Selectivity and Kinetic Requirements of HDAC Inhibitors as Progranulin Enhancers for Treating Frontotemporal Dementia

Summary

Frontotemporal dementia (FTD) arises from neurodegeneration in the frontal, insular, and anterior temporal lobes. Autosomal dominant causes of FTD include heterozygous mutations in the GRN gene causing haploinsufficiency of progranulin (PGRN) protein. Recently, histone deacetylase (HDAC) inhibitors have been identified as enhancers of PGRN expression, although the mechanisms through which GRN is epigenetically regulated remain poorly understood. Using a chemogenomic toolkit, including optoepigenetic probes, we show that inhibition of Class I HDACs is sufficient to upregulate PGRN in human neurons and only inhibitors with apparent fast binding to their target HDAC complexes are capable of enhancing PGRN expression. Moreover, we identify regions in the GRN promoter in which elevated H3K27 acetylation and transcription factor EB (TFEB) occupancy correlate with HDAC-inhibitor mediated upregulation of PGRN. These findings have implications for epigenetic and cis-regulatory mechanisms controlling human GRN expression and may advance translational efforts to develop targeted therapeutics for treating PGRN-deficient FTD.

eTOC blurb

She et al. show that inhibition of Class I HDACs is sufficient to upregulate progranulin (PGRN) in human neurons and only fast-binding HDAC inhibitors are capable of enhancing PGRN expression. These findings may advance translational efforts to develop targeted therapeutics for treating PGRN-deficient diseases.

Introduction

Frontotemporal dementia (FTD) spectrum disorders represent the second most common form of presenile dementia after Alzheimer’s disease, accounting for 5–15% of all cases of dementia in individuals 45–65 years of age (Rademakers et al., 2012). Clinical phenotypes of FTD are heterogeneous, including subtypes associated with changes in personality and behavior, and deterioration of language or movement, and arise from frontotemporal lobar degeneration (FTLD), a family of neurodegenrative pathologies with a predeliction for the frontal, insular, and anterior temporal lobes (Haass and Neumann, 2016). The past decade of basic FTD research has been instrumental in the characterization of the biological and genetic underpinnings of the disease. However, the specific pathways involved in FTLD etiology are only beginning to be understood, and further elucidation of FTLD pathophysiology is necessary for the development of efficacious treatments for these devestating diseases.

One of the known autosomal dominant forms of FTD is caused by mutations in the GRN gene on Chr17 encoding the multifunctional secreted glycoprotein progranulin (PGRN) (van Swieten and Heutink, 2008). Mutations in one copy of the gene lead to PGRN haploinsufficiency and associated neurodegeneration with characteristic accumulation of ubiquitin and TAR DNA-binding protein (TDP)-43 (FTLD-TDP) (van Swieten and Heutink, 2008). Since PGRN-deficient FTD is a disease of haploinsufficiency, one course of therapeutic action may be to increase PGRN protein levels by upregulating GRN expression from the remaining wild-type allele allowing restoration of total PGRN levels. Consistent with this notion, it has been shown that exogenous PGRN can rescue the wild type neurite outgrowth phenotype in Grn^−/− mouse primary neural cultures, and increasing GRN expression has proven to be beneficial in several animal models (Gass et al., 2012a, Kleinberger et al., 2013). It follows that increasing PGRN levels in human neurons may help to restore a wild type phenotype, and may be able to delay disease pathogenesis and neurodegeneration.

HDAC inhibitors in particular the multi-HDAC targeting compound SAHA (Vorinostat), have been shown previously to increase GRN transcription in a mouse Neuro-2A reporter cell line, human lymphoblastoid cell lines from a healthy control and an FTD subject with a nonsense GRN mutation, and in human SH-SY5Y neuroblastoma cells (Cenik et al., 2011, FORUM Pharmaceuticals Inc, 2015, Alquezar et al., 2015). Crebinostat, another potent, multi-HDAC targeting compound, has also been shown to enhance Grn mRNA expression in mouse cortical neurons (Fass et al., 2013). More recently, SAHA has been demonstrated to enhance PGRN protein expression in human induced pluripotent stem cell (iPSC)-derived cortical neurons, albeit at doses that caused altered regulation of many other genes (Almeida et al., 2016). These observations regarding the epigenetic regulation of GRN has led to interest in using HDAC inhibitors as a rational therapeutic approach to treat PGRN-deficient FTD (FORUM Pharmaceuticals Inc, 2015). However, since SAHA is a relatively non-selective HDAC inhibitor targeting Class I (HDAC1/2/3,8), IIb (HDAC6/10) and IV (HDAC11) HDAC isoforms, the precise molecular mechanisms behind PGRN regulation in human neurons have yet to be elucidated (Bradner et al., 2010, Bantscheff et al., 2011). In particular, it is unknown if SAHA, or other HDAC inhibitors, directly affect histone acetylation within the GRN locus in a manner related to transcriptional changes, which members of the Zn²⁺-dependent family of HDACs are the relevant targets of inhibition, or whether engagement of other targets and mechanisms are critical for the enhancing PGRN production. Answering these questions in the context of human neuronal HDAC complexes is essential to advance the development of next-generation HDAC inhibitors as a targeted therapeutic for PGRN-deficient FTD and overcome potential limitations of first-generation HDAC inhibitors that lack isoform and functional selectivity like SAHA.

Here, in a human iPSC-based neuronal culture system with robust and scalable mRNA and protein level assays, we use a chemogenomic strategy to systematically dissect the roles of different HDAC isoforms in GRN mRNA and PGRN regulation. Our data provide strong evidence to support the conclusion that inhibition of Class I HDACs pharmacologically is sufficient to enhance PGRN protein production in human neural progenitor cells (NPCs) and differentiated neurons. In addition, only HDAC inhibitors with apparent fast-on binding to their target HDACs are capable of potently enhancing GRN mRNA expression and PGRN protein secretion, whereas HDAC inhibitors that exhibit slow-on binding kinetics are largely inactive despite measurable effects on histone acetylation and other gene expression. Importantly, in terms of mechanism, we also link these observed changes in GRN mRNA induction to changes in specific histone acetylation sites located within the promoter proximal region of GRN locus, as well as implicate a master regulator of lysosomal and autophagy gene expression, transcription factor EB (TFEB), in this transcriptional response.

Results

Epigenetic probes for GRN mRNA and PGRN protein expression in human NPCs and neurons

Previously, SAHA and crebinostat, HDAC inhibitors known to target Class I and Class IIb HDACs, were found to increase GRN gene expression in proliferative mouse Neuro-2A cells and in post-mitotic, mouse primary neurons (Cenik et al., 2011, Fass et al., 2013). We initially sought to confirm these observation in cultured human iPSC-derived NPCs and in post-mitotic, differentiated neuronal cultures using assays that reported on GRN mRNA levels, intracellular PGRN protein, and extracellular secreted PGRN protein (Figure 1A). Treatment of these NPC and neuronal cultures with SAHA and crebinostat caused a robust and consistent induction of both GRN mRNA and PGRN intracellular and secreted protein levels (Figure 1B–D). In addition, we demonstrated that treatment with panobinostat (LBH-589), also a highly-potent, cinnamic hydroxamic acid, broad spectrum, HDAC inhibitor (George et al., 2005), likewise significantly increased GRN mRNA and PGRN protein levels in NPCs and differentiated neurons (Figure 1E).

Figure 1. Profiling hydroxamic acid HDAC inhibitor effects on GRN mRNA and PGRN protein expression in human iPSC-derived NPCs and neurons.

