PAPP-A and the IGF system in atherosclerosis: what’s up, what’s down?

Abstract

Pregnancy-associated plasma protein-A (PAPP-A) is a metalloproteinase with a well-established role in releasing bioactive insulin-like growth factor-1 (IGF-1) from IGF-binding protein-2, -4, and -5 by proteolytic processing of these. The IGF system has repeatedly been suggested to be involved in the pathology of atherosclerosis, and both PAPP-A and IGF-1 are proposed biomarkers and therapeutic targets for this disease. Several experimental approaches based on atherosclerosis mouse models have been undertaken to obtain causative and mechanistic insight to the role of these molecules in atherogenesis. However, reports seem conflicting. The literature suggests that PAPP-A is detrimental, while IGF-1 is beneficial. This raises important questions that need to be addressed. Here we summarize the various studies and discuss potential underlying explanations for this seemingly inconsistency with the objective of better understanding complexities and limitations when manipulating the IGF system in mouse models of atherosclerosis. A debate clarifying what’s up and what’s down is highly warranted going forward with the ultimate goal of improving atherosclerosis therapy by targeting the IGF system.

INTRODUCTION

The intention of this review is to account for the seemingly inconsistent studies reporting an effect of the insulin-like growth factor-1 (IGF-1) system on atherosclerosis. Even among studies based on widely used and reproducible atherosclerosis mouse models reports are incongruent. In this review, we focus on all reported studies accessible in the PubMed database to date in which the IGF system is directly manipulated in mouse models of atherosclerosis. We discuss potential reasons for the inconsistent conclusions emphasizing problematic issues resulting from manipulating the IGF system in a whole organism setting, which could complicate interpretation of results.

It should be noted that the majority of reported in vivo studies involve genetic manipulation of vascular smooth muscle cells (VSMCs). The reason for this apparent bias is presumably that VSMCs are known to be the principal producers of IGF components in the artery wall (see expression of igf-1, igf-1r, and papp-a in atherosclerosis) and could therefore be assumed to be most relevant. Indeed, endothelial cell- and macrophage-dependent processes of atherosclerosis have been proposed, and future genetic in vivo models addressing these cell types will most likely deepen our understanding of the involvement of the IGF system in atherosclerosis. A summary of the numerous cellular mechanisms proposed to be affected by the IGF system in atherosclerosis is not within the scope of this review. Instead we refer to recent reviews (44, 137).

PREGNANCY-ASSOCIATED PLASMA PROTEIN-A AND THE IGF SYSTEM

In the classic view of the somatotroph axis, growth hormone (GH) secreted from the pituitary gland stimulates hepatocytes to secrete IGF-1, which in turn acts on tissues expressing IGF-1 receptors (IGF-1R) and/or insulin receptors (INSRs) (72). As its name implies, IGF-1 is structurally and functionally related to insulin (90, 91, 95); however, while insulin secretion is restricted to β-cells of the endocrine pancreas (10), IGF-1 is expressed in most tissues and its actions are exerted by endo-, auto-, and paracrine modes (30, 105, 135). IGF-1 is not a hormone like insulin as it is present at a relatively steady level in the blood (106). Consequently, several levels of extracellular regulation are required between IGF-1 secretion and receptor activation. First, the majority of IGF-1 is bound by one of six IGF-binding proteins (IGFBPs), which typically sequester IGF-1 and prevent receptor activation (34). Second, IGFBP proteinases can liberate IGF-1 by proteolytic cleavage of IGFBPs (64), and third, IGFBP proteinases are regulated by proteinase inhibitors (52, 60, 80, 81). IGFBP-bound IGF-1 can therefore be considered a reservoir of IGF-1 that can be liberated in a spatiotemporal manner by IGFBP proteolysis depending on the presence and activity of various proteinases in different tissues (82) (Fig. 1).

Fig. 1.

