The growth factor progranulin (PGRN) regulates cell division, survival, and migration. PGRN is an extracellular glycoprotein bearing multiple copies of the cysteine-rich granulin motif. With PGRN family members in plants and slime mold, it represents one of the most ancient of the extracellular regulatory proteins still extant in modern animals. PRGN has multiple biological roles. It contributes to the regulation of early embryogenesis, to adult tissue repair and inflammation. Elevated PGRN levels often occur in cancers, and PGRN immunotherapy inhibits the growth of hepatic cancer xenografts in mice. Recent studies have demonstrated roles for PGRN in neurobiology. An autosomal dominant mutation in GRN, the gene for PGRN, leads to neuronal atrophy in the frontal and temporal lobes, resulting in the disease frontotemporal lobar dementia. In this review we will discuss current knowledge of the multifaceted biology of PGRN.
Our laboratories began to investigate the function of the granulins (GRNs) while looking for peptides of the innate host response. While purifying members of the defensin family of anti-microbial peptides from extracts of human leukocytes, using a protocol we devised to screen for fractions containing cysteine-rich peptides, the GRNs were identified in minor side fractions.1 Members of the GRN family were also shown to be significant components in extracts of carp myeloid tissue.2 Sequence analysis of human and carp GRNs revealed a unique 12-cysteine motif consisting of four pairs of cysteines flanked by two single cysteines at the amino and carboxyl termini (Fig. 1A). All cysteines were found to be fully crossed-linked to form six disulfide bridges. Two-dimensional nuclear magnetic resonance analysis of purified carp GRN-1 revealed that the peptide backbone adopts a unique conformation of a parallel stack of beta-hairpins in the form of a left-handed helix.3 Recombinant mammalian GRN modules show a similar, but less rigid structure4 (Fig. 1B). Subsequent work showed that the mammalian GRN peptides were fragments of a larger protein, progranulin (PGRN), bearing seven and one-half GRN repeats (Fig. 1A). While both PGRN and its constituent GRN peptides have biological activity, most research has focused on the function of the larger PGRN protein.
Initially the function of the GRNs was obscure, and our continued interest in them was based more on the concept that their unusual chemical structures strongly suggested that they would prove biologically interesting, than on a clear idea of function. Slowly work from a number of groups including our own began to elucidate tentative roles and about 10 years ago we wrote a review called, optimistically, “Granulins: the structure and function of an emerging family of growth factors.”5 In the intervening years, members of the GRN gene family (GRN) have indeed emerged as significant players in the extracellular regulation of cell function, although often in surprising contexts that we had not predicted 10 years ago. We know now, for example, that the human family member PGRN is critical in maintaining neuronal survival, since mutations of the GRN gene lead to cell death in the frontal and temporal lobes of the brain.6, 7 The role of PGRN in cancer has been repeatedly demonstrated.8–26 PGRN has been invoked in early embryogenesis,27 wound repair, and inflammation.28–30 These are diverse roles, extending from control of embryonic development during the first days of life to the survival of long-lived post-mitotic neurons of the adult brain. In this article we wish to identify unifying themes in the biology of PGRN and other members of the GRN family that would help rationalize this complexity.
Evolutionary origins and structure
The unique nature of the 12-cysteine GRN motif makes unambiguous identification of homologous structures possible. A GRN motif is found at the carboxyl terminus of the cathepsin K family of cysteine proteases found in numerous members of the plant kingdom whose expression is up-regulated by environmental stressors31 (Fig. 1A). A form of GRN is found in the slime mold Dictyostelium discoideum,32 a social amoeba that is a modern representative of a primordial organism that is thought to have predated the divergence of the plant and animal kingdoms. Fungi do not have a GRN gene. In fish, and many invertebrate organisms, multiple GRN genes are found, whereas mammals possess only one known member of the GRN gene family. The zebrafish genome, for example, harbors four GRN genes.33zPGRN-A and zPGRN-B are co-orthologs of the human gene bearing multiple copies of the GRN motif (Fig. 1A). Two smaller forms (zPGRN-1 and zPGRN-2) are also present (Fig. 1A). Syntenic conservation of gene location shows that zPGRN-A is the ortholog of mammalian GRN. Why only one family member was retained in mammals and other land vertebrates is not known.
