Lipidomic and Transcriptomic Basis of Lysosomal Dysfunction in Progranulin Deficiency

1Lysosomes are central to lipid homeostasis via clearance of extracellular lipoproteins and autophagy. Upon fusion with lysosomes, the contents of autophagic and endocytic vesicles are catabolized by lysosomal hydrolases, and the breakdown products are released for re-use. Inherited defects in lysosomal hydrolases and other lysosomal proteins lead to the accumulation of lipids and other cellular constituents within dysfunctional lysosomes. These rare inborn errors of metabolism have been broadly grouped as lysosomal storage disorders (LSDs). In comparison to other cell types, neurons are disproportionately affected in many of the prototypical LSDs, including neuronal ceroid lipofuscinose (NCL), Niemann-Pick type C (NPC), and Sandhoff disease.

2Lysosomes are implicated in the pathogenesis of adult neurodegenerative diseases. In human subjects with Alzheimer's disease (AD), lysosomal dysfunction may underlie changes in brain lipid metabolism, particularly in diacylglycerides (DAGs) and phosphatidylcholines (PCs). Mutations in granulin (GRN), the gene encoding progranulin (PGRN), causes frontotemporal lobar degeneration (FTLD). The 88-kDa propeptide PGRN is trafficked to the lysosome through direct interaction with sortilin or in complex with prosaposin (PSAP), an established regulator of lysosomal lipid metabolism, via sortilin-independent mechanisms. PGRN is thought to be proteolytically processed within the lysosome into smaller subunits, called granulins, which have been implicated in the regulation of cell growth and embryonic development. However, the precise role of PGRN and granulins within the lysosome is unknown.

3FTLD causes a variety of changes in behavior and in language production and comprehension. The neuropathological features of FTLD include selective atrophy of the frontal and temporal lobes due to neuronal loss. At the cellular level, FTLD can be divided into subtypes based upon the immunoreactivity of characteristic subcellular protein aggregates, most commonly either transactive response (TAR) DNA-binding protein 43 (TDP-43) (FTLD-TDP) or the microtubule-associated protein tau (FTLD-tau), within neurons and/or glia. Most FTLD cases are sporadic: ∼40% of cases are familial and usually demonstrate autosomal-dominant inheritance. A number of genes, such as C9orf72, MAPT (encoding tau), and GRN, have been associated with these familial forms of FTLD.

4Nearly all human patients with FTLD due to PGRN deficiency harbor heterozygous loss-of-function (LoF) GRN mutations and demonstrate FTLD-TDP neuropathology. Human subjects with heterozygous LoF GRN mutations have elevated lysosomal proteins, including saposin D, cathepsin D, and LAMP1, within brain tissue. Brain tissue from Grn knockout (Grn −/−) mice exhibited increased levels of lipofuscin particles, phosphorylated TDP-43 protein, and lysosomal proteins, such as cathepsin D and LAMP1. Patients with homozygous LoF-GRN mutations exhibit symptoms of adultonset NCL: skin biopsies revealed enlarged lysosomes containing lamellar, pseudomembranous “fingerprint”-like inclusions of the type seen in NCL. Therefore, the pathogenesis of FTLD-TDP and NCL caused by PGRN deficiency may share gene-dosage-dependent mechanism involving lysosomal dysfunction.

5We hypothesized that PGRN is involved in lysosomal homeostasis and that it participates in the metabolism of membrane lipids within the lysosome. Here, we show that PGRN deficiency alters lysosome morphology and abundance in hippocampal neurons. Using an unbiased lipidomic analysis, we show that the lipidomic profiles of human subjects and mouse models with PGRN deficiency exhibit characteristic differences in brain lipids that are gene dosage and disease specific. Loss of PGRN leads to the accumulation of long, polyunsaturated triacylglycerides (TAGs) and a concomitant reduction of DAGs and phosphatidylserines (PSs) in both the whole-cell and enriched lysosome lipidomes. Comparison between gene expression arrays from brain tissue from NPC1- and PGRN-deficient mice reveals distinct, but partially overlapping, expression patterns of lysosomal, immune-related, and lipid metabolic genes. Thus, neurodegenerative diseases caused by PGRN deficiency may resemble classical LSDs like NPC and Sandhoff disease.

