Progranulin: A Proteolytically Processed Protein at the Crossroads of Inflammation and Neurodegeneration*

GRN mutations cause frontotemporal lobar degeneration with TDP-43-positive inclusions. The mechanism of pathogenesis is haploinsufficiency. Recently, homozygous GRN mutations were detected in two patients with neuronal ceroid lipofuscinosis, a lysosomal storage disease. It is unknown whether the pathogenesis of these two conditions is related. Progranulin is cleaved into smaller peptides called granulins. Progranulin and granulins are attributed with roles in cancer, inflammation, and neuronal physiology. Cell surface receptors for progranulin, but not granulin peptides, have been reported. Revealing the cell surface receptors and the intracellular functions of granulins and progranulin is crucial for understanding their contributions to neurodegeneration.


Progranulin Haploinsufficiency Causes Frontotemporal Lobar Degeneration with Ubiquitinated TDP-43-positive Inclusions
In 2006, mutations in GRN were discovered to be a cause of frontotemporal lobar degeneration (FTLD) 3 with ubiquitinated TDP-43-positive inclusions (FTLD-TDP) (15,16). FTLD is the second most common presenile dementia disorder after Alzheimer disease, representing 5-15% of all dementias (17,18). More than 70 mutations in GRN, almost all of which result in null alleles, have been identified in FTLD patients. A few causative missense mutations also result in reduced levels of progranulin (19).
Clinical manifestations of heterozygous loss-of-function GRN mutations include variants of the FTLD spectrum, parkinsonism, and the corticobasal syndrome (20). Neuropathologically, atrophy of the brain parenchyma (most severe in the frontal cortex) is usually observed. The atrophy can be asymmetrical, and different brain regions are affected with varying frequency. Loss of pigmentation of the substantia nigra, hippocampal sclerosis, and atrophy of temporal and parietal lobes are variably observed (20). The characteristic cellular pathology is neuronal cytoplasmic inclusions and dystrophic neurites. These inclusions are positive for TDP-43, an RNA-binding protein and splicing modulator that binds GRN mRNA (21,22). TDP-43 protein in these inclusions is ubiquitinated and hyperphosphorylated and may be proteolytically processed (23). Loss of normal nuclear staining for TDP-43 is typical. Gliosis is also commonly observed (Table 1).
In hereditary cases, the mode of inheritance is autosomal dominant with incomplete penetrance (24,25). Serum progranulin levels are lower in mutation carriers and patients (lower than ϳ60 ng/ml) than in controls (higher than ϳ125 ng/ml) (26,27). Based on the rationale that FTLD-TDP with GRN mutations is caused by haploinsufficiency of progranulin, small molecule enhancers of progranulin expression have been pursued as potential treatments (28,29).
NCLs are genetic progressive lysosomal storage diseases characterized by accumulation of lipofuscin (31). At least 10 related disorders are now classified as NCLs. Causative muta-tions occur in genes encoding lysosomal enzymes and several incompletely characterized membrane proteins.
Lipofuscin is an aggregate of oxidized cross-linked proteins and lipids. It can be toxic to cells by chelating metals, enhancing oxidative damage, and inhibiting mitochondrial and lysosomal function (32). Lipofuscin accumulation occurs during normal aging but is greatly accelerated in NCLs. Interestingly, increased accumulation of lipofuscin has not been reported in cases of FTLD-TDP but has been detected in mouse models of the disease (see below).

Mouse Models of Progranulin Deficiency
Several independent mouse lines with genetic Grn deletions have been generated. Behaviorally, the most consistent finding is social interaction deficits (33)(34)(35). In a classic test of hippocampal learning and memory (Morris water maze), Grn Ϫ/Ϫ mice had mild deficits at old age (18 -21 months) in two studies (33,36) but no deficits at 8 months of age in another study (35). Other reported behavioral deficits include depression-like behavior and either increased (35,37) or decreased (33) anxiety. These behavioral phenotypes are fairly consistent with the clinical manifestations of FTLD, which include early behavioral problems and later deficits in memory.
