵SO₄. Adding insulin to the incubation medium also led to modest increases in incorporation of sulfate, but only with remarkably high concentrations. These experiments indicated that the SF depended on the action of GH, but clearly differed from either GH or insulin, themselves. Further investigation by Salmon and DuVall indicated that a serum fraction that contained SF activity also stimulated incorporation of amino acids into protein-polysaccharide complexes, and nucleic acids into RNA and DNA in cartilage from hypophysectomized rats [10]. Under the same conditions, GH added to the incubation medium again had no effect. The plasma sulfation activity in the circulation was associated with larger molecular weight proteins, but could be transferred to smaller molecular components by denaturing the plasma proteins by boiling [10] or by extraction with acid ethanol [11]. Starting with plasma from patients with acromegaly, Judson Van Wyk et al. purified two highly active fractions (SF and thymidine factor) with a provisional molecular weight of about 8 kDa [11]. The biological activity of SF was not limited to cartilage. SF had insulin-like actions on rat diaphragm and rat adipose tissue. The ILA of SF was not neutralized using anti-insulin serum. This demonstrated that SF was identical to the smaller molecular weight component of NSILA [12].

In 1972, the terms SF and NSILA were replaced by the term “somatomedin,” denoting a substance capable of mediating the effects of GH [12]. The somatomedin hypothesis states that the stimulation of sulfate uptake (by costochondral cartilage) is due to a GH-dependent factor, rather than GH itself. The key experimental finding (noted above) was that bovine GH added to the incubation medium produced little or no stimulation of uptake, even when added to the serum from hypophysectomized rats; however, direct GH treatment of the hypophysectomized rat restored the ability of their serum to stimulate sulfate uptake. This hypothesis has been expanded with the “dual effector theory,” which defines individual functions for GH and IGF-I. GH promotes the differentiation of precursors, for example, cartilage cells at the epiphysis and IGF-I leads to their clonal expansion [13]. Thus, the original endocrine hypothesis had become one of paracrine/autocrine action with some biological activities due to GH (principally metabolic actions, such as lipolysis and diabetogenic effects), while skeletal growth is mediated through the IGFs. Whether or not all growth-promoting actions of GH are mediated through the IGF system remains a subject of active controversy and research. In an editorial, Daughaday and other investigators proposed the generic name “somatomedin” for all of these factors, since “the insulin-like activity cannot be neutralized by anti-insulin serum and follows SF through multiple fractionation procedures [11], suggesting that SF is identical with or very similar to the smaller molecular weight component of the non-suppressible insulin-like activity (NSILA-s)…” [14]. The prefix “somato” was intended to connote both the hormonal relationship to somatropin (GH) and to the soma (body), which is the target tissue of the agent. “Medin” is included in the name to indicate that it is an intermediary in the action of somatropin (GH) [12]. The somatomedins identified at that time included somatomedin A, B, and C.

More or less concurrently, Dulak and Temin studied the activities of multiplication-stimulating activity (MSA) [15, 16]. Rat MSA, a mitogen for cultured fibroblasts, was compared with SF in multiple assays and stimulated DNA synthesis with superimposable dose-response curves. The unlabeled growth factor inhibited the binding of the iodinated ligands with comparable potency; however, there were some differences between them, as they differentially bound to cell surface receptors on specific cell lines [15]. For example, somatomedin A had ∼10% the binding activity of NSILA-s and MSA was fully active as NSILA-s at the NSILA-s receptor. Each liver cell had apparently 50 times more insulin receptors that those for NSILA-s [17]. Rechler and colleagues noted the marked similarity of purified human somatomedin A and rat MSA, both to stimulate DNA synthesis in chick embryo fibroblasts and in binding assays [17]. Unlabeled MSA and somatomedin A inhibited the binding of ¹²⁵I-labeled MSA and ¹²⁵I-labeled somatomedin A to each of the receptors with indistinguishable potency [17].

Further evidence that these factors were distinct from insulin was presented by Megyesi et al., who showed that there were distinct receptors for these factors on plasma membranes from hepatocytes. NSILA-s had <1% affinity for the insulin receptor and the most highly purified preparations of NSILA-s had >300,000-fold greater affinity for the NSILA-s receptor than for the insulin receptor [18].