A) Model and experimental scheme for measuring the HDAC inhibitor effect on progranulin expression, from mRNA (Assay 1 – qPCR) to intracellular protein (Assay 2 – Western Blot and ELISA) to secreted protein (Assay 3 – ELISA). With the HDAC complex, S_n = subunit and TF = transcription factor. B) Chemical structures for fast on/off HDAC inhibitors: (1) SAHA, (2) crebinostat, and (3) panobinostat (Herman et al., 2006). IC₅₀ values for (1) from (Bradner et al., 2010), (2) from (Fass et al., 2013), and (3) from (Khan et al., 2008). All increase GRN gene expression and PGRN protein levels in human NPCs (black) and 18-day neurons (gray) after 24 hours as shown in C (SAHA, 10 μM), D (Crebinostat, 2.5 μM), and E (Panobinostat, 0.5 μM); vehicle = DMSO. Dividing lines on Western blots indicate samples from separate gels. Protein quantification was measured by ELISA. Each condition shown is the mean + SEM of n=3 biological replicates × 3 technical replicates with significance calculated by unpaired t-test. **** p < 0.0001

While these broad specificity-HDAC inhibitors increase PGRN levels upon treatment, the simultaneous inhibition of multiple members of the HDAC family of enzymes is known to have significant effects on many genes at the same time (Almeida et al., 2016). Understanding the inhibition of which HDAC isoforms are necessary and sufficient for upregulation of PGRN will help answer questions about the suitability of HDAC inhibitors as viable therapeutics for disorders of PGRN-deficiency. To this end, we first tested the Class IIb HDAC6-selective inhibitors, ACY-1215 (ricolinostat) and tubastatin A, and a Class I HDAC8-selective inhibitor, PCI-34051 in human NPCs and neurons at a range of concentrations previously reported to be effective in the literature (Iaconelli et al., 2015, Butler et al., 2010, Balasubramanian et al., 2008). Neither tubastatin A nor PCI-34051 increased GRN mRNA or PGRN protein expression (Figure 2), although tubastatin A had a robust effect on H3K9 acetylation (Figure S1). We also tested a broadly acting Class IIa HDAC inhibitor TMP269 (Lobera et al., 2013), and found that it was not effective at increasing GRN mRNA in NPCs (Figure S2). These data suggest that inhibition of HDAC6, HDAC8, and Class IIa HDACs are not sufficient to increase GRN mRNA or PGRN protein expression in a human neuronal context, which is consistent with data from mouse neuronal cultures for HDAC6 inhibitor treatment (Cenik et al., 2011).

Figure 2. HDAC6 and HDAC8-selective inhibitor effects on GRN mRNA and PGRN protein expression in human iPSC-derived NPCs and neurons.

A) Chemical structures for HDAC6-selective inhibitors (4) ACY-1215 and (5) Tubastatin A and HDAC8-selective inhibitor (6) PCI-34051. IC₅₀ values for (4) from (Santo et al., 2012) (5) from (Butler et al., 2010), and (6) from (Balasubramanian et al., 2008). B) ACY-1215 (5 μM) increases GRN and PGRN expression in human NPCs (black) and 18-day neurons (gray) after 24 hours. Tubastatin A (10 μM) was not effective. C) PCI-34051 (10 μM) does not increase GRN and PGRN expression in human NPCs (black) or 18-day neurons (gray) after 24 hours. Vehicle = DMSO. Protein quantification was measured by ELISA. Each condition shown is the mean + SEM of n=3 biological replicates × 3 technical replicates with significance calculated by unpaired t-test. **** p < 0.0001

Although tubastatin A did not increase GRN mRNA or PGRN protein expression, ACY-1215, the additional reported HDAC6-selective inhibitor, significantly increased both GRN mRNA and PGRN protein levels at a dose optimized for maximal PGRN enhancement activity (Figure 2). It may be that ACY-1215 at cellular concentrations of 5 μM has lost its selectivity for HDAC6 since its IC₅₀ for HDAC6 is 5 nM, which is only ~10× selective over HDAC 1, 2, 3 based upon in vitro potencies (Santo et al., 2012). Consistent with this notion, ACY-1215 treatment at doses that were effective for increasing PGRN protein levels (5 μM) caused as strong of an induction of H3K9 acetylation as SAHA and panobinostat that also potently inhibit Class I HDACs (Figure S1).

Class I-selective HDAC inhibitors enhance PGRN expression

To test whether Class I HDAC inhibition is sufficient to enhance PGRN expression, we tested the efficacy of apicidin, a cyclic tetrapeptide natural product that does not contain the strong metal-chelating hydroxamate group and found that apicidin potently enhanced both GRN mRNA and PGRN protein expression (Figure 3). These data decouple the hydroxamic acid moiety from being necessary for efficacy in increasing PGRN levels in human neurons and provide evidence that targeting Class I HDACs is sufficient to increase GRN mRNA and intracellular and secreted PGRN protein.

Figure 3. Apicidin and Valproate increase GRN mRNA and PGRN protein expression in human iPSC-derived NPCs and neurons.

A) Chemical structure for tetrapeptide (7) Apicidin, IC50 values from (Huber et al., 2011). B) Apicidin (2.5 μM) and positive control SAHA (10 μM) increase GRN mRNA and PGRN intracellular and secreted protein expression human NPCs (black) and 18-day neurons (gray) compared to vehicle (DMSO). C) Chemical structure for the carboxylic acid (8) valproate. IC₅₀ values from (Fass et al., 2011). D) Valproate (5 mM) significantly increases GRN mRNA and PGRN intracellular and secreted protein expression human NPCs (black) and 18-day neurons (gray) compared to vehicle (ddH₂O). Protein quantification was measured by ELISA. Each condition shown is the mean + SEM of n=3 biological replicates × 3 technical replicates with significance calculated by unpaired t-test. * p < 0.05 **** p < 0.0001

We also tested the efficacy of valproate (valproic acid), an FDA-approved drug used for the treatment of epilepsy and bipolar disorder that in earlier studies was shown to exhibit Class I HDAC-selectivity (Fass et al., 2011). As predicted based upon its selectivity, in human NPCs and neurons, valproate indeed enhanced both GRN mRNA and PGRN protein expression, albeit at much higher dose (1–5 mM) than any of the hydroxamates or apicidin (Figure 3). However, these efficacious doses correlate with previously reported values in mouse neuronal assays, and correspond to the relative lack of potency of carboxylic acids compared to hydroxamic acids in HDAC biochemical assays (Fass et al., 2011, Cenik et al., 2011).

Class I, slow-binding HDAC inhibitors, do not increase PGRN expression in human neuronal cultures

To further probe the contribution of different HDAC isoforms on PGRN epxression, we tested the effects of Class I-selective, ortho-aminoanilide HDAC inhibitors, CI-994 (HDAC1/2/3) and Cpd-60 (HDAC1/2) on human NPCs and neurons. Surprisingly, given their biochemical potency and ability to significantly increase H3K9 acetylation levels in human NPCs and neurons, neither CI-994 nor Cpd-60 significantly increased GRN mRNA or PGRN levels (Figure 4; see also Figure S1). Rather, following treatment with CI-994, in human neurons, there is a statistically significant decrease of GRN mRNA and intracellular PGRN protein, and in NPCs a decrease of intracellular PGRN and secreted PGRN (Figure 4; see also Figure S1). To ensure that this lack of enhancement by CI-994 was not unique to human neuronal cells, we also repeated these assays in a human, non-neuronal cell line, MSR293, a derivative of the HEK293 cell line. In MSR293 cells, treatment with CI-994 was also ineffective at upregulating GRN mRNA or PGRN, whereas SAHA was effective (Figure S3). In contrast, both CI-994 and Cpd-60 significantly increased FXN (frataxin) mRNA levels (Figure 4E; see also Figure S4), a gene that has been shown to be epigenetically enhanced by other ortho-aminoanilide HDAC inhibitors in human cell lines (Chou et al., 2008, Soragni et al., 2015).