Insulin-like growth factor-1 (IGF-1): from secretion to receptor activation. Blue arrows: growth hormone (GH) secreted from the pituitary gland (1) stimulates hepatic IGF-1 secretion into the circulation (2). In turn, IGF-1 reduces GH secretion by negative feedback (3). Yellow arrows: IGF-1 is bound by IGF-binding protein 3 (IGFBP-3) alone or in a ternary complex with IGFBP-3 and acid labile subunit (ALS) (4). Liberated IGF-1 (5) is able to cross the endothelium. Upon entry into the artery interstitium, IGF-1 is met by an IGFBP-4-dominated environment and is bound by IGFBP4 (6). Red arrows: IGF-1 is liberated from IGFBP4 by proteolytic activity of pregnancy-associated plasma protein-A (PAPP-A) (7). PAPP-A is associated with cell surfaces by electrostatic interaction with heparan sulfate glycosaminoglycans (HS-GAGs). IGF-1 is thereby liberated in close proximity to its receptors. Free IGF-1 and IGF-2 and insulin can bind and activate IGF receptor-1 (IGF-1R), insulin receptors (INSR; depending on isotype), and hybrids hereof, while IGF-2 can bind IGF-2R (8). Green arrows: GH can enter the artery wall and stimulate GH receptor (GHR)-expressing cells (e.g., vascular smooth muscle cells (VSMCs) to synthesize IGF-1 (9). IGF-1 produced by local vascular cells (10) is bound by IGFBP-4 and thereby enters the reservoir of PAPP-A-dependent releasable IGF-1 (11).

Pregnancy-associated plasma protein-A (PAPP-A) is a well-established IGFBP proteinase capable of leaving IGFBP-2, -4, and -5 (65, 67, 75). To date no other proteinases are known to target IGFBP-4 under physiological conditions (in contrast to IGFBP-2 and -5), thus PAPP-A is believed to be critical for liberation of IGFBP-4-bound IGF-1 (82). By its elongated zinc-binding motif, PAPP-A is classified as a metzincin metalloproteinase (15). However, unlike other members of this superfamily, e.g., several of the matrix metalloproteinases, which often cleave promiscuously, substrates other than IGFBPs have not been identified for PAPP-A (82). PAPP-A consists of multiple protein modules, some of which are well characterized while the function of others remains unknown. Besides the proteolytic domain, PAPP-A has three Lin12-Notch repeats (LNRs) critical for IGFBP-4 substrate specificity (14, 130) and five complement control protein (CCP) modules, also termed short consensus repeats, of which CCP3 and CCP4 are important for the ability of PAPP-A to bind to cell surfaces via heparan sulfate glycosaminoglycans (HS-GAGs) (66). The ability of PAPP-A to interact with HS-GAGs on cell surfaces enables efficient regulation of IGF-1 signaling (66). By liberating IGF-1 from IGFBP-4 in close proximity to IGF-1Rs, rebinding of IGF-1 to other high-affinity IGFBPs is minimized. In fact, although counterintuitive, Igfbp4^−/− mice are smaller because IGFBP-4 cooperates closely with PAPP-A to ensure efficient delivery of IGF-1 to its receptor (64, 77, 78).

By tight association with cellular surfaces, PAPP-A thus functions in tissues, close to the receptors of target cells. This is in contrast to PAPP-A2, the only paralog of PAPP-A, which has proteolytic activity toward IGFBP-3 and IGFBP-5 and is unable to associate with cellular surfaces (66). PAPP-A2 is therefore believed to function mainly in the circulation to ensure a steady level of free IGF-1 for tissues (3). Two families with PAPP-A2 deficiencies have recently been identified, directly supporting this interpretation (29). In humans, no families with inactivating PAPP-A mutations have yet been identified.

Little is known about the regulation of PAPP-A gene expression, but three endogenous PAPP-A inhibitors have been identified: The proform of eosinophil major basic protein, which inactivates the vast majority of circulating placenta-derived PAPP-A in pregnancy (80, 81), and stanniocalcin-1 and -2 (STC1 and -2), which appear to regulate PAPP-A within different tissues, including arteries (51, 52, 60, 112). STC1 inhibits by high-affinity (pM) binding to PAPP-A or PAPP-A2, while proform of eosinophil major basic protein or STC2 inhibition requires the formation of a covalent bond to PAPP-A or PAPP-A2.