In 1976, Rinderknecht and Humbel isolated two active substances from human serum, which, owing to their structural resemblance to pro-insulin, were renamed insulin-like growth factors I and II [19]. They later purified and determined the amino acid sequence of IGF-I [20]. These two peptides had several structural similarities with proinsulin: the three disulfide bonds and the six half-cystine residues were conserved and the hydrophobic amino acid residues that make up the cores of the monomers were also conserved, permitting the individual peptides to maintain their 3-dimensional configurations. The surface-facing (hydrophilic) residues were not as well conserved, perhaps explaining why the peptides reacted so differently immunologically. Another striking difference between the IGFs and proinsulin is the lack of the double basic amino acids at the termini of the connecting peptide, which makes it relatively easy for the C-peptide to be enzymatically cleaved from proinsulin to produce insulin, but not so easily cleaved for the IGFs, both of which retain their respective C-peptides.

By 1978, therefore, the relatedness among the peptides denoted NSILA, SF, MSA, and somatomedins A and C had been confirmed [21]. Somatomedin A was determined to be IGF-II and somatomedin C was determined to be IGF-I. Where did somatomedin B go? As of 1977, somatomedin B, although under GH control, was no longer considered a somatomedin, because it did not stimulate sulfate uptake into cartilage or have ILA [22]. Somatomedin B was ultimately found to represent the amino terminus of vitronectin (1–44) and its growth promoting activity was determined to be the result of contamination with epidermal growth factor [23]. In 1985, Bell and colleagues isolated the genes for IGF-I and IGF-II and determined that the genes for IGF-II and insulin were contiguous [24].

Historically, IGF concentrations had been determined by relatively nonspecific bioassays and, later, radioreceptor assays. More sensitive and specific assessment of IGF-I (somatomedin C, Sm-C) levels in serum from humans and animals was facilitated by the development of a specific radioimmunoassay (RIA) by Richard Furlanetto et al. [22]. Specific radioreceptor assays and RIAs provided comparable results and GH dependency of Sm-C [25]. Using these assay methods, A. Joseph D’Ercole et al. [26] demonstrated production of Sm-C in multiple tissues supporting potential autocrine and paracrine mechanisms of action. The IGF-I-specific RIA and subsequent generations of IGF-I-specific assays [27, 28] have been used to demonstrate that human IGF-I levels are GH dependent, have little diurnal or day-to-day variation, but vary with age, sex, and pubertal status [27]. In contrast, IGF-II levels show little GH dependence and are relatively stable in the postnatal period. IGF-I levels are used as a screening tool for the diagnosis of GH deficiency (GHD) and GH excess [29]. The development of an assay for free IGF-I in 1994 by Jan Frystyk and colleagues using ultrafiltration by centrifugation was an important tool for quantifying the IGF-I and IGF-II available to interact with the IGF-1R in serum, pleural effusion, cerebrospinal fluid, and other samples collected in animal models and clinical conditions [30-32]. In a series of epidemiological studies of circulating IGF-I and IGF-II levels in the late 1990s, Michael Pollak and colleagues demonstrated that individuals with levels in the highest tertile compared to the lowest tertile have an increased risk of several different cancers including breast, colon, and prostate [33-35], although not all subsequent studies have confirmed these observations.

In clinical practice, adjustment of biosynthetic human (h)GH dosing has been based upon weight or body surface area, height velocity, and/or IGF-I levels. The use of IGF-I levels to guide dose adjustment of biosynthetic hGH therapy in children and adults has been recommended by some for both safety and short-term efficacy purposes [36]. From a safety perspective, it has been recommended that biosynthetic hGH replacement therapy result in IGF-I levels that fall within the normal range (i.e., ≤+2 SDS) [37, 38]. Targeting an IGF-I in the upper part of the normal range (+1 to +2 SDS) has been suggested to increase short-term linear growth in children with GH deficiency (GHD) and idiopathic short stature (ISS) [39]. However, long-term efficacy of IGF-based dosing in children has not been demonstrated. Conventional weight-based dosing of biosynthetic hGH in children with GHD achieves an IGF-I level close to 0 SDS depending upon the dose used [40].