Figure 4. Profiling ortho-aminoanilide HDAC inhibitors on GRN mRNA and PGRN protein expression in human iPSC-derived NPCs and neurons.

A) Chemical structures for slow-on/off ortho-aminoanilide HDAC inhibitors (9) CI-994 and (10) Cpd-60. IC₅₀ values for (9) from (Bradner et al., 2010) and (10) from (Schroeder et al., 2013). B) Neither CI-994 nor Cpd-60 had a significant effect on GRN mRNA expression or PGRN protein expression in human NPCs (black) or 18-day neurons (gray). C) Despite longer treatment times, CI-994 did not increase GRN mRNA and in fact, decreased PGRN secreted protein expression in NPCs compared to vehicle (see p-values below); 2 biological × 3 technical replicates. D) CI-994 and Cpd-60 increase H3K9 acetylation in 18-day neurons despite not enhancing PGRN protein levels (see also Figure S1). E) Slow-on/off HDAC inhibitors CI-994 and Cpd-60 increase Frataxin (FXN) mRNA expression in 18-day neurons, while fast on/off inhibitor SAHA does not (see also Figure S3). Cells were treated for 24 hours, unless otherwise indicated, with vehicle (DMSO), SAHA (10 μM), CI-994 (10 μM) or Cpd-60 (5 μM). Protein quantification was measured by ELISA. All graphs are shown as mean + SEM, with significance relative to vehicle calculated by unpaired t-test. * p < 0.05 ** p < 0.01 **** p < 0.0001

Besides varying in their isoform selectivity from the hydroxamates, CI-994 and Cpd-60 also differ in terms of their binding kinetics (Bantscheff et al., 2011, Lauffer et al., 2013, Wagner et al., 2016, Chou et al., 2008). Whereas hydroxamates have a fast-on/fast-off mechanism of action, CI-994 and Cpd-60 have a slow-on/slow-off mechanism of action (Schroeder et al., 2013, Chou et al., 2008). This slow-on/slow-off mechanism of action has been shown to be essential for induction of the FXN gene, with the hypothesis that residence time on the HDAC is key for this induction (Herman et al., 2006, Chou et al., 2008). Taking these observations into consideration, we treated our human NPCs with CI-994 for up to 72 hours to account for slow binding properties and still found that CI-994 did not significnatly enhance GRN mRNA or PGRN protein levels in neuronal cells (Figure 4). These results indicate that simply extending the treatment time for slow-binding HDAC inhibitors is not sufficient to enhance GRN mRNA or PGRN protein levels.

Using recombinant human Class I HDACs in an enzymatic assay measuring deacetylase activity, we determined that apicidin and valproate, which both increase PGRN expression, have an in vitro kinetic profile simliar to panobinostat; their potency did not increase with prolonged incubation, unlike the slow-binder CI-994 (Figure S5). This indicates that binding kinetics and not merely residence time may play a crucial role in HDAC inhibitor-mediated GRN enhancement.

Optoepigenetic control of PGRN expression in human neuronal cultures with a photoswitchable HDAC inhibitor

To further test our kinetic model for HDAC inhibitor regulation of PGRN and dissect the relative contribution of the on- and off-rates of binding to PGRN enhancement, we turned to our recently developed, optically controlled HDAC inhibitors (Reis et al., 2016). This series of photoswitchable compounds, or COMET (chemo-optical modulation of epigenetically regulated transcription) probes, has been shown to bind to and competitively inhibit the deacetylase activity of HDAC targets only upon a light-induced trans-to-cis isomerization, with target engagement and binding rates necessarily faster than the microsecond cis-to-trans relaxation rate (Reis et al., 2016). Once bound to the active site of an HDAC, the compounds are locked into a stable cis conformation and have been shown to have long residence times after binding (Reis et al., 2016). Here, we treated NPCs with BG47, a prototypical HDAC1/2 selective, optoepigenetic probe similar in structure to Cpd-60 but possessing an azobenzene scaffold and an ortho-hydroxyanilide chelator (Figure 5). BG47 treatment for 16 hours elevated H3K9 acetylation only in the presence of light (Figure S6) and also significantly upregulated PGRN only in the presence of light (Figure 5). The apparent fast-on/slow-off kinetic profile of BG47 coupled with its ability to upregulate PGRN suggests that the apparent rate of binding rather than inhibitor-HDAC residence time is key to HDAC inhibitor-mediated modulation of PGRN.

Figure 5. Optoepigenetic probe BG47 upregulates PGRN in the presence of light.

A) Activation schematic for optoepigenetic probe (11) BG47. BG47 is activated to its active cis-isomer by 470 nm light and engages target within μs, whereupon it stays bound for hours. B) BG47 (10 μM) enhanced PGRN in NPCs after 16 hours of treatment with light (25 ms on/75 ms off). This effect is not seen without light. 4 biological replicates. Graphs are shown as mean + SEM with significance calculated by unpaired t-test. **** p < 0.0001 C) Model for HDAC inhibitor-mediated GRN enhancement based on kinetics of binding of the HDAC inhibitor. If the on-rate of the HDAC inhibitor is slower than the HDAC complex-chromatin association rate, then the HDAC inhibitor will be unable to upregulate GRN.

Relationship between GRN mRNA and PGRN protein level in human NPCs and neurons

Having tested HDAC inhibitors with a range of efficacies toward enhancing GRN expression, we sought to determine the correlation between changes in GRN mRNA levels and levels of intracellular and secreted PGRN protein. We found that GRN mRNA levels were highly positively correlated with both PGRN protein expression and secretion (Figure 6). These correlations were similar in both proliferative NPCs and post-mitotic neurons. Taken together, these data suggest that multiple human neuronal cell types have a conserved mechanism of epigenetic regulation involving sensitivity to Class I HDAC inhibitors with fast-binding kinetics.

Figure 6. Correlation between effects of HDAC inhibitors on GRN mRNA expression, PGRN intracellular protein expression, and PGRN secreted protein in A) NPCs and B).

18-day differentiated neurons, with representative compounds CI-994 and panobinostat labeled. Data show a positive correlation between mRNA expression, protein expression, and protein secretion with correlation coefficient, r, and p-value enumerated above.

Comparison of RNAi-mediated targeting of individual HDAC isoforms to small molecule HDAC inhibitors

To further probe the HDAC inhibitor selectivity requirements for HDAC inhibitor-mediated regulation of GRN mRNA and PGRN protein production, we sought to complement our chemogenomic profiling with functional genomic studies using RNA interference (RNAi). Using lentiviral-mediated delivery of short hairpin RNAs (shRNAs) to selectively target individual Class I HDACs predicted to be relevant, along with four separate control shRNAs (RFP, lacZ, luciferase, and GFP), we created a panel of stable human NPC lines designed to have selective silencing of HDACs. Upon selection and expansion of NPC lines we were able to obtain robust silencing of each of HDAC1, HDAC2, and HDAC3 compared to the four control shRNAs (Figure S7). However, RNAi silencing of none of these HDAC isoforms had a statistically significant effect on GRN mRNA or PGRN protein expression in human NPCs (Figure S7).