The phylogenetic distribution of the GRN motif suggests that it evolved only once about 1.5 billion years ago. Has it always functioned as a signaling factor? The appearance of complex life forms was concomitant with the evolution of a multiplicity of cell signaling and cell-cell adhesion proteins responsible for the complexity and diversity of multicellular organisms. The sponge Oscarella carmela has a simple branching body plan but surprisingly has nearly all the classical growth factor signaling mechanisms including the Wnt, Hedgehog, and TGF-beta pathways.34 In contrast, Monosiga brevicollis, an example of the Choanoflagellates, which are the closest known unicellular relatives of metazoans, lacks all these critical pathways.35 However, GRN genes are found in O. carmela (EC372216), M. brevicollis (XP_001748993), and D. discoideum (XP_638956), suggesting that the GRN signaling system evolved before most other contemporary growth factor pathways.
The mammalian GRN gene encodes a multifunctional secreted glycoprotein with tandem repeats of cysteine-rich GRN modules1, 36–39 (Fig. 1A). It is known synonymously as PGRN,8 granulin-epithelin precursor (GEP),40 proepithelin,38 PC cell-derived growth factor (PCDGF),10 acrogranin,39 and epithelial transforming growth factor (TGFe).41 This rather dense nomenclature for a single gene is revealing since it captures many of the essential features of the biology of PGRN. The GRN nomenclature emphasizes its association with granulocytes and the cells of the innate immune system, while the epithelin nomenclature emphasizes its association with epithelial cells. The PCDGF and TGFe designations emphasize the functional aspects of the protein as a growth modulator, while acrogranin (from acrosome, a compartment of the sperm head) brings out the likely roles of PGRN in reproduction and early development.
Tissue remodeling and development
PGRN is often expressed under conditions of tissue remodeling where cells are dividing and actively migrating. For adult epithelia it is abundant in regions that are rapidly turning-over, notably in the intestinal deep crypt and epidermal keratinocytes.42 Other less mitotically active epithelia usually express PGRN at far lower levels. Fibroblasts and endothelial cells, which are normally mitotically quiescent, show corresponding low levels of PGRN.43 However, these cells can rapidly deploy very active tissue remodeling programs of increased proliferation and migration following wounding, for example. As they do so, their expression of PGRN increases dramatically.28 The increased expression of PGRN in wounds is likely to contribute to the repair process,28 since adding PGRN to skin wounds in mice increases the number of fibroblasts and capillaries that enter the wounds in the early stages of healing.28 In tissue culture PGRN stimulated the proliferation and migration through collagen of dermal fibroblasts and endothelial cells, recapitulating the effects that were observed in the intact wounds.28 The actions of PGRN in injury extend to regulating inflammation, since PGRN is a potent inhibitor of the inflammatory cytokine tumor necrosis factor-α (TNF-α).29, 30 Tumors exhibit pathologically disordered tissue remodeling and PGRN expression is often highly elevated in cancers of many types including carcinomas,11, 22, 44–47 gliomas,26 multiple myelomas,13 and uterine smooth muscle sarcomas.12 In most cases there is a relationship between the cancer progression and the expression of PGRN, the higher-grade tumors being more likely to express elevated PGRN.
PGRN is intimately involved in early embryogenesis and, importantly, shows specificity of both expression and effect. In the blastocyst (the fluid-filled form of the embryo prior to implantation into the uterus), PGRN immunoreactivity is located in the trophoblast,48 the outermost shell of cells around the inner cell mass that expand after implantation to create the fetal compartment of the placenta. PGRN becomes detectable in the inner cell mass only after the conceptus has attached to the uterus.49 This is consistent with biological studies in which PGRN was found to stimulate cavitation, the process whereby the solid ball of cells of the morula becomes the fluid-filled blastocyst, and shown to have growth-promoting activity on trophoblasts, but not the inner cell mass.48, 49 In addition, PGRN stimulates the hatching, adhesion, and outgrowth of the blastocyst in experimental models of implantation.48 PGRN continues to be expressed in the placenta after implantation50 and in the embryo, particularly in the epidermis and developing nervous system,51 which may be significant given the recent discovery of the role of PGRN in neurodegenerative diseases.