Several lines of evidence underline the crucial role of PAPP-A in growth regulation and in the IGF system. First, genetic variants of PAPP-A strongly associate with height in genome-wide association studies (GWAS) such as the GIANT GWAS (62, 132). Second, Pappa^−/− mice are proportional dwarfs with a 40% reduction in birth weight (26). Interestingly, this effect is abolished in Pappa^−/−/Igfbp4^−/− mice (78) underlining that IGFBP-4 is the principal substrate of PAPP-A. Third, STC2 is one of seven genes including other principal components of the IGF system to have major impact on dog breed weight (89). Moreover, in a GWAS including >700,000 individuals, coding variants of STC2 resulting in compromised proteolytic inhibition of PAPP-A were reported to increase human height by up to 2.1 centimeters per allele (71). Collectively, these findings strongly indicate that PAPP-A and other key factors in the regulation IGF bioavailability are crucial in human physiology.

EXPRESSION OF IGF-1, IGF-1R, AND PAPP-A IN ATHEROSCLEROSIS

IGF-1 and IGF-1R are expressed by endothelial cells, VSMCs, and macrophages (32, 47, 76, 92). In early human atherosclerotic lesion, IGF-1 and IGF-1R are associated primarily with medial VSMCs (79), and expression of IGF-1 and IGF-1R is reduced in advanced lesions (11, 79, 84, 128), possibly as a consequence of VSMC apoptosis (32).

PAPP-A is expressed by medial and intimal VSMCs (73) in the healthy artery wall and throughout lesion development (112). In advanced plaques, PAPP-A colocalizes with endothelial cells, VSMCs, and macrophages, and plaques classified as vulnerable have elevated levels of PAPP-A (6). Indeed, factors present in the atherosclerotic milieu such as various proinflammatory cytokines (TNFα and IL-1β) stimulate VSMC PAPP-A expression (24). Cultivated VSMCs from various species produce abundant PAPP-A (7, 24, 112) and express approximately ten times more PAPP-A than endothelial cells (24). Interestingly, although macrophages stain positive for PAPP-A in atherosclerosis specimens, macrophages do not express PAPP-A, even upon cytokine stimulation (23).

In contrast to human atherosclerosis, little is known about expression patterns in mouse plaques.

CONFLICTING REPORTS?

Circulating PAPP-A was first identified as a biomarker for acute coronary syndrome in 2001 (6). Since then, a multitude of studies have reported diagnostic and prognostic values of PAPP-A in cardiovascular cohorts (53). The late realization that administration of the anticoagulant heparin artificially elevates circulating PAPP-A by detachment from tissue HS-GAGs has, however, seeded doubt in studies not reporting whether or not heparin was used as well of the timing of heparin administration relative to blood sampling (118). Nonetheless, these positive findings prompted hypotheses as to the involvement of PAPP-A in atherosclerosis. Papp-a-deficient Apoe^−/− mice have a marked reduction in lesion development compared with Apoe^−/− littermates in both constitutive and inducible knockout model systems (4, 43) (Table 1). Vice versa, SM22α promoter (altered to prevent decreased promoter activity during plaque formation (125)-driven transgenic overexpression of human PAPP-A by VSMCs but also bladder and uterus smooth muscle cells aggravated atherosclerosis (25). To assess which domains of PAPP-A the proatherogenic effect is attributed to, mice overexpressing mutants of human PAPP-A (driven by the altered SM22α-promoter) were generated (13). Disruption of either the proteolytic domain, CCP3 (responsible for HS-GAG binding) or LNR3 (responsible for IGFBP-4 specificity) reduced the atheropromoting effect, highlighting the critical involvement of these domains and hence a crucial function of the interplay between PAPP-A and IGFBP-4 in atherosclerosis. Finally, in more therapeutic-oriented studies, treating Apoe^−/− mice with either human STC2 (adeno-associated virus-mediated hepatic overexpression) or an antibody inhibiting IGFBP-4-specific proteolytic activity of PAPP-A reduced atherogenesis (21, 112). PAPP-A expression is increased in inflammatory environments (24) and atherosclerotic plaques (6, 112), perhaps through downregulation of miR-430–3p, which in turn inhibits PAPP-A translation (115).

Table 1.