Histopathologically, robust microgliosis, astrogliosis, and increased ubiquitin staining are observed in the brains of aged Grn Ϫ/Ϫ mice (7,(33)(34)(35)38). Ahmed et al. (7) recently showed that intracytoplasmic ubiquitinated aggregates observed in these mice are probably composed of lipofuscin. This finding was replicated in an independent Grn Ϫ/Ϫ line (36). Although some vacuolation was observed in the habenula and hippocampus in very old (23 months) Grn Ϫ/Ϫ mice (7), overt neuronal loss seems to be very mild or absent (34), in contrast with the severe atrophy observed in human FTLD. Yin et al. (33) observed increased staining with an antibody against phosphorylated TDP-43 in the brains of 18-month-old Grn Ϫ/Ϫ mice; however, none of the other studies detected overt TDP-43 proteinopathy. Wils et al. (36) recently reported somewhat increased phospho-TDP-43 immunoreactivity in detergent-insoluble fractions of Grn Ϫ/Ϫ mouse brains; nonetheless, they did not detect a significant difference in TDP-43 staining by immunohistochemistry. Interestingly, none of these findings were reported in heterozygous Grn ϩ/Ϫ mice, which would be analogous to the haploinsufficient condition in human FTLD-TDP.
Progranulin-deficient mice display dysregulated immune responses in the brain (38) Macrophages from Grn Ϫ/Ϫ mice express higher levels of proinflammatory cytokines (MCP-1, CXCL1, IL-6, IL-12p40, and TNF-␣) in response to LPS, but they express less IL-10. Microglia cultured from these animals have toxic effects on co-cultured wild-type neurons. However, the immunomodulatory role of progranulin in the periphery may be different. In a recent study, Grn Ϫ/Ϫ mice on a high fat diet had reduced IL-6 concentrations in blood and adipose tissue. Interestingly, Grn ablation was protective against insulin resistance (5). Finally, loss of the progranulin homolog results in accelerated clearance of apoptotic cells in C. elegans (14) and disruption of motor neuron development in zebrafish (39).

What Does Progranulin Do?
Reported biological activities of progranulin fall into three broad categories: growth factor-like activities, modulation of immune responses, and neuronal effects. Progranulin is overexpressed in many human and experimental tumors, including carcinomas (40 -43), gliomas (44), and sarcomas (45). It may act akin to a growth factor, stimulating proliferation (46,47), survival (48), and invasion (49). Progranulin has been reported to activate many of the typical cell proliferation signaling pathways, including ERK, PI3K, and Akt pathways (48,50,51), not only in tumor cells but also in neurons (52,53). Progranulin may be a prognostic marker (54) or a therapeutic target in cancer: progranulin overexpression confers an aggressive phenotype to adenocarcinoma (46), immortalized ovarian epithelial cells (55), breast cancer (49,56), and hepatocellular carcinoma (57), and anti-progranulin treatment reduces in vivo tumorigenicity of teratoma (58) and breast cancer cell lines (59). However, it should be emphasized that although these early studies strongly linked progranulin to cancer, no cell surface receptor has been shown to mediate these effects.
Full-length progranulin is generally anti-inflammatory; whereas proteolytically released granulins may have the opposite effect (see below). Progranulin reduces reactive oxygen species production by immune complex-activated neutrophils (60) and blocks TNF-␣-induced immune responses, namely respiratory burst, degranulation, and spreading of adherent human neutrophils (61). Progranulin also attenuates TNF-␣-induced IL-8 release (62). A recent finding suggests that these activities may be mediated at the level of TNF receptors (TNFRs) (63). Importantly, recombinant or proteolytically released granulins do not antagonize TNF-␣.