Progranulin and neurodegenerative diseases
Although the relationship between PGRN expression and tissue remodeling is compelling, there are cases where this evidently does not apply, perhaps the most obvious being in the post-mitotic cells of the brain and spinal cord, many of which express PGRN very strongly,42, 51 but are neither proliferating nor migrating. Recent evidence, mostly from the genetics of neurological disease, reveals that PGRN protects neurons from premature death. Mutation of a single copy of the human GRN gene results in neuronal atrophy of the frontal and anterior temporal lobes (frontotemporal lobar degeneration, FTLD)6, 7 (Fig. 2A).
Clinically, this manifests as a disease called frontotemporal dementia (FTD). FTD is often a familial disease, and typically appears at a relatively young age, being the second most common dementia for people under 60 years. The frontal and temporal lobes have key roles in regulating behavior, empathy, and social understanding as well as language and this is reflected in the initial clinical presentation of the disease. Generally speaking, three variants of FTD are recognized,52, 53 a behavioral variant (bvFTD), and two forms that are associated with language problems, namely progressive non-fluent aphasia, which is characterized by loss in the ability to produce speech, with the individuals eventually becoming effectively mute, and semantic dementia, in which patients retain the ability to speak but in an increasingly disorganized and meaningless manner as the disease worsens. GRN mutations are often associated with bvFTD, but many instances of language-deficit FTD linked to GRN mutations have been reported.54 FTD may be accompanied by motor neuron disease, although this is rare in individuals with mutant GRN.
FTD is genetically heterogeneous. Other mutations in addition to GRN cause FTD, most notably the microtubule-associated protein tau (MAPT),55 and, more rarely, chromatin-modifying protein 2B (CHMP2B)56 and the valosin-containing protein (VCP).57MAPT was the first FTD gene to be identified, and maps in remarkably close proximity to GRN on chromosome 17q21.3.6, 7 Presently the Alzheimer Disease and Frontotemporal Dementia Mutation Database (http://www.molgen.ua.ac.be/FTDMutations/) records 66 distinct GRN mutations.58 Most, if not all the GRN-dependent FTDs result from a decrease in the amount of PGRN expressed or secreted, rather than an acquired toxic effect of mutant protein. Many of these mutations lead to nonsense-mediated mRNA decay, a process that eliminates the mutant transcripts and therefore lowers the expression level of PGRN mRNA by 50%.59 Other mutants may affect the translated protein, for example, by changing one of the cysteines to another residue and therefore disturbing disulfide bridge formation, or by impeding secretion through mutations of the signal peptide.60, 61 Insufficient production of PGRN protein as the underlying cause of disease was confirmed with the identification of FTD associated with allelic loss of the entire GRN locus.62
Although neuronal death is the hallmark of FTLD, the GRN- and MAPT-dependent phenotypes are strikingly different at the cellular level. In the GRN-linked condition, the neurons accumulate cytoplasmic and nuclear inclusions that stain strongly for ubiquitin and phosphorylated fragments of a protein called TAR DNA-binding protein 43 (TDP-43).63, 64 Ubiquitin inclusions are rarely found when FTLD is due to mutant MAPT, instead the affected neurons display aggregations of the tau protein, identifying this form of FTLD as a tauopathy along with other dementias such as Alzheimer's disease.65 Thus far, all GRN-linked forms of FTLD examined exhibit ubiquitin inclusions. However, inclusions occur in cases of FTLD that do not arise from mutant GRN. Moreover, the ubiquitin/TDP-43 inclusions occur in other neurodegenerative diseases, such as amyotrophic lateral sclerosis (ALS), where there is no strong link with GRN mutations.66
Given the complex histopathological relationship between GRN mutations and ubiquitin/TDP-43 inclusions, are there any functional relationships between PGRN and TDP-43? The ubiquitin/TDP-43 inclusions do not contain PGRN,67 and therefore the direct seeding of ubiquitin inclusions by PGRN is highly unlikely. When PGRN mRNA levels were depleted in cell culture, TDP-43 underwent a partial caspase-dependent proteolysis.68 The fragmentation of TDP-43 appears to be an initial step in the formation of ubiquitin/TDP-43 inclusions, although it should be pointed out that other investigators found no alteration in TDP-43 localization or stability following depletion of PGRN mRNA.61 The 25 kDa carboxyl-terminal fragment of TDP-43 accumulates in neurodegenerative tissue, and, when over-expressed in cells, it was phosphorylated, ubiquitinated, and cytotoxic.69, 70 Among its several functions, TDP-43 is a specific mRNA-binding protein for human neurofilament mRNA,71 and may be involved in the response to neuronal injury since its levels increased and it was translocated from the nucleus to the cytoplasm in injured motor neurons.72 As TDP-43 increased in the injured neurons, the cytoplasmic levels of PGRN dropped in parallel.72 Assuming that neuronal PGRN levels also decrease after injury in the brain, the depletion of the remaining 50% normal PGRN in GRN-mutant carriers following cerebral stress or trauma may on occasion bring neuronal PGRN to very low levels compared to non-carriers.