6Human patients and mice with homozygous LoF-GRN mutations develop NCL with characteristic lysosomal morphology and pseudomembranous lamellar inclusions. Moreover, human FTLD-TDP patients with GRN mutations and Grn knockout (Grn−/−) mice exhibit increased lipofuscin and immunoreactivity of lysosomal markers, such as cathepsin D and LAMP1, in brain tissue, further implicating disruption of lysosomal homeostasis with PGRN deficiency.

7We used electron microscopy (Figures 1A–1C) to examine our Grn mutant mouse model for differences in lysosomal abundance and morphology. The shape of the lysosomes and their contents were altered in Grn−/− mouse brains, with neuronal lysosomes exhibiting the classic fingerprint inclusions of NCL (Figures 1A and 1C). These enlarged, inclusion-filled lysosomes were ultrastructurally distinct from lipofuscin particles, which were excluded in the quantification studies (Figures 1B and 1C). Whereas there was little discernible difference in the appearance and size (Figures 1A and 1C) of neuronal lysosomes in Grn+/+ and Grn+/− mouse brains, the total number of lysosomes was increased in neurons from Grn+/− and Grn−/− mice (Figure 1B).

8The prominent lamellar lysosomal inclusions in human GRN−/− skin fibroblasts and in mouse Grn−/− neurons (Figure 1A) are suggestive of a defect in lysosomal metabolism leading to the pathologic accumulation and storage of at least partially membrane-derived lipids. We therefore performed an unbiased lipidomic analysis of brain tissue from human subjects with autopsy-confirmed and genetically defined neurodegenerative pathologies (Table S1; Figures 2A and 2B) and from 12-month-old Grn mutant mice (Figures 2C and 2D). Infusion mass spectrometry (MS/MSALL) yielded 8,602 (human) and 13,214 (mouse) identified and unidentified lipid species based exclusively on mass features (Tables S2A and S2D). Linear discriminant analysis (LDA) of all 8,602 detected mass features revealed distinct changes in lipid composition in human brain lipid extracts (Figures 2A and 2B; Table S2B). Intriguingly, these changes were characteristic for different neuropathologies and allowed for discrimination between AD, FTLD-TDP due to PGRN haploinsufficiency, and non-PGRN-related forms of FTLD-TDP, based on lipid composition alone (Figure 2A). The differences between the neuropathologies were mainly driven by TAG, DAG, phosphatidylethanolamine (PE), and PS subspecies (Figure S1A; Table S2C). Similar to the human brain lipidomic analysis, a LDA score plot using all 13,214 detected mass features from mouse brain lipid extracts demonstrated distinct, genotype-specific grouping between wild-type and Grn mutants (Figures 2C and 2D; Table S2E). The differences between mouse genotypes were again mainly driven by TAG, DAG, PE, and PS subspecies (Figure S1B; Table S2F).

9Taken together, these data suggest that a dysfunction in lysosomal lipid metabolism may underlie the characteristic and disease-specific changes of brain lipid composition in PGRN-haploinsufficient human subjects and that either PGRN or individual granulins are involved in the metabolism of these lipids.

10Because human and mouse (Figures 1A-1C) PGRN-deficient cells showed abnormal lysosomal morphology and abundance, we determined whether the characteristic differences in the lipidomic patterns seen in PGRN-haploinsufficient human and Grn mutant mouse brains (Figures 2 and S1) could also be observed in peripheral tissues. Total cellular lipid extracts from immortalized, non-clonal Grn+/+, Grn+/− and Grn−/− mouse embryonic fibroblasts (MEFs) were analyzed using our infusion-based, lipidomic mass spectrometry system. 11,642 mass features were detected, and 392 were positively identified (Table S2G). With decreasing Grn gene dosage, TAGs progressively accumulated and both DAGs and PSs were reduced in Grn+/− and Grn−/− MEFs as compared to wild-type MEFs (Figure 3A). When individual TAG species were plotted against increasing acyl chain length and unsaturation level, PGRN deficiency led to a preferential accumulation of TAGs with long acyl chains and high unsaturation levels (ranging to more than an 8-fold increase in some TAG species) in Grn+/− and Grn−/− MEFs as compared to wild-type MEFs (Figure 3B). Cholesteryl esters (CEs) also exhibited a distinct genotype-specific pattern, with a relative increase in CEs containing saturated and monounsaturated acyl chains in Grn+/− and a concomitant reduction in CEs containing polyunsaturated acyl chains in Grn−/− cells, potentially arising from reduced substrate availability to acyl-coenzyme A (CoA):cholesterol acyltransferase caused by the complete loss of PGRN and granulin function.