Summary of seemingly conflicting reports

Model System (Apoe−/− Background Unless Stated Otherwise)	Effect on Atherosclerosis Burden	Plasma Cholesterol	Endogenous Igf-1	Other Reported Effects	Reference
Papp-a−/− (global knockout, constitutive)	Reduction	Unaffected	Similar in circulation in newborns (26)	Reduced body weight by ~40% (26)	43
Papp-a−/− (global knockout, induced 5 wk postdiet initiation)	Reduction	Unaffected	ND	None	4
Papp-a inhibition (AAV-mediated hepatic secretion of human STC2 into the circulation)	Reduction	Unaffected	ND	None	112
Papp-a inhibition (antibody targeting IGFBP-4 proteolysis)	Reduction	Unaffected	ND	None	21
Human PAPP-A overexpression from VSMCs (and other tissues, especially bladder and uterus) (altered SM22α promoter)	Increase	Unaffected	Unaffected	None	25
Mutated human PAPP-A overexpression from VSMCs (altered SM22α promoter)	Increase	ND	ND	None	13
Human IGF-1 treatment	Reduction	Unaffected	Reduced in circulation by ~40%	None	114
Long R3 IGF-1 treatment	Reduction	Unaffected	Unaffected	None	120
Rat Igf-1 overexpression from VSMCs (altered α-SMA promoter)	None (stabilizing)	Unaffected	Unaffected	Body weight unaffected Reduced aorta size	99
Hepatic Igf1−/− (induced at age 3 mo), Paigen diet (not Apoe−/−)	Increase (only females)	Increase (females)	Reduced in circulation by ~80% Artery mRNA: increased by ~200%	Reduced body weight by ~20%	117
Hepatic Jak2−/− (both Apoe−/− and Ldlr −/−) As above + Long R3 IGF-1 treatment or hepatic overexpression of rat Igf-1	Increase Reduction	Unaffected ND	Major reduction in circulation Reversing reducing effect	Hepatic steatosis: increased GH in circulation by ~400% Reversing effects above	104
Myeloid Igf1r−/− (lyz-cre)	Increase	Unaffected	Unaffected	None	45
VSMC and fibroblast Igf1r−/− (SM22α-cre)	Increase	ND	Reduced in circulation by ~25% Vascular mRNA unaffected	Reduced body weight by ~25% Reduced aorta size Increased circulating GH	113
C3H.6T+/+ (with Igf-1 genomic region transferred from CH3 background)	Increase	Unaffected	Reduced in circulation by ~20%	None	100
Igf2−/− (global knockout)	Reduction	Reduced by ~45%	ND	Reduced body weight by ~60% Reduced aorta size Increased insulin level by ~200%	138

IGF-1, insulin-like growth factor 1; AAV, adeno-associated virus; VSMC: vascular smooth muscle cell; α-SMA, α-smooth muscle actin; PAPP-A, pregnancy-associated plasma protein-A; STC2, stanniocalcin-2; GH, growth hormone.

The uniform conclusion of these studies is that by increasing IGF-1 bioavailability in the artery tissue, PAPP-A is an atheropromoting molecule.

However, this conclusion is in opposition to a number of studies focusing on IGF-1 itself (Table 1) (44). Treating Apoe^−/− mice with human recombinant IGF-1 (114) or the LR3 IGF-1 analog (120) had atheroprotective effects. Vice versa, Apoe^−/− mice harboring a genomic region from the C3H strain have a 20% reduction in circulating IGF-1 and increased atherosclerosis; however, it was stated that other genetic components of the C3H strain could contribute to this phenotype (100). Likewise, adult-onset hepatocyte-specific Igf-1 knockout mice with an 80% reduction in circulating Igf-1 showed more pronounced fatty streak formation in Paigen diet (a high-fat diet supplemented with cholate)-fed female mice (but not in males), although this effect may in part be attributed to increased plasma cholesterol and IL-6 levels, as stated by the authors (117). In a more complex setting, hepatocyte-specific Jak2 knockout mice on Apoe^−/− background, which develop profound hepatic steatosis and reduced circulating Igf-1, also showed aggravated atherosclerosis, which could be counteracted by either infusion of LR3 IGF-1 or by hepatic overexpression of rat Igf-1 (104). As emphasized by the authors of this study, a direct effect of IGF-1 on the vascular wall could not be concluded as hepatic steatosis was also attenuated by IGF-1 treatment. Transgenic rat Igf-1 expression driven by a α-smooth muscle actin (α-SMA) promoter fragment (SMP8) in Apoe^−/− mice was used to determine the effect of local VSMC and fibroblast produced IGF-1 in the artery wall (ex-vascular production was also detected) (99). Although there was no effect on plaque burden (in contrast to studies based on elevating Igf-1 in the circulation), features of plaque stability (content of SM22α- and α-SMA-positive cells and reduced necrotic core size) were enhanced upon Igf-1 overexpression. On the other hand, Igf-1r deletion in VSMCs and fibroblasts (SM22α-CRE driven) resulted in atherosclerosis aggravation (113), as did Igf-1r depletion (lyz2-CRE driven) in macrophages (45).