The circulating concentrations of IGFs exceed those of insulin by approximately 1,000 fold [41]. Even if the IGFs have a small amount of insulin-like biological activity, one would expect that to be sufficient to keep all animals permanently hypoglycemic [41]. The explanation for this apparent paradox probably lies in the IGF-binding proteins (IGFBPs). Evidence for one or more circulating protein(s) that bind or “carry” IGFs (and inhibit their biological activity) comes from biochemical observations. When serum is subjected to gel filtration at neutral pH, NSILA-s elutes with proteins of about 70–150 kDa [14, 42]; however, when an acetone powder of serum is subjected to gel chromatography in 6 M acetic acid, an additional biologically active fraction is also noted, at about 6–10 kDa. This suggests a ligand/binding protein interaction in which the ligand dissociates in the strong acid environment. Further work showed exquisite specificity of binding with a limited capacity for a biologically active peptide with the characteristics of a carrier protein-binding of ¹²⁵I-labelled NSILA with high sensitivity to unlabeled NSILA, but not to other hormonal ligands, including insulin and GH [42].

Raymond L. Hintz et al. identified high affinity, low capacity, reversible, saturable binding sites for human somatomedin C in human plasma in 1977 [43]. Others, using multiple biochemical techniques, particularly western ligand blotting, identified the currently known family of six IGFBPs in the 1980s [44, 45]. The IGFBPs are a family of 6 proteins with high-affinity binding for the IGFs in conserved regions in similar sites in each of the molecules [45-50]. Based on their amino acid sequences (determined by cDNA cloning) the six human IGFBPs have similar sizes and sequences, featuring closely corresponding cysteine residues. SF was initially purified from large molecular weight complexes in the plasma. Further purification of these complexes from plasma identified their components as IGFBP-3 and acid labile subunit, IGFALS. Subsequently, IGFBP-5 was also shown to be involved in formation of a minority of the circulating ternary complexes. IGFBP-3 and IGFBP-5 are in the greatest concentration in the circulation, both with very high affinities for IGF-I and -II, and are the major carriers of IGFs in serum. The IGFBPs act to deliver IGF-I and IGF-II to target tissues, act as a local reservoirs of IGFs through regulating the bioavailability of their metabolic and mitogenic actions, and have been shown to have potential IGF-independent effects [47, 48]. In a series of knockout mouse experiments by John Pintar and colleagues, the role of IGFBPs on growth and metabolism was evaluated [51]. Single knockout of IGFBP-1 and IGFBP-2 had no impact on linear growth or metabolism [51]. Single knockout of IGFBP-3 and IGFBP-5 had no impact on growth, [52] demonstrating redundancy in the action of these IGFBPs. Single knockout of IGFBP-4 had a mild (5–10%) negative impact on prenatal growth [52]. However, the triple knockout of IGFBP-3, IGFBP-4, and IGFBP-5 led to 20% reduction in adult size [52] associated with low levels of total and bioactive IGF-I. In addition, the triple IGFBP knockout mice had an increased insulin secretory response to glucose, demonstrating the metabolic role of these IGFBPs [52]. The IGFBPs may act in a “local” or systemic manner. The former actions are for autocrine/paracrine effects, such as at the growth plate and the latter are mediated via transport through the circulation (endocrine) for growth and metabolism. They bind to an additional protein, the IGFALS, to form large ternary complexes which prolong the half-lives of the IGFs in the circulation and thus their biological half-life [53]. The autocrine/paracrine effects are tissue specific and thus difficult to summarize succinctly. They may act as intermediaries between nutrient availability and growth regulation, thereby leading to potential roles in cancer, growth, and metabolism [48, 54-56].

In 1989, Robert C. Baxter et al. identified the acid labile subunit (IGFALS) as the 85-kDa glycoprotein “alpha subunit” of the circulating high molecular weight IGFBP complex [57]. IGFALS is produced primarily in hepatocytes and secreted into the serum. IGFALS binds to IGFBP-3 (or IGFBP-5) and IGF-I or IGF-II in serum to form the 150-kDa ternary complex [53, 58]. The formation of the ternary complex allows prolongation of the half-life of circulating IGFs [59] and delivery of IGF-I produced by the liver to the target tissues, thereby supporting the GH dependency of IGF-I’s endocrine function. Although the igfals knockout mouse did not demonstrate any growth impairment [60], the important role of IGFALS in delivery of endocrine IGF-I in humans was highlighted by identification of individuals with IGF-I deficiency who harbored mutations in the IGFALS gene resulting in growth impairment by Domene et al. in 2004 [61].