Progranulin expression is induced by inflammatory stimuli in astrocytes (64). Consistent with its immunomodulatory role, progranulin expression was found to be induced in multiple sclerosis, a classic example of neuroinflammation (65). How- Neuronal loss and gliosis, neuronal cytoplasmic and intranuclear inclusions, dystrophic neurites, ubiquitinated phosphorylated TDP-43 aggregates "Fingerprint profiles" seen by EM in skin biopsy samples, brain pathology unknown ever, no difference in cerebrospinal fluid progranulin levels was reported in multiple sclerosis in an earlier study (66). Pickford et al. (67) reported that progranulin may also be a chemotactic cue for microglia. In this study, intracerebral injection of progranulin led to microgliosis in excess of that seen in a control lesion. This is somewhat dissimilar to the knock-out studies discussed above, where the deletion of progranulin was shown to lead to gliosis. In contrast, Kessenbrock et al. (60) have reported that recombinant progranulin reduced neutrophil infiltration in a reverse passive Arthus reaction model. These disparate results may perhaps be explained by differential in vivo proteolysis of progranulin into proinflammatory granulins or by differential effects on microglia and neutrophils. Progranulin expression is stimulated in the early phases of wound healing together with proinflammatory mediators (61). In this context, progranulin may be an attractant for neutrophils, monocytes, fibroblasts, and endothelial cells. It also stimulates tube formation of endothelial cells (68). In mice with genetic deletion of antileukoproteinase (secretory leukocyte protease inhibitor (SLPI)), an inhibitor of progranulin degradation, exogenous progranulin was shown to enhance cutaneous wound healing (61). Recently, progranulin was found necessary for efficient activation of TLR9 (Toll-like receptor 9) by CpG oligodeoxynucleotides (69). In this study, macrophages from Grn Ϫ/Ϫ mice had a muted response to CpG. However, it was somewhat unclear whether full-length progranulin or granulins mediated this effect.

How Does Progranulin Affect Neurons?
An early study by Van Damme et al. (12) showed that progranulin induced neurite outgrowth. This has been replicated by Gao et al. (53) but could not be replicated by Hu et al. (10). Exogenous recombinant progranulin increased survival of motor neurons in serum-free conditions in the study by Van Damme et al., but knocking down progranulin to ϳ20% of normal levels did not affect survival of hippocampal neurons in another study (70). Two recent studies investigated the effect of progranulin deficiency on neuronal morphology and synaptic transmission. Both genetic deletion of Grn in mice (35) and siRNA-mediated knockdown of progranulin (70) led to reduced dendritic length and reduced spine density in hippocampal neurons. These results seem to support the abovementioned neurite outgrowth-promoting effects of progranulin.
Electrophysiologically, the ratio of field excitatory postsynaptic potential slope to afferent volley amplitude was diminished in hippocampal slices prepared from Grn Ϫ/Ϫ mice (35). This study also reported that induction of long-term potentiation in Grn Ϫ/Ϫ slices was more difficult and that mean longterm potentiation amplitude (change in field excitatory postsynaptic potential slope) was diminished in Grn Ϫ/Ϫ slices. The authors interpreted these findings as suggesting reduced synaptic connectivity and impaired synaptic plasticity in Grn Ϫ/Ϫ mice. However, the variation from slice to slice was large, especially in the Grn Ϫ/Ϫ group.
Progranulin knockdown in cultured hippocampal neurons resulted in reduced numbers of co-localized pre-and postsynaptic markers (i.e. reduced synapse density), an increased num-ber of synaptic vesicles per synapse (as revealed by electron microscopy), and increased miniature excitatory postsynaptic current frequency (70). This suggests decreased synaptic connectivity but enhanced transmission at individual synapses, indicative of a homeostatic response.

What Is the Progranulin Receptor?
Progranulin has a classic N-terminal signal peptide and several N-glycosylation sites (71). It is readily secreted in cell culture and detected in serum and cerebrospinal fluid in animals. Exogenous recombinant progranulin has many biological effects as detailed above. Hence, several groups have searched for progranulin receptors.