Taken together, the current literature appears to be conflicting. However, as discussed in the following sections, there are several potential explanations for this apparent dispute.

NOT JUST A MATTER OF UP AND DOWN

It has previously been emphasized that the effect of manipulating IGF-1 signaling depends on several parameters. Timing (40) and duration (121, 140) of IGF-1 signaling are essential for resulting biological responses and phenotypes in various physiological and pathophysiological settings (27). This notion includes atherosclerosis. While gene expression is regulated dynamically by various disease stimuli, transgene expression is constant (indeed, as mentioned above, efforts are made to alter promoters to ensure chronic transgene expression) and the level not readily controllable. Therefore, chronic overexpression of a transgene is not simply a spatiotemporal parallel increase in the expression of its endogenous counterpart.

This possibly explains paradoxical observations such as the absence of an effect on lesion burden when overexpressing rat Igf-1 by VSMCs (99) as opposed to the increase in atherosclerosis resulting from knocking out endogenous VSMC/fibroblast Igf-1r (113) or why both constitutive and inducible Pappa^−/− mice as well as transgenic VSMC PAPP-A overexpression all reduce neointima formation during mechanical vascular injury (5, 22, 88). On the other hand, in healthy aortas where endogenous gene expression is constant, observations are consistent: overexpression of IGF-1 leads to VSMC hyperplasia (127) and altered dimensions of the aorta including increased thickness of the medial layer (99), while Igf-1r deficiency of VSMCs reduces aorta dimensions (113). Accordingly, overexpression of IGFBP-4 by VSMCs results in VSMC hypoplasia (129), and when a PAPP-A-resistant variant of IGFBP-4 is overexpressed, hypoplasia is accentuated (139). Moreover, overexpression of PAPP-A increases arterial Igfbp-5 expression (25) [an Igf-1-responsive gene (33)] indicating enhanced Igf-1 activity. It is of note that the fact that healthy arteries display marked dimensional differences in these models is a problem, as they likely bias subsequent atherogenesis. Taken together, manipulating the IGF system is not simply a matter of “up and down.”

SOMATOTROPH AXIS FEEDBACK

An additional set of problems arise upon systemic manipulation of the IGF system. Manipulation of any component of any biological system is bound to set in motion counterregulatory effects. The IGF system is not an exception. The problem is evident from many of the studies addressed here. Artificial lowering or raising of circulating IGF-1 feeds back on GH release from the pituitary gland either directly or indirectly via reduced GH-releasing hormone from the hypothalamus. The result is a reciprocal production of endogenous Igf-1. For instance, hepatocyte-specific Igf1^−/− mice having an 80% reduction in circulating Igf-1 display an increase in circulating GH (138) and a twofold increase in arterial Igf-1 mRNA in females (117). Likewise, circulating GH is elevated in hepatocyte-specific Jak2 knockout mice on Apoe^−/− background and reduced upon LR3 IGF-1 treatment (104). Vice versa, administration of recombinant human IGF-1 resulted in a 40% reduction in circulating endogenous Igf-1 (114). These counterregulatory effects give rise to a number of questions. How does the experimental approach affect net IGF-1 signaling in the artery tissue examined? What is the direct effect of GH on cells involved in atherosclerosis [indeed, the relevant cells of atherosclerosis are GH responsive (70, 110, 119)]. Since GH not only regulates expression of IGF-1 but also other factors of the IGF system, including IGFBPs (74, 122), interpretation is further complicated. That said, cell-type specific gene manipulation would be expected to circumvent these issues; however, that does not appear to be the case, as VSMC-specific Igf-1r deficiency resulted in concomitant elevation of circulating GH and an ~25% reduction in circulating Igf-1 as well as reduced body and aorta size (113).

Although circulating Igf-1 was unaffected in Papp-a^−/− mice (26) or upon vascular overexpression of human PAPP-A (25) it is likely that unidentified counterregulatory effects occur in these model systems as well.

