**4. GRP94**

GRP94 is a glucose-regulated protein of 94 kDa molecular size, encoded by the gene *HSP90B1* (*OMIM #*191175), whose chromosomal location is remarkably close to the IGF-I gene. Its expression is ubiquitous and its transcription is upregulated by low glucose tension [39], among other conditions. GRP94 also is commonly known as gp96, ERp99, or endoplasmin [40], referring to its extensive glycosylation and its abundance in the endoplasmic reticulum. It has the domain structure typical of the heat shock 90 (HSP90) family of proteins, including a C-terminal domain that mediates the constitutive dimerization of GRP94 (Figure 1). Like all family members, the N-terminal domain of GRP94 is a typical ATP-binding domain [41] that affects the dimerization of GRP94 and its action cycle [42]. The N-terminal domain also mediates binding of antigenic peptides [43] through which GRP94 activates T cells, the basis for the immunological function of GRP94 [43,44]. The protein chaperone function of GRP94 also requires the N-terminal ATPase domain [45], but the protein binding site is thought to reside in the C-terminal domain, around residues 652–678 [46].

GRP94 is an essential chaperone for multiple receptors and secreted proteins [47] (Table 1). Protein interaction data using GRP94-sufficient and -deficient cells show physical interactions with ~200 proteins and effects on expression levels of ~500 proteins [48], including some of the verified substrates listed in Table 1. Much of this interactome remains to be characterized. For some of the substrates (e.g., IGFs) there is genetic evidence that GRP94 is essential, whereas others can be expressed properly (albeit at lower abundance) even absent GRP94. For many of the susbstrates (e.g., thyroglobulin), data only show physical association without a physiological conseqence.

As can be gleaned from this non-exhaustive listing, GRP94 substrates (also called "clients") are found in a variety of tissues and cell types. These substrates share no common structural motif that would predict their association with GRP94, nor do they share a common protein fold or a characteristic post-translational modification, aside from internal disulfide bonds (Table 1). The only obvious common denominator is that the substrates are secreted or membrane-bound proteins that are made in the endoplasmic reticulum. Importantly, even in cases of verified GRP94-substrate interactions, there can be exceptional isoforms or family members that do not interact, for example, TLR3 vs. most other TLRs [49].


**Table 1.** Protein substrates of GRP94.

No GRP94–susbstrate complex has been purified and analysed so far, so the exact mode of interaction currently can only be simulated, as shown in Figure 1 with human IGF-I [62] and GRP94 [63], using the ZDOCK algorithm [64]. Furthermore, as GRP94, like all chaperones, binds substrates that have not ye<sup>t</sup> reached their final three-dimensional structure, the precise interaction is only approximated based on known mutations in the interacting proteins. This is an implicit limitation of docking studies such as that shown in Figure 1.

**Figure 1.** A predicted complex between GRP94 and IGF-I. The crystal structure of human IGF-1 (1IMX [63]) was docked onto the crystal structure of GRP94 (2O1U [64]) with the ZDOCK algorithm (version 3.0.2 [65]). The two monomers of GRP94, are shown in cyan and green, with the N-terminal and C-terminal domains indicated. The interacting amino acids are colored yellow. The complex shown is the highest scoring predicted complex, and eight other complexes out of the 10 highest scoring ones overlap with it, predicting the same topology of binding. The GRP94 interacting residues are from from the internal face of the Middle domain and the C-terminal domain of the chaperone. The IGF-I interacting residues are mostly derived from its N-terminal 32 amino acids, colored light blue, while the C-terminal 28 residues (deep blue) are mostly predicted as non-interacting.

Along with molecular specificity, another enigmatic feature of GRP94 is the paucity of co-chaperones. Its cytosolic homolog, HSP90, has a well-known set of auxiliary proteins that form transient complexes and impact the quality and/or speed of enhanced folding of the respective substrates [65]. Even other types of ER chaperones have well known co-chaperones, some general for all substrates and some substrate-dedicated [66]. In contrast, GRP94 is currently known to work with only one co-chaperone, CNPY3 [49] (see below). As far as the insulin/IGF substrates, GRP94 co-chaperones are presumably ye<sup>t</sup> to be characterized, because genetic data show that ASNA-1 is an evolutionarily conserved ATPase that is important for insulin/IGF maturation in both worms and mammals [67] (see the next section). Further characterization of co-chaperones will no doubt explain many of the unresolved details about the action cycle of the GRP94 chaperone machine.

GRP94 differs from the cytosolic HSP90 orthologs in inherent, functionally-relevant structural properties [42] such as the nucleotide-dependent conformational changes of the N-terminal domain [63] as well as the interactions mediated by the charged linker domain. These differences lead to a different action cycle of this protein [68,69] and probably also to its ability to chaperone folding of client proteins without the many co-chaperones that are required for activity of the cytosolic HSP90 orthologs [70].

#### **5. GRP94 as an Obligate Chaperone for IGF-I and IGF-II Production**

The dependence of IGF-I maturation and secretion on GRP94 is a property also exhibited by IGF-II [38,71] and insulin [58] (Table 1), and even by the insulin-like proteins of the nematode *C. elegans*, some of which have only weak primary sequence similarity to IGF-I [72], showing that it is evolutionarily conserved. In contrast, within the TLR family of substrates, TLR3 is exceptional in its refractiveness to GRP94, showing the selectivity of substrate selection by GRP94 [49]. The chaperone dependence of the IGFs is based on physical association of pro-IGFs (or pro-insulin) with GRP94, an association that is transient and occurs early during biosynthesis [38,73]. The precise amino acids of the pro-insulins that interact with GRP94 have not been mapped, but some experiments plus molecular modelling indicate that the pro-insulins do not bind at the site of GRP94 that is responsible for binding of antigenic peptides [43], but rather bind at a more distal site encompassing the middle and C-terminal domains of GRP94 [58]. Apparently, their binding site overlaps residues 652–678 [58], the region that was identified for binding integrins and TLRs [46]. Nonetheless, despite such overlap, there is more complex specificity built into client selection, for example, the pair Met658/Met662 residues are essential for integrin folding but not TLRs [46].