Mapping of epigenetic modifications of the GRN promoter upon HDAC inhibitor treatment

To further elucidate the mechanism through which small molecule HDAC inhibitors enhance GRN mRNA, we sought to address the questions of whether HDAC inhibitors directly affect histone acetylation within chromatin associated with the GRN promoter. Although we had found that global histone acetylation was not measurably different when NPCs were treated with fast- or slow-binding HDAC inhibitors (Figure S1), we hypothesized that the resolution of specific acetylation sites associated with the promoter proximal region of the GRN locus would reveal HDAC inhibitor changes that are obscured when analyzing global histone acetylation levels. We thus developed chromatin immunoprecipitation (ChIP)-qPCR assays to understand potential differential histone acetylation changes in the GRN promoter given treatment with fast- or slow-binding HDAC inhibitors. Here, we focused on the promoter-enhancer region of GRN, as computationally defined by enrichment of H3K27 acetylation marks, a histone post-translational modification associated with active transcription, according to the Human Epigenome Atlas data from human embryonic stem cell (H9)-derived NPCs and neurons, as well as the dorsolateral prefrontal cortex and inferior temporal brain regions (Figure 7A) (Zhou et al., 2011). This promoter proximal region corresponds to transcription factor binding and histone modification sites from previously generated global ChIP-seq data from the ENCODE (Encyclopedia of DNA Elements) project, which revealed that across several non-neuronal human cell lines, the promoter proximal region of the GRN locus contains an enrichment of H3K27 acetylation marks, as well as binding sites for multiple Class I HDAC complex members and HDAC-interacting proteins (Figure 7B) (Kent et al., 2002, Rosenbloom et al., 2013). We thus selected three regions of the GRN locus for which dynamic changes in H3K27 acetylation were predicted if they were to play a regulatory role in controlling active transcription.

Figure 7. ChIP-qPCR of H3K27 acetylation on GRN promoter/enhancer region. ChIP-qPCR of H3K27 acetylation on GRN promoter/enhancer region.

A) Visualization of the promoter/enhancer region of GRN in two human brain regions and human embryonic stem cell H9-derived NPCs and neurons, with measured H3K27 acetylation marks in the brain regions and imputed H3K27 acetylation in the H9-derived cell lines as compiled by the WashU EpiGenome Browser (Zhou et al., 2011). TSS = transcriptional start site. B) Promoter/enhancer region of GRN, showing H3K27 acetylation and binding sites for selected ENCODE Transcription Factors for up to 7 cell lines (indicated above). Image was modified from UCSC Genome Browser (Kent et al., 2002) and shown to scale with (A). H3K27 acetylation data is layered, so some colors may not be reflected in the legend. More information may be found on the UCSC Genome Browser. Transcription factors were selected based on binding affinity for HDAC complexes or known HDAC-modulated genes. Transcription factors are color coded in grayscale; the darkness of the box is proportional to maximum ChIP data value seen in any cell line in the region. The ChIP regions denote regions of interest for ChIP-qPCR. C) H3K27 acetylation ChIP-qPCR data in NPCs for the regions on the GRN promoter denoted in (A and B). Cells were treated with vehicle (DMSO), panobinostat (0.5 μM), or CI-994 (10 μM) for 24 hours. Each condition is shown as the mean + SEM of 3 technical replicates, with significance calculated by unpaired t-test. ** p< 0.01 *** p < 0.001

Using these ChIP-qPCR assays, we found that in NPCs, treatment with panobinostat or CI-994 differentially affected H3K27 acetylation within chromatin in the GRN promoter. Panobinostat significantly increased H3K27 acetylation in “ChIP Region 1” on the GRN promoter while CI-994 did not. Neither compound treatment increased H3K27 acetylation in “ChIP Region 2”, and both increased H3K27 acetylation in “ChIP Region 3” to varying levels. Taken together, these data suggest that the increased H3K27 acetylation in “ChIP Region 1” may play a significant role in GRN expression in human neuronal cells and may predict whether an HDAC inhibitor will be effective or not at increasing GRN mRNA levels irrespective of the effect on global histone acetylation. These data provide the first demonstration of a region of chromatin in the GRN locus in the human genome that responds to HDAC inhibition in a manner correlated with GRN mRNA induction.

The role of TFEB in HDAC-mediated GRN regulation in human neuronal cells

Besides directly affecting histone acetylation, HDACs are also known to epigenetically regulate many transcription factors, and it is possible that HDAC inhibitors with different binding kinetics differentially affect key transcription factors governing GRN transcription. For instance, the transcription factor EB (TFEB), a master regulator of autophagy-lysosomal gene expression, is implicated in GRN expression through its specific recognition and binding to E-box consensus sequences (5′-CANNTG-3′) in the GRN promoter region (Belcastro et al., 2011). Since TFEB overexpression has been shown to be sufficient to enhance GRN mRNA and PGRN protein levels in human cells (Holler et al., 2016), we hypothesized that HDAC inhibitors that affect PGRN expression would have a different effect on TFEB levels than those HDAC inhibitors that do not. Indeed, we observed that the level of this enhancement in TFEB levels correlated with an increase in PGRN in a dose-dependent manner (Figure 8). HDAC inhibitors that surpassed a ‘threshold’ of ~10-fold induction of TFEB protein levels were capable of enhancing PGRN levels (e.g. panobinostat), whereas HDAC inhibitors below this level (e.g. the slow-binder CI-994) were not (Figure 8).

Figure 8. TFEB protein expression and occupancy on the GRN promoter after HDAC inhibitor treatment.

A) TFEB, pTFEB(Ser142), and PGRN protein levels in NPCs when treated with HDAC inhibitors. Graph showing best-fit curve with 95% confidence interval (dotted line). NPCs were treated with vehicle (ddH2O for Valproate, DMSO for all others), SAHA (10 μM), Panobinostat (0.5 μM), Crebinostat (2.5 μM), ACY-1215 (5 μM), Tubastatin A (10 μM), PCI-34051 (10 μM), Apicidin (2.5 μM), CI-994 (10 μM), Cpd-60 (5 μM), or Valproate (5 mM) for 24 hours. Quantification was done with ImageJ. B) Potential TFEB binding sites on the GRN promoter/enhancer region shown relative to H3K27 acetylation in 7 cell lines from the ENCODE project and the UCSC Genome Browser (Kent et al., 2002). H3K27 acetylation data is layered, so some colors may not be reflected in the legend. More information may be found on the UCSC Genome Browser. E-box sites (5′-CANNTG′) are shown with blue lines. CLEAR motifs (5′-TCACG-3′) are shown in orange. These sites are overlaid with the ENCODE transcription factor binding regions (yellow boxes). Portions of the genome where both E-box sites and CLEAR motifs occur near known transcription factor binding sites are shown, with the start site of GRN 5′ UTR marked in bold, highlighted in purple. C) TFEB ChIP-qPCR data in NPCs for the regions on the GRN promoter denoted in (B). Cells were treated with vehicle (DMSO), panobinostat (0.5 μM), or CI-994 (10 μM) for 24 hours. Each condition is shown as the mean + SEM of 2 replicates, with significance calculated by unpaired t-test. * p< 0.05 *** p < 0.001

We next sought to determine whether upregulation of TFEB due to HDAC inhibitor treatment correlates with TFEB binding on the GRN promoter. Computational analysis revealed 45 E-box sequences on the GRN promoter region to which TFEB can potentially bind, a proportion of which overlap with other known regions of transcription factor binding (Figure 8). In addition, TFEB has been linked to the expression of members of the computationally-derived CLEAR (Coordinated Lysosomal Expression and Regulation) network, and is thought to bind to either a full CLEAR sequence (5′-TCACGTGA-3′) or a partial CLEAR motif (5′-TCACG-3′) (Palmieri et al., 2011). Although the promoter region of human GRN does not contain any full CLEAR sequences, computational analysis revealed there are three partial CLEAR motifs to which TFEB may bind, two of which occur within DNA regions known to recruit other transcription factors (Figure 8). Consistent with our predictions, performing TFEB ChIP-qPCR studies in our human NPC system showed that upregulation of GRN/PGRN correlates with increased TFEB occupancy on the GRN promoter/enhancer region (Figure 8).