IGF-2/IGF-2R NEGLECT

The presumed reason for the scarcity of experimental data on IGF-2 in rodents and the neglect of IGF-2 in most of the studies summarized above is that rodent Igf-2 expression decreases dramatically after birth (in contrast to humans) (68) and that Igf-2 deficiency does not affect postnatal growth (31) and therefore may be considered insignificant in postnatal life. Although the direct effect on atherosclerosis by Igf-2 cannot be determined from the Igf2^−/−/Apoe^−/− mouse phenotype due to concomitant effects on plasma cholesterol (reduction in both atherosclerosis and total cholesterol) (138), Igf-2 mRNA is not detectable in normal aortas, but elevated in atherosclerotic aortas (138), and evidently the vasculature is highly reactive to Igf-2 as Igf-2 overexpression by VSMCs (driven by an altered α-SMA promoter) results in spontaneous focal neointima formation (138). Moreover, translating to the human situation, IGF-2 seems to be at play in postnatal life. IGF-2 prevails in the circulation throughout life (30), and recently an IGF-2 mutation was shown to cause both pre- and postnatal growth restriction (9). PAPP-A liberates both IGF-1 and IGF-2 from its binding protein substrates, and depending on the relative abundance of Igf-1, Igf-2, and Papp-a substrates in healthy and diseased mouse arteries, the resulting phenotype of manipulating Papp-a would be a combined effect of increased bioavailability of both Igfs. Likewise, manipulating the level of Igf-1 or Igf-1r cannot be done without affecting Igf-2 signaling through the Igf-1 receptor. Although IGF-1 and IGF-2 are similar and both bind to IGF-1R and INSR (depending on INSRA/INSRB isoform) (36, 108), IGF-2 is the only one binding the IGF-2 receptor (IGF-2R), which is structurally and functionally distinct from IGF-1R (61). Whether or not IGF-2R is relevant in the context of vascular biology and atherosclerosis is unknown; however, in vitro studies do point to important roles of Igf-2r in VSMC modulation and migration (41, 42).

In summary, the observed effects on atherosclerosis in the studies summarized above are net effects of altered Igf-1 and Igf-2 signaling, which complicates interpretation of these studies.

INSULIN SIGNALING

The IGFs and insulin are ligands of the homologous IGF-1R and INSR, and hybrid receptors hereof, and bind these with varying affinity (107, 108). Similar to disrupting the balance between IGF-1 and IGF-2, manipulating the bioavailability of IGFs (e.g., by overexpression or deletion of Igf-1, Igf-1r, or Papp-a) would affect the constellation and/or usage of receptors. For instance, reducing Igf-1r expression was shown to increase insulin sensitivity by increasing the relative number of insulin holoreceptors in the mouse aorta (1). Indeed, INSRs are present on all vascular as well as inflammatory cells relevant for atherosclerosis (54). Immunoprecipitation of Igf-1r and Insr indicated a 0:2:1 ratio of Igf-1r holoreceptors, hybrid receptors, and Insr holoreceptors, respectively, on macrophages (45). Deletion of Igf-1r would thereby result in exclusive expression of Insr holoreceptors on macrophages. Although insulin-dependent Akt phosphorylation was unaffected in Igf-1r-deficient macrophages, it is difficult to exclude an effect on insulin signaling, since other known (e.g., the MAPK/Erk pathway) and potentially unknown pathways exist (58). Taken together, manipulation of the Igf-1 system cannot be done in an isolated manner without affecting insulin signaling. The consequence of insulin signaling on atherosclerosis is controversial and may depend on dose, context, and pathway use (87), yet there is no doubt that manipulation of insulin signaling affects atherosclerosis (58). Furthermore, other cell surface proteins (e.g., αVβ3-integrin and integrin-associated protein) interacting with IGF-1R can modulate not only which intracellular pathways that are initiated but also the duration of signaling (20). In this way, the exact cellular response to a given dose of IGF depends on the protein composition of the individual cell. These mechanisms of IGF signaling modulation probably account for the various reported cellular responses to IGF, including proliferation, survival, differentiation, migration, and insulin-like effects on metabolism (57). Depriving cells of Igf-1r could thus affect systems other than the IGF system and insulin signaling. These effects will also contribute to the net phenotype observed in the studies summarized here. In extension, unforeseen tampering with insulin signaling is likely to affect blood glucose levels, especially when manipulating the IGF system on a whole organism level. Since hyperglycemia is a recognized driver of atherosclerosis, blood glucose should be reported in future studies.