The discovery of IGFBP proteases in the serum of pregnant women led to identification of multiple proteases capable of degrading specific IGFBPs and releasing bound IGFs, thereby facilitating the access of IGFs to their receptors [62, 63]. One well-characterized IGFBP protease is human pregnancy-associated plasma protein-A (PAPP-A), first identified in 1974 as a protein of placental origin [64]; however, no unequivocal biological activity was determined until Lawrence et al. showed proteolytic activity toward IGFBP-4 [65]. PAPP-A and the closely related protein, PAPP-A2, are highly specific metalloproteases called pappalysins which can selectively cleave the IGF peptide from its binding protein and thus regulate the amount of free peptide, allowing binding to the IGF receptor [66]. Degradation of the IGFBPs in circulation reduces delivery of circulating IGFs to the target tissue. PAPP-A has particular affinity for IGFBPs- 2, -4, and -5 and -A2 toward -3 and -5, the two binding proteins capable of forming ternary complexes (with IGFALS) with IGF-I. In 2000, Overgaard et al. demonstrated that PAPP-A forms large molecular weight complexes with its physiologic inhibitor, the pro-form of eosinophilic major basic protein (proMBP) [67]. ProMBP binds the majority of serum PAPP-A which limits PAPP-A activity in the circulation [32]. In 2004 and 2011, Conover et al. demonstrated impaired growth in pappa and pappa2 knockout mice, respectively [68, 69]. In 2016, humans with mutations in the PAPPA2 gene encoding PAPP-A2 were identified by Andrew Dauber et al. and found to have IGF deficiency and short stature [70].

In 2015, stanniocalcin (STC), a known regulator of postnatal growth in mice [71], was shown to inhibit the proteolytic action of PAPP-A and PAPP-A2 [72]. Overexpression of STC1 and STC2 in mice led to growth impairment [32, 73]. stc2 knockout mice were 10–15% larger than controls [71]. Although stc2 knockout mice had normal levels of circulating IGF-I and IGFALS, it is likely that the absence of the inhibition of PAPP-A by STC2 at the tissue level led to increased availability of IGFs at the target tissue [71]. In 2017, genome-wide association studies identified coding variants in STC2 in which carriers with the most significant rare variant at STC 2 were ∼2 cm taller than non-carriers [74]. The proteolytic action of PAPP-A and PAPP-A2 on the IGFBPs to release IGFs to interact with their receptors demonstrates the important role of the members of the PAPP-A, STC, and IGFBP protein families as intermediaries between nutrient availability and growth regulation [32, 75].

The cell surface receptors for insulin (INSR) and IGF-I (IGF-1R) are closely related integral membrane proteins purified from placental membranes and sequenced by Ullrich and colleagues in the mid-1980s [76, 77]. The purification was facilitated by the known functional activities and binding characteristics of each receptor.

In the late 1980s, the interactions between the IGFs and their receptors were delineated. The extracellular alpha subunits contain ligand binding domains [78, 79] which interact with the requisite ligand. Due to the close homology of the receptors and the similarities in three-dimensional structure of the ligands, each receptor can bind and transduce signals for either ligand. However, the affinity of each receptor for its cognate ligand is approximately 100-fold greater than for the alternate ligand. For the NSILA-s (IGF-1R) the 50% inhibition points were 3–5 ng/mL for NSILA-s, but 1 mg/mL for insulin; for the INSR these were 20 ng/mL for insulin and 6 μg/mL for NSILA [18].

The IGF-II receptor, also known as the mannose-6-phosphate receptor, has no signal transduction component and functions to clear IGF-II from the target tissue [80, 81]. The actions of IGF-II occur through its interactions with the IGF-1R and, to some degree, the INSR. Using NIH 3T3 fibroblasts transfected with human IGF1R cDNA, Emily Germain-Lee et al. [82] showed that the IGF-1R binds IGF-I with a 15- to 20-fold higher affinity than IGF-II. During fetal life and in some cancers, a fetal form of the INSR (isoform A) and hybrid receptors consisting of an INSR alpha-beta dimer cross-linked with an IGF-1R alpha-beta dimer are present and demonstrate higher affinity for IGF-II [83, 84].