The functional roles of PGRN in neurons and neurodegeneration are only beginning to be explored. PGRN is neurotrophic for cortical and spinal cord neurons,73 which may be relevant in the loss of cortical neurons in FTD; however, other factors are likely to be at work. Not everyone with a GRN mutant develops FTD, and some carriers of pathological variants of GRN have lived well into their 70s and beyond with no apparent loss of cognitive function.59 This would suggest that there are important biological modifiers of GRN pathological outcome, but little if anything is known of what these may be.
In other cell types, PGRN is a potent extracellular anti-apoptotic factor16, 21, 74 and if this is true also for neurons, the loss of half the normal level of PGRN may sensitize the affected neurons to a range of traumatic shocks. Similarly, PGRN has roles in peripheral inflammation.29, 30 The equivalent neuroinflammatory cells in the brain, the microglia, express PGRN very strongly in diseased tissue, and an inflammatory contribution from PGRN to the disease has been hypothesized.75 Although PGRN is widely distributed, the pathology due to GRN mutations is highly restricted. PGRN is expressed in many neurons outside the cerebral cortex,42 but these appear to be far less affected by GRN mutations. Why the cells of the frontal and temporal cortex are so much more sensitive to loss of PGRN than others is not known. The lack of a direct causative influence in other neurodegenerative diseases does not exclude more subtle roles. For instance, genetic variants of GRN may be disease modifiers in ALS,76 although other investigators have not been able to reproduce this correlation.66
PGRN has other functions in the brain. The male hypothalamus undergoes a default female developmental program until late in embryonic development when it is masculinized by the actions of circulating androgens. GRN mRNA was identified as one of the transcripts that were most strongly up-regulated by androgens in neonatal mouse hypothalami and in a series of elegant experiments by Suzuki et al.77 it was shown that depletion of PGRN blunts many of the normal masculine reproductive behaviors, particularly ejaculation.78, 79 Abrogation of the GRN gene in male mice showed unusual levels of anxiety, which were attributed at least in part to decreased expression of the serotonin receptor 5HT1A.79 These findings coupled with the expression of PGRN in many regions of the embryonic mouse brain51 hint at yet further roles for PGRN in brain development. Intriguingly, comparative studies also support a key neuronal role for PGRN, since a PGRN-like protein is evident in the nerve cells of an annelid ragworm, whose last common ancestor with vertebrates presumably dates very early in animal evolution.80
Progranulin as a somatic growth factor
PGRN is a double-edged sword. Of equal importance to the problems that accrue when PGRN levels are low is what happens to cells with increased exposure to PGRN. Elevated PGRN stimulates proliferation,8–10, 15, 16, 28, 40 survival,9, 21, 74 and motility9, 24, 28 of epithelia, fibroblasts, and endothelia. PGRN activates typical growth factor signal transduction pathways such as the phosphorylation of shc and p44/42 mitogen-activated protein kinase (p44/42MAPK) in the extracellular regulated kinase (ERK) pathway as well as phosphatidylinositol 3-kinase (PI3K), protein kinase B/AKT, and the p70S6 kinase in the PI3K pathway.9, 15, 16 PGRN promotes tyrosine phosphorylation of focal adhesion kinase (FAK),9 which is a cytoplasmic tyrosine kinase in the signaling pathways associated with clustered integrins.81 FAK provides a link between integrin and growth factor signaling since it is required for epidermal growth factor (EGF) and platelet-derived growth factor (PDGF) to promote cell motility.82 In bladder cancer cells where PGRN is a strong motogen but a poor mitogen, PGRN promoted the association of FAK with paxillin and p44/42MAPK,24 suggesting, at the very least, a cellular link between PGRN and extracellular matrix (ECM) signaling machinery.