IGF SYSTEM-INDEPENDENT FUNCTIONS OF PAPP-A

One immediate conclusion of the above-listed conflicting studies is that PAPP-A simply has atheropromoting effects independent of its established role in increasing IGF-1 bioavailability. Since the role of several PAPP-A domains has yet to be identified, this is indeed a possibility, and some studies may propound this notion. The most direct evidence for an IGF system-independent function of PAPP-A was demonstrated in zebrafish, as pappa knockdown impairing early zebrafish development could be rescued by overexpression of a proteolytically inactive pappa mutant (59). However, it seems that most of the atheropromoting effect of PAPP-A is dependent on IGFBP-4 proteolysis. As mentioned above, genetic disruption of the LNR3 domain, required for cleavage of IGFBP-4, resulted in a marked reduction of the atheropromoting effect of PAPP-A (13), and targeting the LNR3 domain with an antibody reduced atherogenesis (21). Indeed, since a common strategy of these two approaches is disruption/targeting of the LNR3 domain to selectively inhibit IGFBP-4 proteolysis, an alternative role of LNR3 could theoretically confound the conclusion. Alternatively, cleavage of IGFBP-4 could have other effects than liberating IGF-1. Indeed, IGF system-independent roles have been described for all three PAPP-A substrates (8, 48, 49, 94, 109, 116, 136). In the case of IGFBP-4, one example of this is the reported effect on cardiomyocyte growth and differentiation, which appear to be mediated via the Wnt/β-catenin pathway (131, 134, 142), but other IGF-independent effects have also been suggested (85, 102, 133).

Taken together, PAPP-A may have atheropromoting effects by means of functions that are not related to the IGF system; however, if this is indeed the case, they would have to overpower the apparent atheroprotective effect of increasing IGF-1 bioavailability (45, 99, 100, 104, 113, 114, 117, 120). Moreover, they would have to be specific for atherosclerosis.

GENETIC EVIDENCE

A few hypothesis-driven studies have indicated atheroprotecting effects of IGF-1. In hypertensive but not normotensive individuals, a 192-basepair deletion within the IGF-1 promoter is associated with a 20% decrease in circulating IGF-1 and a 4% increase in carotid intima-media thickness (97), and in a different study, the C-allele of rs35767 was associated with a 7% increase in circulating IGF-1 and a 6.5% decrease in carotid intima-media thickness (98).

Two single nucleotide polymorphisms of PAPPA have also been reported to be associated with atherosclerosis, although the direction of effect is less clear. The intron-located rs13290387 was associated with acute myocardial (83), while the C allele rs7020782 was reported to be associated with ischemic cerebrovascular (126) and carotid plaque development (141). rs7020782 is located in exon 14 and the A to C substitution causes a missense mutation (tyrosine to serine) in the CCP1 module of PAPP-A. CCP3 and CCP4 have been demonstrated to be important in HS-GAG interaction and cell-surface binding (66); however, any specific function of CCP1 remains to be demonstrated. Of note, rs7020782 was previously associated with recurrent miscarriage during pregnancy (118) further highlighting a potential functional importance of the CCP1 module.

Curiously, however, despite that most genes encoding components of the IGF system contain effectful variants evident from strong associations with, e.g., height (17), there is no evidence for associations between these genes and measurable proxies for atherosclerotic disease such as coronary artery disease or myocardial infarction in current leading GWAS (17, 18, 28). The reason for this could be that the observed mouse phenotypes suffer from what has previously been referred to as “the knock out conundrum of experimental atherosclerosis” (69); namely, that the quick pace of experimental atherosclerosis (weeks/months) limits time for compensatory mechanisms to take over as they would in the human decades-long disease development and in this way not be as crucial as suggested by a mouse phenotype. Alternatively, genes encoding components of the IGF system may eventually be added to the ever-growing list of susceptibility loci with increasing power of GWAS. The above-mentioned hypothesis-driven genetic studies indeed propound this possibility. Finally, current large-scale GWAS are based on single nucleotide polymorphisms masking other types of variation such as the 192-basepair deletion within the IGF-1 promoter (97).