The first cell surface protein conclusively shown to bind progranulin was sortilin (10). Sortilin has diverse biological functions, including prosaposin trafficking to lysosomes (72), hepatic VLDL secretion (73), and proneurotrophin-induced apoptosis (74). It is a regulator of extracellular progranulin levels (75). Sortilin knock-out mice are grossly normal, have increased levels of extracellular progranulin, and are resistant to motor neuron injury (76). Sortilin does not seem to be a signal-transducing receptor itself but acts as a coreceptor for the low affinity neurotrophin receptor p75 NTR . Neurotrophins are overexpressed in cancer (77) and implicated in neurodegeneration and inflammation (78). Sortilin binds via the C terminus of progranulin, suggesting that the C-terminal granulin domain can potentially mediate the binding even after proteolytic cleavage. However, the precise role of progranulin binding in sortilin function remains unknown.
Perhaps the most interesting candidates for the elusive progranulin receptor are the TNFRs. In 2011, Tang et al. (63) reported that progranulin bound to TNFRs with high affinity and blocked the binding of TNF-␣. They also showed that progranulin or a synthetic progranulin fragment was therapeutic in an arthritis model (63), essentially acting as an endogenous TNF-␣ antagonist. Replication of these findings will be immensely valuable to the field because TNF-␣ antagonism can potentially explain most biological effects attributed to progranulin. Several lines of evidence, especially the knock-out models, strongly suggest that progranulin has anti-inflammatory effects. Furthermore, Zhu et al. (61) previously reported that full-length progranulin suppressed TNF-␣-induced neutrophil activation. Perhaps, progranulin haploinsufficiency leads to an overabundance of TNF-␣ activity. However, it is somewhat puzzling that an excess of TNF-␣ activity would lead to a specific neurodegenerative phenotype rather than any of the other inflammatory conditions associated with this cytokine (79), most notably arthritis. A recent report suggested that the serum levels of IL-6, TNF-␣, and IL-18 were unchanged in GRN mutation carriers compared with controls (although IL-6 levels were increased in symptomatic GRN mutation carriers) (80). We are not aware of any published work detailing the prevalence of a TNF-␣-related autoimmune disease (e.g. rheumatoid arthritis) in GRN mutation carriers or FTLD patients. An epidemiologic study showing increased risk of inflammatory disease for individuals with progranulin deficiency will be required to strengthen the hypothesis that progranulin is an endogenous TNF-␣ antagonist.
Antagonism of TNF-␣ can potentially explain some of the cancer-promoting effects of progranulin. TNF-␣ was first discovered as a humoral factor that caused rapid hemorrhagic necrosis of experimental tumors. It is cytotoxic to several tumor cell lines in vitro, increases endothelial permeability, may stimulate an antitumor immune response, and is in clinical use for immunotherapy of limb sarcomas (81). By overexpressing progranulin, a presumed TNF-␣ antagonist, tumors may escape TNF-␣ toxicity. However, this hypothesis remains untested.
TNF-␣ is involved in synaptic scaling, maintaining synapses in a plastic state (82). After silencing, TNF-␣ enables up-regulation of surface AMPA receptor expression and miniature excitatory postsynaptic current amplitudes (83). If progranulin is acting as a TNF-␣ receptor antagonist at the synapse, we would expect that progranulin deficiency would be functionally equivalent to overabundance of TNF-␣. However, this does not seem to be the case (70).
Unlike the case of full-length progranulin, no cell surface receptor has been shown to mediate the biological effects of proteolytically released granulins. Individual granulins do not bind TNFRs, whereas only granulin E is expected to bind sortilin. A precursor protein being proteolytically processed into active species is a relatively common mechanism in cellular biology (e.g. TGF-␤ family of growth factors). We critically examine the case of granulins below.

Proteolytic Cleavage of Progranulin and the Case of Granulins
In the 1990s, several ϳ6-kDa granulin peptides were purified from leukocyte granule extracts and bone marrow (84), kidney (85), and various other biological sources (86). It was later discovered that all of these peptides are encoded by a single gene (GRN) translated into a large precursor protein, progranulin (87). Granulins are released following proteolytic cleavage of progranulin. Human progranulin contains seven full-length granulin domains and one half-length granulin domain. These granulins are named, from the N terminus of progranulin to the C terminus, granulins p, G F, B, A, C, D, and E, with "p" denoting the half-length "paragranulin" domain ( Fig. 1). The molecular structure of individual granulin domains has been solved. Each granulin domain consists of parallel stacked ␤-hairpins held together by six disulfide bonds (Fig. 1) (88, 89).