In 1985, Ullrich et al. characterized the tyrosine kinase activity of the INSR and IGF-1R. The intracellular portion of each beta subunit contains a tyrosine kinase domain and an ATP binding site [76, 77]. Numerous cellular proteins are tyrosine phosphorylated in response to activation of INSR and IGF-1R, leading to the respective metabolic, proliferative, and differentiation effects of each receptor. This downstream signal transduction cascade involving the Ras/Raf/MAP kinase pathway was characterized by numerous groups in the 1980s and 1990s (reviewed in LeRoith et al., 1995 [85]). The presence of elevated levels of the IGF-IR in certain cancers led to studies of its role in neoplasia [85] and attempts to block IGF-1R signaling as a means to treat cancer [86].

One key question involving the somatomedin hypothesis was whether somatomedins were responsible for all growth-related effects of GH or if there were GH-dependent effects that were somatomedin independent. The “dual effector” version of the somatomedin hypothesis ascribes separate individual functions to GH and IGF-I [87]. The key example of a GH-dependent, somatomedin-independent event is the recruitment of chondrocytes from the resting zone of the epiphysis into the proliferative zone [13]. Once GH has stimulated this event, IGF-I promotes the proliferation and hypertrophy of the chondrocytes, resulting in lengthening of the long bone and linear growth [13]. The lack of GH-dependent action on the resting zone chondrocytes to promote entry into the growth plate may limit growth of animals and humans with GHI, including their response to IGF-I therapy (described below). The GH-dependent stimulation of production of hepatic IGF-I, IGFALS, and IGFBPs makes up the majority of circulating IGFs. It remains to be determined, however, to what degree the endocrine source of IGF-I versus GH-dependent local autocrine and paracrine IGF production [26] are necessary for growth in humans (discussed below). There are also non-growth-related, GH-dependent, somatomedin-independent effects on metabolism, including lipolysis and regulation of immune function.

Between 1990 and 1996, using targeted mutagenesis of genes inserted into embryonic stem cells to generate knockout mice, Efstratiadis and colleagues described the phenotypes of igf1, igf2, igf1r, and igf2r mutant mice. These studies led to important insights into the pre- and postnatal function of these components of the IGF system [88-92]. When igf2 was knocked out, mice had prenatal growth failure, resulting in a 40% reduction in birth weight, followed by normal postnatal growth. In contrast, when igf1 was knocked out, the mice had similar prenatal growth failure, but accompanied by postnatal growth failure resulting in a 70% reduction in adult size. Knockout of igf1r led to prenatal growth failure resulting in a 55% reduction in birth weight [91]. The igf1r knockout mice reportedly died within the first day of life from respiratory failure, with general organ hypoplasia. The relevance of these findings in mice for human growth was supported by reports of patients with combined prenatal and postnatal growth failure found to have homozygous deletion or mutations in the IGF1 gene or heterozygous mutations in the IGF1R gene [93-96]. In addition, heterozygous carriers of mutations in the IGF1R gene were shorter than non-affected family members [95]. There are no known humans with a homozygous IGF1R mutation.

Mouse models with mutations of both igf1 and igf1r do not differ from the single mutant igf1r, although those with homozygous mutations of igf1r and igf2 are further dwarfed (∼30% normal size) indicating an interaction of IGF-II with the IGF-1R in utero[92]. In the absence of the igf2r, circulating IGF-II levels are high in utero and it is able to act through the IGF-1R to cause fetal overgrowth, but also a lethal phenotype. The insulin receptor mediates a portion of IGF-II’s prenatal growth promoting actions [97]. Unlike the IGF1 gene, the human IGF2 gene is imprinted, so that only the paternally inherited allele is expressed. Thus, mutations of the paternal allele, but not the maternal allele, cause human growth impairment [98]. In addition, loss of methylation at the imprinting center region of the 11p15 leads to reduced local IGF2 expression associated with pre- and postnatal growth failure in children with Russell-Silver syndrome [99].

Following the determination of the impact of global knockout of the IGF system ligands and receptors, a number of studies further guided our understanding of their tissue-specific actions. Using the Cre-loxP expression system and cell-specific promotors, LeRoith and colleagues began by asking the question “What is the function of the IGF-I produced in the liver?” [100]. The majority of the circulating IGF-I is produced in hepatocytes under regulation by GH. The lack of IGF-1Rs on hepatocytes supports the endocrine function of hepatocyte-produced IGF-I. Obliteration of hepatocyte IGF-I production reduced circulating IGF-I levels by 75%, but had virtually no impact on linear growth [101]. This supports a role for the GH-dependent local production of IGF-1 acting in an autocrine and/or paracrine fashion [26] to stimulate growth. Interestingly, the igfals knockout mouse demonstrated a similar reduction in circulating IGF-I levels, with an additional reduction in circulating IGFBP-3 levels without an impact on linear growth [60]. In each of these models, elevated levels of GH may have compensated for the reduction in circulating IGF-I [100]. Additional studies of tissue-specific reduction of igf1 expression investigated the effects on muscle, bone, fat, neural, and hematopoietic tissue function [100].