To appreciate how the growth factor-like properties of PGRN differ from those of conventional growth factors, it is necessary to revisit some of the fundamentals of growth factor biology. In serum-free medium, normal fibroblasts require two distinct growth factor signals to progress through the complete cell cycle.83 One growth factor, the competence factor, prepares the cell to pass into the S-phase where DNA synthesis occurs. The second growth factor, the progression factor, then drives the cell through the S-phase and into the M-phase, where cell division takes place. PGRN, unlike classic growth factors, is simultaneously both a competence and progression factor, that is, it successfully stimulates the completion of both S- and M-phase without assistance from other growth factors.
In murine embryonic fibroblasts, the dominant progression signal is provided through the insulin-like growth factor-I receptor (IGFI-R).83 Disrupting this receptor prevents mouse embryo fibroblasts from completing the cell cycle not only in response to IGF-I, the progression factor, but also to competence factors such as PDGF or EGF.84 PGRN, however, circumvented the requirement for the IGFI-R-mediated signal and supported the traverse of both the S- and M-phase, allowing the IGFI-R-deficient cells to complete the cell cycle in serum-free medium16 (Fig. 2B). PGRN is the only extracellular protein known to do this and appears to achieve this unusual feat because of the kinetics of its signal transduction response.16 Growth factors such as PDGF or EGF elicit relatively transient signaling responses in IGFI-R-deficient cells, whereas the PGRN response is considerably more prolonged.16 Whether this is due to events at the level of PGRN receptor-ligand interactions, or to PGRN eliciting weaker downstream counter-regulatory effects responsible for turning off the ERK signal after stimulation is unclear. Frustratingly, identifying the receptors or other PGRN-binding proteins in cell membranes has proven elusive, making it very difficult to address questions of this kind.
Progranulin in cancer
Signaling does not prove function. Evidence that PGRN is a functional growth factor came from work on cancer. Initially PGRN was found to be an autocrine growth stimulus for an aggressive murine teratoma.10 Reducing PGRN mRNA expression greatly reduced tumor formation by the teratoma cells,14 as well as in breast cancer,17 liver cancer,46 and squamous esophageal cancer18 cell lines in vivo. Clearly therefore PGRN is required for these cells to be tumorigenic. To demonstrate whether PGRN is not only necessary for tumor growth but also actively confers malignancy, it was over-expressed in a cancer cell line (SW13 adenocarcinoma cells) that is normally weakly tumor-forming. The PGRN over-expressing cells then formed substantial tumors in mice8 (Fig. 2C).
The potency of PGRN as a tumorigenic agent was demonstrated using primary human cells from ovarian epithelia20 and uterine smooth muscle.12 Human primary cells are difficult to transform by gene transfer in culture,85 requiring at minimum the increased expression of telomerase activity, blockade of the retinoblastoma (Rb) tumor suppressor system, and a strong mitotic drive provided in most experiments by oncogenes such as mutant RAS. PGRN can substitute for RAS in the transformation process. The combination of telomerase (Tert) and SV40 T-antigen (to block the Rb and P53 tumor suppressors) immortalized, but did not transform primary human ovarian cells20 or uterine smooth muscle cells,12 but when GRN was included with Tert and SV40, the primary cells became very tumorigenic in mice.12, 20 It is interesting to speculate whether this is related to the unusual property of PGRN as a conjoint competence and progression factor, although at present experimental evidence neither supports nor refutes this hypothesis.
PGRN supports tumor growth by increased proliferation,8, 15 decreased apoptosis,9, 29, 86–88 and greater invasiveness through the ECM.9, 19 Each of these actions requires the activity of the ERK and PI3K signal transduction pathways, although the extent to which either pathway contributes is variable.16 If PGRN promotes tumor growth as we propose here, blocking its action should block tumor growth. As discussed above, attenuating PGRN mRNA inhibits tumor growth of cancer cells in mouse models, but this might mean only that losing PGRN prevents the ability of the cells to seed. Is the depletion of PGRN relevant when dealing with large established tumors?