11These data suggest that PGRN or individual granulins participate in the metabolism of long-chain, polyunsaturated fatty acids, possibly by influencing the activity of lysosomal metabolic enzymes. This may involve the conversion of other lipid species, such as phosphatidic acid, to DAGs or the egress of DAGs from the lysosometo other compartments, i.e., the endoplasmic reticulum (ER). The lipid profile changes observed in MEFs are consistent with the data obtained from the human and mouse brain samples, further supporting our hypothesis that PGRN participates in lipid metabolism within the cell, presumably within the lysosome itself.

12The liver is the central clearing house for lipids in the body, where nutritional lipids are first received and then repackaged for redistribution to peripheral tissues, mainly muscle and adipose tissue. It is also the site of the highest peripheral PGRN expression, and cell fractionation procedures for the selective enrichment of biochemical quantities of hepatocyte organelles, including lysosomes, have been well established. Using a standard protocol, we generated subcellular fractions highly enriched for lysosomes (Figures S2A–S2C). Fractionation efficiency was tested using a variety of organelle markers for lysosomes (NPC1, LAMP1, and cathepsin D), endosomes (EEA1), plasma membrane (Na+/K+ ATPase), ER (ACAT1), Golgi (Gm130), and peroxisomes (PMP70; Figure S2D). Unprocessed PGRN did not appear to be enriched in lysosomes compared to whole-cell lysate (Figure S2E). This may be due to PGRN precursor sequestration within secretory vesicles and not in lysosomesorrapid processing of the precursor to granulins within lysosomes.

13Infusion-based lipidomics of the enriched lysosomes (Figure 4; Table S2H) revealed a similar gene-dosage-dependent increase in a subset of TAGs with a pronounced reduction in the majority of PSs and PEs (Figures 4A and 4B) but no significant change in DAGs (Figure 4A). The differences between the MEF and liver lysosome lipidomes may be attributed to physiologic differences between cultured MEF cells and hepatic tissue. For example, MEFs endogenously synthesize virtually all lipids, whereas hepatocytes and other liver cells receive a significant portion of their lipids exogenously from the diet.

14Various LSDs are due to a deficiency of specific lysosomal hydrolases. Though not a hydrolase itself, NPC1 can affect the functioning of lysosomal hydrolases and overall lysosomal activation, as measured by cathepsin B activity, through the aberrant accumulation of lipids within the lysosome. Lysosome enrichment by subcellular fractionation showed, for example, that the complete absence of PGRN in lysosomes increased cathepsin D expression and altered its processing (Figure S3).

15A recent study using microarray profiling showed increased expression of lysosomal and innate immunity genes in the brains of Grn−/− mice. To obtain a complete overview of the entire complement of RNA expression that is altered in Grn deficiency, with a particular focus on transcripts involved in lipid metabolism, we sequenced the mRNA extracted from the brains of 7-month-old Grn+/+, Grn+/−, and Grn−/− mice. Of 22,552 known mouse genes, loss of one or both alleles of Grn altered the transcription of 774 genes compared to Grn+/+ controls (Figure 5A; Table S3). Of the transcripts differentially expressed in either Grn + − or Grn−/− brains, approximately 50% were similarly and significantly altered in both genotypes (Figure 5B; Tables S4 and S5). Several differentially expressed transcripts in both Grn+/− and Grn−/− mice have functions in lipid metabolism (Figure 5C; Tables S4 and S5). Of these, a subset was shared between Npc1−/− and Grn mutants. Moreover, brain mRNA levels of several lipid hydrolases with functions consistent with compensatory changes to the cellular lipidome were significantly altered in Grn+/− and Grn−/− mice as compared to Grn+/+ controls (Figure S4). Our transcriptomic analysis of mRNAs from Grn mutant mouse brains was independently validated by the concordant results of innate immunity genes that matched those previously reported by Lui and colleagues using a microarray approach.