Discussion

In recent years, GRN has emerged as a gene of interest in multiple neurodegenerative disorders including FTD, Alzheimer’s disease, and neuronal ceroid lipofuscinosis (NCL), as well as bipolar disorder (Cenik et al., 2012, Kittel-Schneider et al., 2014). Research into PGRN and its role in diverse aspects of CNS function with a clear causal role for haploinsufficiency in driving FTD makes it an attractive therapeutic target. Insight into GRN modulators may help in the development of therapeutics for this subclass of FTD, as well as for other neurological and psychiatric diseases.

Here, we provide evidence that one potential mechanism of increasing PGRN levels in human NPCs and neurons is through inhibition of Class I HDACs (HDAC1/2/3) and that fast-on inhibitors of these HDACs robustly increase GRN mRNA, intracellular PGRN protein, and secreted PGRN protein. Concentrations of compounds used in our studies were optimized for maximal effect in enhancing progranulin expression, capitalizing on the different potencies and selectivity of different compounds within the chemogenomic toolkit for dissecting Zn²⁺-dependent HDACs. By directly targeting gene expression in this way, we expected and showed not only an increase in gene expression but, correlatively, in protein expression and action (Figure 6).

Contrary to our initial expectations, our functional genomic studies using RNAi-mediated silencing of HDAC1, HDAC2, and HDAC3 failed to provide evidence that loss-of-function of any one of these Class I HDACs is sufficient to enhance PGRN expression (Figure S7). These results suggest that either multiple Class I HDACs need to be inhibited simultaneously or that the loss of HDAC protein from RNAi-mediated silencing does not faithfully recapitulate the same functional consequence of small molecule inhibition. In support of the latter, it is well known that HDACs are part of large, macromolecular complexes that may include other HDAC family members (Bantscheff et al., 2011, Millard et al., 2017). The loss of one component of these complexes may impact epigenetic regulation differently compared to substrate competitive inhibitors within the HDAC active site that otherwise leave the complex intact. Furthermore, we noted evidence for compensatory effects from single HDAC knockdown in the form of enhanced expression of other Class I HDACs, in particular, elevated HDAC1, and to a lesser extent HDAC3, levels upon HDAC2 silencing (Figure S7A). Thus, we conclude that rebalancing of HDAC activity due to loss of a given HDAC isoform in human NPCs may confound the straightforward comparison of single HDAC genetic perturbations to small molecule inhibitor treatments.

Since our functional genomic methods did not further narrow down a single targetable Class I HDAC isoform, further studies must be performed to determine whether small molecule inhibition of a single, or subset of, Class I HDAC isoforms is sufficient for increasing GRN/PGRN expression. The current classes of inhibitors that target a subset of Class I HDACs are ortho-aminoanilides, which we show here in the case of CI-994 and Cpd-60, are unable to significantly increase GRN expression in human iPSC-derived NPCs and neurons. Importantly, our findings demonstrate it is possible to decouple GRN gene expression from global effects on genome-wide acetylation level with HDAC inhibitors. Future studies with an expanded set of HDAC inhibitors, in particular, additional ortho-aminoanilide HDAC inhibitors with varying binding kinetics, will be important to determine whether the conclusions we draw from both CI-994 and Cpd-60 hold for additional members of this chemotype. It will also be important to address potential species-specific differences that may exist between epigenetic regulation in human neuronal cells and mouse neuronal cells.

In light of our observations regarding the effectiveness of various types of HDAC inhibitors, it is of interest to understand why, in human iPSC-derived neuronal cultures, fast binding Class I HDAC inhibitors preferentially increase GRN/PGRN expression compared to compounds with slow-on kinetics. Targeted ChIP studies revealed differential H3K27 acetylation enhancement on a specific region of the GRN promoter when treated with a fast-on/off vs. slow-on/off binding HDAC inhibitor. Since our studies with photoswitchable HDAC inhibitors suggest that HDAC inhibitor on-rate is key for GRN enhancement, we propose a regulatory model in which the binding kinetics of the HDAC inhibitor must be compatible with the kinetics of chromatin and protein complex rearrangement at the GRN promoter for the HDAC inhibitor to regulate GRN/PGRN expression (Figure 5C). Specifically, the apparent HDAC inhibitor binding rate must be faster than the apparent HDAC-chromatin association rate at the GRN promoter to be able to affect acetylation and GRN regulation. When chromatin is rearranging faster than a slow-binding HDAC inhibitor like CI-994, the compound will be unable to upregulate GRN. In addition, since the H3K27 acetylation patterns in this region differ even in different human cell types analyzed in the ENCODE genome wide ChIP studies (Rosenbloom et al., 2013) (Figure 7), these results stress the importance of understanding species-, cell type-, and gene-specific small molecule-mediated changes when trying to understand the role of epigenetic mechanisms in controlling GRN gene expression.

Our mechanistic studies comparing fast-on vs. slow-on binding HDAC inhibitors also revealed that upregulation of TFEB with GRN inducing HDAC inhibitors was correlated with increased TFEB occupancy on the GRN promoter (Figure 8). This is the first time that TFEB has been shown definitively to bind to the GRN promoter in a human neuronal context and suggests both that differential TFEB induction by different classes of HDAC inhibitors may also contribute to the difference between HDAC inhibitor chemotypes on GRN expression through its cis-regulatory functions and that increased H3K27 acetylation on the GRN promoter may promote TFEB binding to the GRN promoter.

In the context of translational efforts for PGRN-deficient FTD, the finding that the FDA-approved drug valproate upregulates PGRN may be important in the consideration of next-generation therapeutics for the disease. With chronic treatment, valproate is known to reach serum concentration of 300–700 μM (Kalu, 2010) with the brain concentration estimated to have a C_max = 400 μM (Goldenberg, 2010). Although in our studies valproate at doses lower than 1 mM only increased PGRN protein expression by 1.3–1.5× compared to vehicle in cultured NPCs and neurons (Figure S8), it is possible that sustained treatment (i.e. weeks to months) with valproate may cumulatively have a beneficial effect on PGRN levels. Additionally, it may be feasible to find other agents that synergize or further enhance the effects of valproate toward PGRN while minimizing undesired effects.