When liver cancer cells were transplanted into mice, allowed to form tumors and then exposed to injections of a PGRN monoclonal antibody, tumor growth was impeded by approximately 50%.25 Treating cancers in mice is far from treating people, but the liver cancer experiments show that PGRN has potential as a cancer drug target and is an extremely promising area for future investigation. Interestingly, the PGRN monoclonal antibodies not only inhibited proliferation of the cancer cells directly, but also had a striking anti-angiogenic action,25 possibly due to decreased secretion of the angiogenic growth factor VEGF in the treated tumors,19 although given that PGRN stimulated angiogenesis in wounds,28 a more direct effect on tumor vascularization can be postulated.
Progranulin interaction with other proteins
Extracellular protein-protein interactions regulate the activity of many growth factors, and PGRN is no exception. During inflammation, neutrophils release proteases such as elastase and proteinase-3 that digest PGRN into its individual GRN domains, which are then liberated as 6 kDa GRN peptides. In many instances the GRN peptides oppose the effects of intact PGRN. Thus while PGRN stimulates proliferation and inhibits the actions of TNF-α on neutrophils,29, 30 some of the GRN peptides inhibit cell proliferation,36 and stimulate inflammation by eliciting the secretion of interleukin-8.29 The critical balance between intact PGRN and the 6 kDa GRN peptides is maintained by a third party, the secretory leukocyte protease inhibitor (SLPI), which binds PGRN, and prevents proteolysis by neutrophil proteases. The activity of PGRN in a wound, or other site of inflammation, depends therefore on the levels of PGRN itself, the protective factor SLPI, and the neutrophil proteases (Fig. 2D). The significance of the triad effect – PGRN, SLPI, proteases – has been confirmed in compound knockouts of neutrophil elastase and proteinase-330 and in slpi knockout mice whose severely disrupted wound repair could be rescued by treating the wounds with PGRN.29 Moreover, the SLPI-PGRN interaction is not limited to inflammation since it has been implicated in ovarian tumor progression.89, 90
The interplay of proteases and PGRN may be a widespread determinant of PGRN activity. Metalloproteinases such as MMP-1491 and ADAMTS-792 (a disintegrin and metalloproteinase with thrombospondin-7) digest PGRN, with ADAMTS-7 inactivating the growth factor-like effects of PGRN during endochondral bone formation.92 This interaction has also been implicated in the pathogenesis of arthritis.93 PGRN binds the ECM proteins perlecan94 and chondrocyte oligomeric matrix protein (COMP);95 the perlecan interaction decreases the proliferative activity of PGRN, whereas the COMP interaction enhances it. In both cases PGRN binds its partner weakly, with affinities in the micromolar range,94, 95 and associates with the target protein through an EGF-containing module. PGRN reportedly binds the membrane protein Dlk,96 which is also rich in EGF modules, and although not yet tested, an affinity for EGF modules may prove a recurrent pattern in PGRN-protein interactions. Interactions of PGRN with intracellular proteins such as cyclin T have been reported, although their physiological significance is uncertain.97
GRN knockout and transgenic models
Grn knockout mice display behavioral abnormalities but few other recorded phenotypes.79 Given the biological actions that have been attributed to PGRN, the mild knockout phenotype suggests that PGRN is a molecular generalist, contributing to many tasks but acutely essential for few. Dissecting the biological functions of PGRN is being revealed through conditional transgenic and knockout strategies. For instance, keratinocyte-specific over-expression of PGRN leads to abnormal hair development, suggesting a role for PGRN in maintenance of hair follicles.98
When we first identified the GRN peptides as minor side fractions on a chromatogram, there was nothing to suppose that what would emerge was an extracellular signaling gene family that extends back to green plants and slime mold, or that would be implicated in so many biological functions. In recent years knowledge of the structure and function of products of GRN gene expression has expanded enormously, particularly in the areas of cancer and neurobiology, but also in the fields of development, tissue repair, and inflammation. Much remains to be discovered, and hopefully the next few years will see breakthroughs that will put the latest insights on a firm footing through identification of PGRN receptors and binding proteins, definition of its function in nerve cells, and its application as a therapeutic target.
The research carried out in the laboratories of the authors and reviewed in this article was supported by grants from the Canadian Institutes for Health Research, the National Cancer Institute of Canada, and the Canadian Breast Cancer Research Association Ideas Program. We gratefully acknowledge Dr. Thomas D. Bird, University of Washington School of Medicine, Seattle, for permission to use the image in Fig. 2A.