ASSOCIATING CIRCULATING PAPP-A AND IGF-1 WITH MEASURES OF ATHEROSCLEROSIS

Human studies associating either circulating IGF-1 or PAPP-A with measures of atherosclerosis suffer from methodological issues in part due to biological complexity. Recognizing that heparin administration causes artificial elevations in circulating PAPP-A (118) (presumably by outcompeting its binding to tissue heparan sulfate) complicates interpretation of studies not reporting on the use of this anticoagulant. Moreover, since a yet to be assessed proportion of circulating PAPP-A may be in complex with STC1 or STC2 in nonpregnant individuals, it may be that a variable fraction of quantified circulating PAPP-A is inactive. Also, total circulating PAPP-A could be underestimated if epitopes are masked by these inhibitors.

IGF-1 has for long been known not to circulate freely. Around 99% of IGF-1 in the circulation is in a complex with IGFBP-3 and acid labile subunit (ALS) (12, 101). Efforts are made to measure the free or “bioavailable” portion of IGF-1 as this is considered more physiologically relevant than the reservoir of latent IGF-1 present in complex with an IGFBP. It remains elusive what mechanism governs the transfer of circulating IGF-1 (free or bound) to its receptor on the cell surface of target cell in the vascular wall; however, it likely involves transfer of IGF-1 from the IGFBP-3/ALS complex to IGFBP types dominating the vascular tissue, e.g., IGFBP-4 (2, 39). In support of this idea, proteins regulating IGF signaling locally (i.e., PAPP-A and its inhibitors) appear to be more critical than the concentration of circulating IGF-1, since knockout or transgenic overexpression of genes encoding these has profound effects on physiology without affecting circulating Igf-1 (19, 26, 38, 124). Taken together, although highly convenient in terms of obtaining a blood sample, the biological and methodological complexities complicate data interpretation. Consequently, reports linking IGF-1, circulating binding proteins, or PAPP-A to measures of atherosclerosis point in all directions (16, 35, 37, 50, 53, 55, 56, 63, 93, 96, 103, 111, 123).

The next problem is the interpretation of an elevated level of circulating IGF-1 or PAPP-A. A positive correlation between a circulating molecule and a disease measure is usually interpreted as the molecule being a disease-promoting role. However, it may also be that the tissue expresses the molecule as a way to counteract disease progression.

Circulating PAPP-A is low in healthy individuals, and circulating elevations in disease states are probably an artifact with no physiological purpose. This notion is substantiated as PAPP-A must be bound to the cell surface in close proximity to IGF receptors to enable local liberation of IGF to avoid that IGFs are rebound to new IGFBPs (66). Circulating levels of IGFBP-4 fragments may circumvent the above-mentioned complications and be a better measure of tissue PAPP-A activity (46, 86).

CONCLUSION AND FUTURE DIRECTIONS

Therefore, what’s up and what’s down? As discussed in this review, there are many potential ways to explain the seemingly conflicting reports all of which point to the complexity of the IGF system and the problems of manipulating it. Experimental approaches addressing whether or not the proatherosclerotic effects of PAPP-A are in fact mediated by IGF-1 or whether IGF system-independent effects are involved should be a high priority. This would require, e.g., complex model systems enabling assessment of consequences of PAPP-A overexpression in an Igf-1r-deficient system. Moreover, future studies should strive to use inducible cell type-specific genetic manipulation (e.g., tamoxifen- or doxycycline-inducible cre-recombinase activation systems) to circumvent many of the problems highlighted above.

At this point, the literature does not provide a clear message, but we hope that the issues discussed here will be considered when navigating in the literature of this field and when designing future experiments aiming at dissecting the role of PAPP-A and the IGF system in atherosclerosis with the overall aim of identifying novel treatment regimens for this disease.

GRANTS

This work was supported by the Centre for Individualized Medicine in Arterial Diseases, Odense University Hospital (to L. B. Steffensen); National Heart, Lung, and Blood Institute Grant R01-HL-074871 (to C. A. Conover); and Independent Research Fund Denmark (to C. Oxvig).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

L.B.S. prepared figures; L.B.S. drafted manuscript; L.B.S., C.A.C., and C.O. edited and revised manuscript; L.B.S., C.A.C., and C.O. approved final version of manuscript.