The essential role of IGF-I in mammalian growth was established by the creation of transgenic “knockout” mice lacking a functional igf1 gene, which have birth weights ∼60% of normal and grow poorly thereafter [88]. Corresponding but naturally occurring mutations in the human IGF1 gene can cause severe intrauterine growth retardation, with severely compromised subsequent growth [96]. As might be predicted, compound heterozygous loss-of-function mutations in the human IGF1R can cause human growth retardation, but this is rare: only two of 92 children with severe growth retardation surveyed carried loss-of-function mutations in their IGF1R genes [93]. These findings thus demonstrate that the human IGF system is essential for normal growth, but that their mutations are not common causes of growth failure in the human.

The role of knockout mouse models of the pappalysins, stanniocalcins, and IGFBPs is described in the section on IGFBPS and proteases (above).

The GH receptor (GHR) belongs to the cytokine family of receptors involved in regulation of the immune system. However, individuals with GH insensitivity due to mutations in the GHR gene (GHR) do not have an immune phenotype. The downstream signaling of the GHR involves activation of JAK kinase, NFkB, and STATs (Janus kinase; nuclear factor, kappa B; signal transducers and activators of transcription) to translate the cell surface signal triggered by activation of the GHR by GH to the nucleus with resultant transcriptional activation of GH-responsive genes including IGF1. The GHR shares some members of its downstream signaling pathway with other cytokine receptors. Due to this relationship, it was hypothesized that humans with defects in the components of the downstream signaling cascade would have immune deficiency in addition to the growth characteristics of GHI [102]. The recognition of this potential phenotype led to the search for individuals who met this criterion resulting in identification of mutations in STAT5B. In 2003, Kofoed et al. identified a 16-year-old female with a classical phenotype of GHI including severe postnatal growth failure, but accompanied by immune dysfunction associated with recurrent lung infections [103]. Identification of additional individuals with STAT5B mutations demonstrated variable growth impairment with adult heights ranging from –2 to –10 SDS [102]. These studies established the mechanism of GH-dependent IGF-I production via activation of STAT5b. This work led to the expanded definition of GHI to include defects in the GHR, the GHR signaling cascade (STAT5B), defects in the IGF1 gene, defects in IGF transport (IGFALS mutations), defects in IGF bioavailability (PAPPA2 mutations), and defects in IGF signaling (IGF1R mutations) [102]. In addition, identification of the link between STAT5b and severe short stature expanded the definition of GHI to incorporate immunodeficiency as a potential associated finding [102].

A cornerstone of therapy in endocrinology is the replacement of the native hormones missing in deficient individuals. The concept of “IGF deficiency” as a biochemical diagnosis was first proposed by Ron G. Rosenfeld in 1996 [104]. IGF-I is the logical therapeutic agent for those with IGF-I deficiency. Therapy with biosynthetic hIGF-I for primary IGF-I deficiency due to GHI was developed through clinical investigations by Zvi Laron in Israel [105]; Louis Underwood, Judson Van Wyk, Steven Chernausek, and Phillippe Backeljauw in North Carolina, USA [106-108]; Ron Rosenfeld, Jaime Guevara-Aguirre, and Arlan Rosenbloom in Ecuador [109, 110]; and Michael Ranke, Martin Savage, and Pierre Chatelain in Europe [111]. Laron et al. reported growth data on 5 children with Laron-type short stature receiving biosynthetic hIGF-I therapy for a period of 3 to 10 months [105]. In a report from Chernausek et al. on the long-term treatment of GH-insensitive (IGF-I deficient) patients with biosynthetic hIGF-I [107], height velocity increased in virtually all, but not to the degree noted in patients with severe GH deficiency when treated with biosynthetic hGH [107]. In the first year, children with GHI grew 8.0 cm/year. In six children who reached adult height, the height SDS change from baseline ranged from 1.6 to 4.3 SDS and 5 of the 6 children gained an estimate of more than 10 cm in adult height compared to untreated historical controls with GHI [107]. Biosynthetic hIGF-I was approved for commercial therapy by the United States Food and Drug Administration in 2005 and European Medicines Agency in 2007 and continues to be used in the treatment of children with severe short stature and primary IGF-I deficiency. More recent data from the real world use of biosynthetic hIGF-I from the European Union Increlex® Growth Forum Database (IGFD) Registry demonstrated a first-year height velocity in prepubertal children of 7.3 ± 2.0 cm/year [112]. Adverse events associated with biosynthetic hIGF-I therapy include hypoglycemia, lipohypertrophy, tonsillar hypertrophy, injection site reactions, and headaches.