Major unanswered questions for therapeutic efforts aiming to restore PGRN expression to wildtype in PGRN-deficient FTD patients include: 1) what levels of PGRN protein are sufficient to rescue neurodegeneration; 2) when in the course of disease would PGRN levels need to be increased to modify disease progression; and 3) what is/are the optimal cell types for restoring PGRN expression in the CNS. Given that GRN is expressed in many cell types, including brain cells such as glia and microglia, the cumulative effect of increasing progranulin intracellular and secreted protein in the human body is currently unknown (Gass et al., 2012b). It is also possible that the selectivity patterns are different in different cell types of the human body. Future studies would benefit from profiling a similar set of HDAC modulators in other cell types in which PGRN is expressed, most notably microglia where GRN mRNA levels are higher than other cell types (Martens et al., 2012). The potential for cell-type specific epigenetic regulation of GRN expression would open up new avenues for treating FTD due to PGRN deficiency possibly at different stages of the disease process and through potentially synergistic combinations.

In addition, one caveat of our studies is the inability to comprehensively detect all PGRN proteins species, in particular cleaved granulins, due to limited availability of antibodies. Future testing would benefit from enhanced coverage of individual or combinations of PGRN domains. Our study was also focused on regulation of GRN mRNA and PGRN protein expression in control NPCs and neurons and further work must also be done to assess the degree to which Class I selective, fast-binding HDAC inhibitors increase PGRN in patient-derived neuronal cell lines where the levels of PGRN expression are already low. There is evidence that SAHA increases GRN in FTD patient fibroblasts (Cenik et al., 2011) and FTD-patient neuronal cell lines (Almeida et al., 2016). Because many patients with PGRN-deficient FTD exhibit accumulation of TDP-43 within neurons (van Swieten and Heutink, 2008), future studies can be directed towards understanding whether HDAC inhibitors can help to alleviate these phenotypes in cultured patient-derived NPCs and neurons and ultimately, in patients.

Significance

In the case of PGRN-deficient FTD, enhancement of GRN gene expression through HDAC inhibition has been proposed as a disease-modifying strategy to rescue haploinsufficient disease phenotype. However, the mechanisms behind which the GRN gene is epigenetically regulated in different neural cell types and the requirement for inhibition of specific HDAC family members has remained poorly understood. Here, in human iPSC-derived neuronal cultures, we used a chemogenomic toolkit consiting of a set of HDAC inhibitors with varying specificity, potency, and binding kinetics to show that pharmacological inhibition of Class I HDACs is sufficient to enhance PGRN protein production in human neurons. Moreover, we show that only HDAC inhibitors with an apparent fast association on their target HDACs are capable of potently enhancing GRN mRNA expression and PGRN protein secretion in human iPSC derived neuronal cells. The panel of human neuronal assays described here that enable tracking enhancement of GRN mRNA and PGRN protein increases, along with ChIP assays for key acetylated lysine modifications in histones and TFEB binding sites located in the GRN promoter proximal region that correlated with GRN mRNA levels, holds promise for profiling next-generation HDAC and other epigenetic target modulators to discover novel therapeutics for treating PGRN-deficient FTD and other CNS disorders caused by reduced PGRN levels. Overall, our results demonstrate the potential for using human iPSC-derived neuronal cells as an ex vivo platform for chemogenomic studies and systematic validation of next-generation neurotherapeutic targets.

STAR Methods

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Stephen J. Haggarty (shaggarty@mgh.harvard.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Human iPSC-Derived Neural Progenitor Cell Culture

All neural progenitor cells (NPCs) used were obtained from the previously reprogrammed, iPSC derived from a clinically unaffected, human fibroblast cell line GM08330 (Coriell Institute for Medical Research, Camden, NJ) as described in (Sheridan et al., 2011). These NPCs were cultured on plastic tissue culture ware coated first with poly-ornithine (Sigma Aldrich #P3655) in H₂O for two hours, then laminin (Sigma Aldrich #L2020) in DPBS (Gibco # 14190-144) at 37°C overnight or 4°C for at least 48 hours. Med ia for NPC culture (NPC media) was composed of 70% DMEM (Dulbecco’s modified Eagle’s Medium, Gibco #11995), 30% Ham’s F12 with L-glutamine (Modified Cellgro/Mediatech #10-080-CV), 1X penicillin/streptomycin, and 1X B27 Supplement (50X, Gibco #17504-044). NPC media was then supplemented with EGF (20 ng/mL, Epidermal Growth Factor, Sigma Aldrich #E9644, prepared as 1000X stock in DMEM), bFGF (20 ng/mL, basic Fibroblast Growth Factor, ReproCELL #03-0002, prepared as 1000X stock in PBS), and heparin (5 μg/mL, Sigma Aldrich #H3149, prepared as 1000X stock in Ham’s F12) just before use.

NPCs were maintained in complete NPC media at 37 °C with 5% CO₂ in a humidified atmosphere and passaged twice per week at a 1:3 ratio or 4 × 10⁶ cells per T75 flask. For passaging, confluent cultures in T75 flasks were washed once with 10 mL PBS and then treated with 1 mL TrypLE Select (Life Technologies #12563029) until cells detached. TrypLE treatment was stopped by adding 9 mL NPC media. Cells were gently triturated to obtain a single-cell suspension and were centrifuged at 1000 rpm (700 × g) for 5 min and then resuspended in complete NPC media.

Human iPSC-Derived Neuron Culture

All human iPSC-derived neurons were derived from above stock of NPCs by growth factor withdrawal. The 8330-8 NPCs were grown on plastic tissue culture ware in 6-well plates (Falcon #353046) that were coated with a combination of 20 μg/mL polyornithine (Sigma Aldrich #P3655) and 5 μg/mL laminin (Sigma Aldrich #L2020) in DPBS. The resulting plates were stored at 37°C overnight or at 4°C for at least 48 hours b efore use and were washed with 1 mL DPBS before cells were plated. NPCs were plated at 0.5×10⁶ cells per well in a poly-ornithine/laminin single coated 6-well plate in 2 mL NPC media with above growth factors. The cells were allowed to incubate at 37°C for 48–72 hours to grow to confluency. The media was aspirated and the cells were incubated in NPC media with no growth factors for desired amount of time, with media changes every 3–4 days.

Human MSR293 Cell Culture

Human MSR293 cells were cultured as described by commercial protocols (ThermoScientific #R79507). MSR293 medium was as follows: 90% DMEM (Dulbecco’s modified Eagle’s Medium, Gibco #11995), 10% FBS (HyClone, GE Healthcare Life Sciences #SH30910.03), 1X penicillin/streptomycin. For treatment, cells were grown in uncoated 6-well plates (Falcon #353046) with application of Geneticin at 0.8 mg/mL (ThermoScientific #10131027) upon plating.

METHOD DETAILS

Compound Preparation of HDAC inhibitors

Sources of compounds are noted in Key Resources Table. Stock concentrations of all compounds were made at 1000× in DMSO (Sigma Aldrich #D2438), except for Sodium Valproate which was made in ddH2O. Upon treatment, the stock compounds were diluted 1:1000 in NPC media (DMSO concentration 0.1%) and 1 mL was added to desired well in a 6-well plate.

Treatment of NPCs with HDAC inhibitors

NPCs were plated at 0.5×10⁶ cells per well in a poly-ornithine/lamanin double-coated 6-well plate in 2 mL NPC media with above growth factors. The cells were allowed to incubate at 37°C for 48 hours to grow to 95–100% confluency. The media was aspirated and the cells were treated with compound in fresh NPC media for 24 hours.

Treatment of Neurons with HDAC inhibitors

Human iPSC-derived neurons were generated as described above. On day 18, the media was aspirated and the cells were treated with compound in fresh NPC media for 24 hours.