Progranulin is proteolytically cleaved by neutrophil elastase (61), proteinase 3 (a neutrophil protease) (60), MMP-12 (matrix metalloproteinase 12; macrophage elastase) (64), MMP-14 (90), and ADAMTS-7 (a disintegrin and metalloproteinase with thrombospondin motifs 7) (91). Zhu et al. (61) have mapped the neutrophil elastase cleavage sites and shown that cleavage occurs in the linker regions between granulin domains. However, they did not detect any cleavage sites between granulins F, B, and A ( Fig. 1) even though granulins A and B had been individually purified from various sources. Notably, incubation of recombinant progranulin with these proteases does not always result in the release of solely 6 -12-kDa fragments as would be expected if the ϳ80-kDa precursor protein was completely processed to 7 1 ⁄ 2 granulin domains. At least five intermediate products larger than 15 kDa seem to be present after overdigestion with any of these proteases according to data presented previously (60,61,64). However, Kojima et al. (62) observed mostly Ͻ6-kDa bands after 16 h of digestion at 37°C with elastase. After in vitro incubation with MMP-12, several fragments (15-45 kDa) were still detectable with an antibody against the C terminus of progranulin (64).
SLPI protects progranulin from proteolysis by elastase. Interestingly, SLPI binds progranulin directly, and this interaction is protective against proteolysis even when the active site of SLPI  (61). Asterisks denote linker regions where proteolytic cleavage also takes place, but the protease that releases granulins A and B has not been conclusively identified. The amino acid sequence of granulin A is shown at the bottom. Cysteines are highlighted in red. Numbers denote approximate positions of granulin domains relative to full-length human progranulin (593 residues).
Even though progranulin is known to be secreted and circulates in the full-length form in blood, proteolytic processing probably takes places intracellularly to some extent. Progranulin cleavage products are detected in cellular fractions, and double deletion of neutrophil elastase and proteinase 3 increases intracellular progranulin levels (60). MMP-12 cleaves progranulin intracellularly in microglia but not in conditioned medium (64). An early study that identified an acrosomal glycoprotein later shown to be identical to progranulin also showed that progranulin was partially proteolyzed as the sperm moved along the epididymis (93). MMP-14 is active in the Golgi apparatus, but the physiological significance of progranulin cleavage by MMP-14 is undetermined.
Biological effects have been attributed to the granulin peptides. Granulin A has been reported to induce anchorage-independent growth of cultured keratinocytes and fibroblasts while apparently inhibiting proliferation of other cancer cell lines (85,87,94). The effect of granulin B was generally inhibitory and antagonistic to granulin A. At least in one case (87), fulllength progranulin did not have the same activity as granulin A. An independent group reported that granulin D increased DNA synthesis in cultured astrocytes and, to a limited extent, in primary glioblastoma cells (44). Granulin E was reported to act similarly to progranulin and to support neuronal survival in cell culture (12). Notably, purified recombinant granulins do not have the anti-inflammatory activities of full-length progranulin but, on the contrary, seem to be proinflammatory (61). For example, elastase-digested progranulin induced IL-8 release from A549 cells, whereas recombinant granulin B induced IL-8 release from both A549 and SW-13 cells.

Perspectives
The number of distinct biological activities attributed to progranulin is remarkable. How can we reconcile these seemingly nonspecific functions with the fact that progranulin deficiency leads to two specific neurodegenerative conditions (FTLD in the case of haploinsufficiency and NCL in the case of a homozygous mutation)? Progranulin is almost certainly a multifunctional protein, and multifunctional proteins are common in eukaryotic organisms (95). Nevertheless, independent replication, particularly with cautious scrutiny toward impurities in progranulin and granulin preparations, may whittle this large number of biological activities down to a shorter list of bona fide progranulin functions. Furthermore, we believe that granulin peptides are almost certainly bioactive and that a search for granulin receptor(s) is warranted. Of particular interest is whether the recombinant progranulin is proteolyzed in tissue culture conditions, an observation that we feel has been insufficiently addressed.