Folding of client proteins often involves not just a chaperone protein, but also recruitment of additional proteins dedicated for the client, which serve as co-chaperones. In the case of GRP94, the ER luminal protein CNPY3 binds to GRP94 when it is engaged in biosynthesis of Toll-like receptors, but not other clients. CNPY3 and GRP94 interact with each other and with the TLR client in nucleotide-dependent manner [49]. Similar complexes have not been defined for the IGF/insulin proteins, but they likely exist; ASNA1, for example, is an ATPase expressed in insulin/IGF-producing cells in both worms and humans which regulates insulin secretion [67].

The biological importance of the IGF–GRP94 interaction is highlighted by the discovery of a hypomorphic variant of human GRP94, P300L, that a ffects the IGF chaperone activity and limits IGF biosynthesis [74]. Only four homozygotes have been identified so far, a lower frequency than expected from genetic principles [74], and heterozygous carriers of P300L are a noncommon single nucleotide polymorphism with frequencies of 1–4% in various populations. Carriers have 9% lower circulating IGF-1 concentration. In cell models of P300L heterozygosity, half as much IGF was secreted relative to wild type GRP94 [74]. It should be noted that the marked dependence of IGFs on GRP94 activity is unusual—depletion of the chaperone has much milder e ffects on the expression of some GRP94 client proteins compared to the secretion of the insulin family clients [52].

Why does the insulin/IGF structure require GRP94? At present, this question is not properly answered, and the available data only provide hints. The insulin-like family of proteins is unusual in that they are made initially as small (less than 100 amino acids) pro-proteins, that are processed proteolytically [72,75]. Furthermore, most of these sequences encode for three disulfides [72] that need to be bonded in a precise order within a small molecular space, a considerable folding challenge [76,77]. The surprising finding that at least one IGF-I variant has alternative folded states [78] underscores the folding di fficulty, which is one likely reason for the need for molecular chaperones. As GRP94 has been found to interact with PDIs [60,79] it may act as a sca ffolding protein in the recruitment of PDIs during the folding of the substrates [80].

The essential chaperoning role of GRP94 towards IGFs has implications for cell growth, for normal tissue di fferentiation and for cancer progression. A common cellular stress situation is the withdrawal of growth factors from cells, many of which respond to such stress by autocrine production of the growth factors [81]. However, cells with mutated or drug-inhibited GRP94 cannot produce these growth factors [38], leading to arrested growth/di fferentiation and, in extreme cases, cell death. The requirement for functional GRP94 in development is illustrated by the dramatic impact of tissue-specific GRP94 depletion on striated muscle [57], where myotube fusion and expression of contractile proteins downstream of the master MyoD transcription program are inhibited, coincident with the known need for synergistic input from growth factor signaling [73]. In cancer, elevated expression of GRP94 is observed in melanoma, ovarian cancer, multiple myeloma, lung cancer, and inflammation-associated colon cancer. GRP94 expression in cancer cells is closely linked to cancer growth and metastasis through a number of its clients, as listed above [82]. In part, this is due to response of the GRP94 promoter to some aspects of the tumor microenvironment that may include low glucose level [83], but is distinct from hypoxia [84]. The increase in GRP94 expression in tumors is tightly linked to their increased cellular proliferation rate and migration capacities and to their increased production of growth factors [85].

Constitutive overexpression of GRP94 is a common survival pathway that is usually used during oxidative stress [86], reflecting the many pathways that involve GRP94. The above three examples highlight situations that upregulate GRP94 more specifically, because of subsets of interacting proteins.

Association with GRP94 is by no means the only protein–protein interaction that IGF-I undergoes. Circulating IGF-I is secreted mainly by the liver and circulates bound to IGF-binding proteins (IGFBPs), either as binary complexes or ternary complexes primarily with IGFBP-3 or IGFBP-5 and an acid-labile subunit (ALS). The components of these circulating complexes are produced by di fferent cells and the complexes assemble after secretion to the circulation [87]. The complexes are important for the stability of circulating IGFs and also for their signaling function; in the absence of IGFBPs, there is much lower level of serum IGF-1, but surprisingly, this neither predicted growth potential or skeletal integrity nor defined GH secretion or metabolic abnormalities [88].

Each IGF-I associated protein appears to play a distinct role in determining musculoskeletal phenotype, with di fferent e ffects on cortical and trabecular bone compartments and the striated muscles [88,89]. The di fferential e ffects of hepatic vs. autocrine/paracrine IGF-I is likewise attributable to di fferent complexes, either due to di fferential assembly or to di fferent proteases at the target tissue that cleave the IGFBP to release IGF-I to interact with IGF1R, IGF1R and insulin receptor density, etc. [90]. Similarly, when skeletal muscle deletion of GRP94 is used to limit production of IGF-I, endocrine and paracrine IGF-I are shown to regulate both tissue growth and body plan [57,88].