16To determine whether PGRN deficiency results in similar transcriptional changes as seen in other mouse models of lysosomal storage diseases, we used a publicly available microarray dataset of brain mRNAs isolated from Npc1-deficient (Npc1−/− mice. Significantly altered transcripts were compared between the transcriptomic datasets from Npc1−/− Grn+/− and Grn−/− mice (Figures 5B and 5C; Tables S4 and S5). Of the differentially regulated genes in Grn+/− and Grn−/− brains, a fraction was similarly up-or downregulated in Npc1−/− brains. A subset of the shared upregulated genes in the Npc1−/− and Grn mutant brains (Figure 5B) are associated with either lysosomal function or the immune response (Table S5). These findings are consistent with the results in Npc1-deficient mouse brains where immune-related and lysosomal genes are upregulated.

17The magnitude of change among lysosomal, immune-related, and lipid metabolic genes that were differentially regulated in Grn+/− and Grn−/− brains was Grn dose dependent (Figures 5D–5F). Importantly, the transcriptomic data revealed distinct, but partially overlapping, expression patterns (Figures 5D-5F). This observation is consistent with the notion that pathophysiological molecular mechanisms are, in part, differentially controlled by NPC1 and PGRN.

18These data suggest that a reduction or complete loss of PGRN induces characteristic changes in the expression of genes associated with lysosomal function, the immune response, and lipid metabolism, indicative of lysosomal dysfunction. These findings further indicate that neurodegenerative diseases due to PGRN deficiency may be driven by similar pathophysiologic mechanisms as classic LSDs, such as NPC disease.

19PGRN deficiency manifests itself clinically as two phenotypically distinct diseases, depending on whether the loss of one functional GRN allele induces a state of haploinsufficiency or complete loss of function when both alleles are defective. In the first case, PGRN haploinsufficiency results in a dominant form of FTLD-TDP, in the latter a recessive form of NCL. An explanation for these dramatically different manifestations of the same gene defect likely resides in the lysosomal function of PGRN and the granulins that are hypothesized to be proteolytic cleavage products of PGRN. A fraction of newly synthesized PGRN travels through the secretory pathway in a complex with PSAP, another propeptide that undergoes cleavage to mature saposins. In contrast to PGRN and its granulins, the role of saposins as cofactors for lysosomal lipid hydrolases is well established. We hypothesized that PGRN and the granulins might have similar functions in the lysosome. This hypothesis was supported by the striking lysosomal storage phenotype that occurs in the complete absence of PGRN and by the characteristic lysosomal inclusions of lamellar pseudomembranous structures (Figure 1), which are indicative of an inability of the lysosome to digest and recycle cellular membranes.

20Decreasing lysosomal capacity is now being recognized as a characteristic feature of the aging process, which is also the greatest risk factor for dementias, such as AD and FTLD. Hence, we hypothesized that the clinical manifestations of FTLD-TDP in humans might be indicative of an unmasking of a latent lysosomal deficiency caused by PGRN haploinsufficiency. Characteristic changes in the brain lipidome have been previously reported for AD. Analysis of brain lipids extracted from age-matched and autopsy-confirmed control, AD, non-GRN FTLD-TDP, and GRN FTLD-TDP human subjects reproduced and extended these findings (Figures 2 and S1). Intriguingly, using LDA as an analytical tool, none of the distinct pathologic or control lipidomes overlapped with each other, indicating that normal aging and different age-related neurodegenerative disorders differentially affect cellular lipid metabolism and homeostasis. More specifically, the lipid changes seen in GRN FTLD-TDP human subjects closely resembled the lipidome of Grn+/− and Grn−/− mice, with preferential changes in TAGs, DAGs, and PSs (Figures 2 and S1). This was largely recapitulated in the whole-cell lipidome from Grn mutant MEFs, with a preferential, gene-dosage-dependent accumulation of CEs and TAGs and a corresponding decrease of DAGs and PSs of increasing chain length and unsaturated bonds (Figure 3). Similar results were obtained when we analyzed lysosomes enriched from wild-type and Grn mutant mouse livers (Figure 4A), which further revealed a striking approximate 3-fold decrease in the PS/total lipid ratio (Figure 4B). In addition, Grn−/− ver lysosomes also showed a significant increase in both unprocessed and processed cathepsin D, consistent with a compensatory response of the lysosome to a PGRN or granulin-deficiency-induced malfunction (Figure S3). Taken together, these findings indicate a PGRN-or granulin-dependent mechanism in the transfer of long, unsaturated fatty acids from CEs and TAGs to PSs and PEs and to a potential role in the function of cathepsins. Homozygous deficiencies of cathepsin D and cathepsin F independently cause their own genetically distinct forms of NCL, indicating that normal functioning of lysosomal hydrolases is essential for preventing NCL-type storage diseases.