The relatively modest height velocity response to biosynthetic hIGF-I therapy in children with GHI compared to more robust growth responses to biosynthetic hGH therapy in children with GHD may have several causes [113]. First, administration of biosynthetic hIGF-I in individuals with low levels of IGFALS and IGFBPs may limit the ability of biosynthetic hIGF-I to reach the target tissue through decreased half-life and lack of delivery mechanism. Suppression of GH production by feedback during administration of biosynthetic hIGF-I may play a role in individuals with partial GHI. There may be long-term negative growth effects of prenatal IGF deficiency on postnatal growth that impair the response to biosynthetic hIGF-I therapy. Finally, the lack of GH-dependent action on the resting zone chondrocytes to promote entry into the growth plate may limit the growth response to biosynthetic hIGF-I.

Attempts to overcome the short half-life of biosynthetic hIGF-I were made by combining it with biosynthetic hIGFBP-3 prior to administration, leading to approval of this combination, mecasermin rinfabate, for treatment of severe growth failure in 2005 [114]. However, this drug was removed from the market in 2007 due to patent infringement. Studies of mecasermin rinfabate as a continuous infusion to prevent bronchopulmonary dysplasia in premature infants are ongoing [115].

We have highlighted the historical events in identification and elucidation of the structure and function of the components of the IGF system at the molecular, cellular, tissue, and organism levels. We have attempted to capture the contributions of key investigators and research teams in the evolution of the somatomedin hypothesis, the definition of GHI, and the clinical utility of IGF-I as a diagnostic tool and therapeutic agent as new evidence has become available from experiments at the molecular and cellular levels, whole animal and tissue-specific gene knockouts, studies of cancer epidemiology, identification of prismatic human cases, and short- and long-term clinical trials of IGF-I therapy in humans. Since the identification of the insulin-like activity, we have made much progress in understanding how the amount of free IGF available to interact with the IGF-1R is tightly regulated by the balance of IGFBPs, IGFBP proteases, and IGFBP protease inhibitors and how that relates to growth disorders in humans. Many astute scientific and clinical observations, new techniques, and many hours have led to these findings resulting in incremental and, at times, exponential leaps in our understanding of these important and increasingly complex pathways that regulate normal and abnormal growth in animals and humans. Many questions remain unanswered and new questions will arise as new information becomes available. It is our hope that these studies, and those that remain to be conducted by new investigators, new techniques, and more hours, will continue to lead to an improved understanding of the pathophysiology of human growth conditions, like GHI, resulting in better tools and paradigms for the diagnosis and treatment of these conditions and other conditions that the GH-IGF cascade impacts.

Dr. Miller is a consultant for Abbvie, Ascendis Pharma, BioMarin, Bristol Myers Squibb, EMD Serono, Endo Pharmaceuticals, Novo Nordisk, Orchard Therapeutics, Pfizer, Tolmar, and Vertice and has received research support from Alexion, Abbvie, Aeterna Zentaris, Amgen, Amicus, Lumos Pharma, Lysogene, Novo Nordisk, OPKO Health Pfizer, Prevail Therapeutics, and Sangamo Therapeutics. Dr. Rogol is a consultant for Antares Pharma, Ascendis Pharma A/S, Clarus Therapeutics, Lumos Pharma, United States Anti-doping Agency (USADA), and Ultragenyx Pharmaceutical. Dr. Rosenfeld is a consultant for Ascendis, BioMarin, Cavalry, DNARx, Lumos, OPKO, and Pfizer.

There was no external funding source.

Brad Miller, Alan Rogol, and Ron Rosenfeld conceptualized the paper, contributed to writing the original manuscript, and approved the final version.

There were no data generated for this manuscript.