Determination of Gene Expression Changes in Human NPCs and Neurons

RNA was generated from each well of a 6-well treated plate of NPCs or neurons. Media was aspirated. The wells were washed with 1 mL of DPBS and then lysed with 1 mL TRIzol Reagent (ThermoFisher #11596026). The cells in TRIzol were incubated at room temperature for 5 minutes and RNA was extracted with the DirectZol RNA MiniPrep Kit (Zymo Research #R2052). All RNA was stored at −80°C until ready to use.

cDNA was generated using the High Capacity cDNA synthesis Kit with RNase inhibitor (ThermoFisher #4368814) with 1200 ng RNA. cDNA was used immediately or stored at −20°C until ready to use. Before use, cDNA was diluted 1:4 with DNase/RNase-free H₂O (Invitrogen #10977-015).

qPCR was conducted on the Roche 480 Light Cycler in a 384-well plate. Into each well was added 5 μL of TaqMan 2X Gene Expression Master Mix (ThermoFisher #4369510), 0.5 μL of 20X commercial TaqMan primer probe (ThermoFisher, GRN: Hs00963707_g1, FXN: Hs00175940_m1, GAPDH: Cat.#432924E), 0.5 μL of DNase/RNase-free H₂O, and 5 μL of above diluted cDNA. Results were normalized to GAPDH and replicate mean values and standard error of the mean are reported. Significance was determined with unpaired t-tests with GraphPad Prism software.

Western Blot Antibodies and Analysis

Cell pellets were collected from each well of a 6-well plate, frozen in dry ice, and stored at −80°C until ready to use. Cell pellets were lysed in radio immunoprecipitation assay (RIPA) buffer (Boston BioProducts #BP-115) with EDTA-free protease inhibitors (Sigma Aldrich #4693159001) and rocked at 4°C for 30 minutes. The lysates were centrifuged at 14,000 rcf at 4°C for 25 minutes and the supernatant was collected. Protein quantification was done with the Pierce bicinchoninic assay (BCA) (ThermoFisher #23227). Lysates were diluted to 800 ng/μL in RIPA Buffer and stored between −20 and −80°C until ready for use. Before use, lysates were boiled at 95°C with SDS loading buffer (New England BioLabs #B7703S) + DTT (New England BioLabs #B7705S) for 5 min.

In probing for PGRN, proteins were separated on NuPAGE 4–12% Bis-Tris gels (ThermoFisher #NP0335BOX) in MOPS SDS Running buffer (ThermoFisher #NP0001). For H3K9 acetylation blots, proteins were separated on 16% Tricine gels (ThermoFisher #EC6695BOX) in Tricine running buffer (ThermoFisher #1691442). To each well was loaded with 8 μg total protein and gels were run at 125V for 1 hour. Gels were then transferred onto 0.45 μm PVDF membranes (ThermoFisher #88518). Membranes were blocked in 5% milk in TBST for 1 hour and probed overnight at 4°C with primary antibodies in 5% BSA + 0.02% sodium azide (PGRN: Invitrogen #40-3400, 1:1000; H3K9Ac: Sigma Aldrich #H9286, 1:5000; TFEB: Bethyl #A303-673M, 1:1000; pTFEB: ThermoFisher #ABE1971MI, 1:5000; HDAC1: Abcam #ab7028, 1:10,000; HDAC2: Abcam #ab7029, 1:10,000; HDAC3: BD Transduction #611125, 1:1000; GAPDH: Abcam #ab8245, 1:10,000). Membranes were then washed with PBS, incubated with secondary antibody in TBST containing 5% milk (for all proteins except HDAC3 and GAPDH: anti-rabbit-HRP, Cell Signaling #7074S, 1:2000; for HDAC3 and GAPDH: anti-mouse-HRP, Cell Signaling #7076S, 1:2000) for 1 hour, washed with PBS for 1 hour, and developed with chemiluminescence reagents (Pierce ECL Western Blotting Substrate, ThermoFisher #PI32106; SuperSignal West Dura Extended Duration Substrate, ThermoFisher #PI34076).

ELISA

All ELISA was done with a Progranulin (human) ELISA kit (AdipoGen #AG-45A-0018YPP-KI01) at the manufacturer’s instructions. Cell culture supernatant was collected from treated cells after 24 hours and incubated with protease inhibitors (Sigma Aldrich #4693159001). The collected media was spun down to remove debris and supernatant was collected and stored at −80°C until ready to use. Supernatant was diluted 1:5 in provided ELISA buffer, and protein lysates (as prepared above) were diluted 1:100 in ELISA buffer. ELISA results were collected with SpectraMax Plus 384 Microplate Reader (Molecular Devices). ELISA data was normalized to control values. Replicate mean values and standard error of the mean are reported. Significance was determined with unpaired t-tests with GraphPad Prism software.

In vitro HDAC Enzymatic Binding Assay

The in vitro HDAC enzymatic assay was done with HDAC1, HDAC2, and HDAC3 using an endpoint assay with fluorogenic HDAC substrate MAZ1600 (Bradner et al., 2010) in HDAC assay buffer (in Milli-Q water, 100 mM KCl; 50 mM HEPES, GIBCO #15630-114; 0.05% BSA, Invitrogen #P2489; 0.001% Tween-20 Zymed #00-3005). To a well of a 96-well plate was added 20 μL of 6X HDAC inhibitor stock in HDAC assay buffer (compounds purchased from commercial vendors) and 60 μL enzyme at appropriate concentration in HDAC assay buffer (HDAC1, BPS Biosciences #50051, 30 ng; HDAC2, BPS Biosciences #50052, 45 ng; HDAC3, BPS Biosciences #50003, 12ng) and left to incubate for desired amount of time (0–3 hours). After desired incubation time, 40 μL 3X MAZ1600 in HDAC assay buffer (at 3X substrate K_m per enzyme) was added to each well. The reaction was left to proceed for 40 minutes whereupon a background read was taken with an Envision Multiwell plate reader (Perkin-Elmer; excitation 355 nm, emission 460 nm). The reaction was then quenched with 150 nM Trypsin (Worthington Biochemical Corporation #LS003744) + 10 μM panobinostat (Broad Institute) and was allowed 30 minutes to develop after which a fluorescent read was taken on the Envision Multiwell plate reader (Perkin-Elmer; excitation 355 nm, emission 460 nm).