Could GRN haploinsufficient FTLD-TDP and NCL with homozygous GRN mutations be manifestations of the same spectrum of diseases? In NCL, time of disease onset varies from the neonatal period to young adulthood according to the mutated gene, but patients with "mild" hypomorphic mutations are known to present with later onset than those with "classic" loss-of-function mutations of the same genes (31). The patients with homozygous loss-of-function mutations of progranulin reported previously (30) presented in their 20s, which is quite late compared with other known NCL mutations. Perhaps, GRN haploinsufficient FTLD-TDP represents a very late onset form of atypical NCL. This hypothesis seems to be supported by the mouse models: Grn Ϫ/Ϫ mice display behavioral characteristics of FTLD together with lipofuscinosis. However, exaggerated lipofuscinosis has not yet been reported in FTLD-FIGURE 2. Trafficking of prosaposin and progranulin. Both prosaposin and progranulin consist of several repeats of saposin and granulin domains, respectively (step 1). Both proteins are N-glycosylated (step 2) and secreted (step 3). Prosaposin is also directly transported to the lysosomes (Lys) via the mannose 6-phosphate receptor (step 4). Sortilin also plays a role in lysosomal trafficking of prosaposin (not shown). Reuptake of progranulin is mediated by sortilin, whereas prosaposin reuptake is mediated by the LDL receptor-related protein (LRP), the mannose receptor (not shown), and the mannose 6-phosphate receptor (not shown) (step 5). Both proteins are probably proteolyzed intracellularly, although this has not been shown directly for progranulin (step 6). In the lysosome, the saposins activate lysosomal enzymes (pink pentagon) partly by lifting their substrates (green) out of the lipid bilayer (step 7). Lysosomal functions of granulins remain unknown. Prosaposin is shown in blue, progranulin is shown in red. ER, endoplasmic reticulum. TDP patient brains (or other tissues), in which the main ubiquitinated protein is known to be TDP-43. Furthermore, different regions of the brain seem to be primarily affected in FTLD-TDP compared with NCL. Thus, it remains possible that FTLD-TDP and GRN mutant NCL are caused by disruptions of disparate pathways that are differentially sensitive to progranulin dosage.
Another relatively poorly understood concept is the role of intracellular progranulin or granulins. Given the recent finding of adult onset lysosomal storage disease caused by homozygous progranulin deficiency and the high affinity binding between progranulin and the lysosomal transport protein sortilin, an essential intracellular role for progranulin in lysosomal biology is a possibility. Moreover, another recently identified function of progranulin/granulins, binding of CpG oligodeoxynucleotides to TLR9, takes place intracellularly in lysosomes. It is also likely that proteolytic cleavage of progranulin into granulins occurs in lysosomes (or somewhere along the way from the endoplasmic reticulum to the lysosome). Could progranulin or granulins be acting as activators or transporters of lysosomal enzymes? Accumulation of the autophagy-related receptor p62 and the lysosomal protease cathepsin D was reported in Grn Ϫ/Ϫ mice. Perhaps progranulin functions akin to prosaposin, which is likewise synthesized as a precursor protein, secreted, and taken up again into the cells, followed by proteolytic processing into small (8 -11 kDa) bioactive saposin polypeptides (Fig. 2) (96,97). Saposins act as activators of lysosomal enzymes, and genetic abnormalities of prosaposin cause lysosomal storage disorders of varying severity. Lysosomal enzymes also play roles in the regulation of immune responses, apoptosis, and defense against pathogens (98,99). Therefore, lysosomal functions of progranulin/granulins could explain their immunomodulatory properties. We believe that progress along these lines will ultimately lead to the identification of the biological mechanisms underlying selective vulnerability of frontotemporal regions of the brain to progranulin deficiency and, we hope, to novel drug targets for the treatment of GRN-deficient conditions.