21Our unbiased comprehensive lipidomic analysis was complemented by a transcriptomic survey of mRNA extracted from Grn+/+, Grn+/− and Grn−/− mouse brains. Analysis of the changes in lysosomal and immune-related gene expression by RNA sequencing (RNA-seq) (Figure 5) yielded results that reproduced those previously published using a technically different microarray-based approach. Moreover, our RNA-seq data allowed us to quantitatively compare transcriptomic changes in Grn deficiency with a previously reported dataset obtained in Npc1 knockout (Npc1−/−) mouse brains (Figures 5B–5F), revealing selective changes in the expression of lysosomal, immune-related, and lipid-related genes between NPC1 and PGRN deficiency states. Despite a state of severe lysosomal dysfunction that is common between Npc1 and Grn mutant mice, there was a highly selective subset of transcripts in all three categories that were significantly altered in Grn−/− mice, but not in Npc1−/− mice (Figures 5D-5F). These selective differences point toward pathology-specific mechanisms by which dysfunctional lysosomes communicate with the cellular gene expression machinery. A possible mechanism might involve the differential complement of cellular lipids that accumulate as a result of the different biochemical defects, leading to the activation or suppression of lipid-sensitive nuclear hormone receptors that regulate the transcription of numerous metabolic genes, including those involved in lipid metabolism. These include a group of lipid hydrolases that exhibited significant differential regulation in Grn deficiency (Figure S4), suggesting roles in metabolic pathways that depend on the function of PGRN or of individual granulins.

22In summary, we present here a comprehensive set of lipidomic and transcriptomic data that support a role for PGRN and granulins in the metabolism of lipids and proteases within the lysosome. We propose that neurodegenerative diseases due to PGRN dysfunction (FTLD-TDP with PGRN haploinsufficiency) involve latent lysosomal dysfunction. This metabolic defect consequently exacerbates the progressive reduction of lysosomal and autophagic capacity during normal human aging and becomes the driver for the neuronal loss that culminates in FTLD-TDP.

23The generation of Grn mutant (Grn+/− and Grn−/−) mice has been previously described. See Supplemental Experimental Procedures for more detail. All procedures were performed in accordance with the protocols approved by the Institutional Animal Care and Use Committee of the University of Texas Southwestern Medical Center.

24Total RNA was extracted from the brains of female Grn+/+, Grn+/−, and Grn−/− mice (n = 2; average age = 200 ± 23 days) with Trizol (Ambion). mRNA libraries were prepared from each RNA sample and sequenced via Illumina sequencer. High-quality reads were mapped to the genome and analyzed to determine differentially expressed transcripts between genotypes. See Supplemental Experimental Procedures for more detail.

25Frozen pieces of fresh brain tissue were isolated from Grn+/+, Grn+/−, and Grn−/− mice and from autopsy-confirmed cases (Table S1) of AD, FTLD-TDP with and without GRN hemizygous loss (GRN FTLD-TDP and non-GRN FTLD-TDP, respectively), and controls with only age-related neuropathologic changes (control) and were prepared for lipidomic MS/MSALL analysis. MEFs and hepatic lysosomes were harvested and prepared for lipidomic MS/MSALL analysis. See Supplemental Experimental Procedures for more detail.

26The sample size, number of replicates, and statistical tests are indicated either in the text, figure legends, and/or Experimental Procedures. No data were excluded from analysis, and any and all outliers encountered within the data were included in the statistical analysis. Data generated from the transcriptomic analysis and high-throughput lipidomic analysis has been provided as supplemental tables and Excel files.

27The accession number for the RNA-seq data reported in this paper is SRA: SRP114906.