COMET Cell Culture and Immunofluorescence

NPCs as cultured above were switched to phenol red free COMET NPC medium composed of 48.5% fluorobrite DMEM (Life Technologies #A1896701), 48.5% DMEM/F12 Nutrient mix (Life Technologies #21041025), 1X B27 Supplement (50X, Gibco #17504-044), 1X penicillin/streptomycin, and growth factors EGF, FGF, and Heparin as detailed above. NPCs were plated in a 96-well poly-ornithine/laminin coated microplate (Corning #3904) at 15,000 cells/well in 100 μl COMET NPC medium. On the following day, media in wells was supplemented to 250 μl with COMET NPC media containing HEPES (Gibco #15630, final concentration 4%), glycerol (final concentration 4%), and oxyrase (Oxyrase Inc, #OF-0005, final concentration 1%). Compounds (DMSO, BG47 at 10 μM, panabinostat at 0.5 μM) were added into wells with an HP D300 Digital Dispenser (Hewlett-Packard). The cell plate was placed onto the COMET LED microplate stage (PCB board v6.2, as described in Reis et al, 2016) and cells were either exposed to no light or to modulating light cycles of 25 ms on/75 ms off at a frequency of 1 Hz for 16 hours. Subsequently, cells were fixed with 4% formaldeyde for 1 hour, and then incubated with 100% MeOH for 1 hour, followed by a 1 hour permeabilization and blocking step with a 0.1% Triton-X 100 and 2% BSA solution. Cells were either incubated overnight with anti-H3K9Ac (Millipore #07-352, 1:1,000) or anti-PGRN antibody (Invitrogen #40-3400, 1:50), followed by a 1 hour incubation with secondary antibody Alexa Fluor 594-conjugated antibody (Molecular Probes #A11012, 1:500) and nuclear probe Hoechst (Invitrogen #H3570, 1:1,000). 25 images/well were captured with a 20× objective using the In Cell Analyzer 6000 (GE Healthcare). For quantification of the immunofluorescence, cells were either segmented for nuclei to assess H3K9Ac intensity or subcellular regions (nuclei and cytoplasm) were analyzed for PGRN intensity. For quality assessment, the following criteria were used: 1) Cell Count per Well > 1600/well. 2) 200 > Cell Count Per captured Field > 99. For H3K9Ac, the mean nuclear intensity was determined, while for PGRN, the mean cytoplasmic intensity was determined and data was normalized to light condition-dependent DMSO controls. Replicate mean values and standard error of the mean are reported.

HDAC shRNA Studies

NPCs were seeded at 50% confluency in poly-ornithine/laminin coated 12-well plastic dishes (1.6 × 10⁵ cells/well). After a 24 hour incubation, the cells were infected at an MOI = 10 with lentivirus packaged with short-hairpin RNAs (shRNAs) cloned into lentiviral vector pLK0.1 (http://www.broadinstitute.org/rnai/public/vector/details?vector=pLKO.1) targeting the following sequences: HDAC1 – CCTAATGAGCTTCCATACAAT; HDAC2 – CAGTCTCACCAATTTCAGAAA; HDAC3 – CAAGAGTCTTAATGCCTTCAA; RFP (control) — CTCAGTTCCAGTACGGCTCCA; LacZ (control) – TCGTATTACAACGTCGTGACT; luciferase (control) —ACGCTGAGTACTTCGAAATGT; GFP (control) – ACAACAGCCACAACGTCTATA. All lentivirus-packaged shRNAs were obtained from the Broad Institute Genetic Perturbation Platform (http://portals.broadinstitute.org/gpp/public/). The cells were spun down at 2000 rpm for 30 minutes at room temperature and incubated at 37 degC for 24 hours. After 24 hours, the media was changed and complete NPC media was added and the cells were incubated at 37 degC for 24 hours. When the cells were 90–100% confluent, complete NPC media with puromycin (0.8 μg/mL; Sigma #P8833) was added to select for infected cells. Cells were thereafter grown and split as denoted above with puromycin in the media.

ChIP-qPCR Studies

NPCs were grown in 15-cm dishes and treated with vehicle (DMSO), panobinostat (0.5 μM), or CI-994 (10 μM) for 24 hours. For H3K27 acetylation ChIP, 4 × 10⁷ cells were used per immunoprecipitation. Cells were fixed with NPC media containing 1% formaldehyde (Tousimis #1008A) for 10 minutes at room temperature and quenched with glycine at a final concentration of 125 mM. Cells were collected, lysed in ChIP lysis buffer (1% SDS, 10 mM EDTA, 50 mM Tris-HCl pH8 in ddH₂O), and sonicated with EpiShear Probe Sonicator for 12 cycles of 20 seconds each at 50% amplitude.). The sonicate was spun down and the supernatant was diluted 1:10 in ChIP Dilution Buffer (0.01% SDS, 0.275% TritonX-100, 1.2 mM EDTA, 16.7 mM Tris-HCl pH8, 0.167 M NaCl in ddH₂O) and incubated with Anti-H3K27Ac antibody (Abcam #ab4729) at a final concentration of 3 μg/mL overnight at 4°C. Dynabeads Protein A (ThermoFisher #10001D) were then added to the antibody-lysate mixture and incubated for 1 hour at 4°C, after which the beads were collected and washed for 20 minute intervals, twice with a Low Salt Buffer (0.1% SDS, 1% TritonX-100, 2 mM EDTA, 20 mM Tris-HCl pH8, 150 mM NaCl in ddH₂O), twice with a High Salt Buffer (0.1% SDS, 1% TritonX-100, 2 mM EDTA, 20 mM Tris-HCl pH8, 500 mM NaCl in ddH₂O), twice with a Lithium Chloride Buffer (250 mM LiCl, 10 mM Tris-HCl pH8, 1 mM EDTA, 1% deoxycholate, 1% Igepal CA-630 in ddH₂O), and twice with 1X TE Buffer pH8. The DNA-protein complex was eluted from the beads at 65 °C using ChIP Elution Buffer (1% SDS, 150 mM NaCl in 1X TE Buffer pH8) with 1:200 1.25 M DTT added just before use. This elution was performed twice with 125 μL of ChIP Elution Buffer. The eluates were combined and underwent reverse crosslinking at 65°C overnight. The DNA was then purified using the MinElute reaction cleanup kit (Qiagen #28206), and quantified using Qubit 3.0 fluorometer. qPCR was done directly with the eluted DNA with Taqman primer-probe sets as follows: ChIP Region 1: Forward primer: AGGATAGAAAGGCGAGCACA, Reverse primer: CACCCCATTTCTAGGGATCA, Probe: TTCATAACACTCCCTCGCACT; ChIP Region 2: Forward primer: CCACCCCACTGAACTAGCTG, Reverse primer: GCCCTTGCCTCTCCATCTAT, Probe: TAGCTGAGCCTGGGAGAAGA; ChIP Region 3: Forward primer: CTCACGTTTGCTCCTCTTCC, Reverse primer: CCACAGAGCCCCTGTAAGGT, Probe: TGGTTCTACCTGCTGTGAGCT. Results were normalized to concentration of DNA used per qPCR reaction. Controls included rabbit IgG controls, as wells as input controls.

For TFEB ChIP studies, NPCs were plated, treated, fixed, and collected as above and the SimpleChIP Enzymatic Chromatin IP Kit (Cell Signaling Technology #9003) was used to conduct chromatin immunoprecipitations with 4×10 cells per immunoprecipitation according to manufacturer’s instructions. Antibodies used were as follows: anti-TFEB (Cell Signaling Technology #37785), Normal Rabbit IgG (Cell Signaling Technology #2729), positive control Histone H3, data not shown (Cell Signaling Technology #4620). qPCR was done with eluted DNA with custom TaqMan primer-probe sets denoted in Supplemental Experimental Procedures. Results were obtained as percent of input DNA and normalized to vehicle. Significance was determined with unpaired t-tests with GraphPad Prism software.

QUANTIFICATION AND STATISTICAL ANALYSIS

Protein was done either by ELISA or by ImageJ quantification of Western blot. Graphical data are generally presented as mean + SEM with statistical analyses carried out by GraphPad Prism. P < 0.05 is considered significant.

Highlights.

• One cause of frontotemporal dementia is haploinsufficiency of progranulin (PGRN)

• Inhibition of Class I HDACs is sufficient to upregulate PGRN in human neurons

• Only fast-binding HDAC inhibitors enhance PGRN expression in human neuronal cells

• H3K27Ac and TFEB occupancy on the GRN promoter correlate with PGRN expression