Kinetic Tuning of HDAC Inhibitors Affords Potent Inducers of Progranulin Expression

Abstract

Histone deacetylases (HDACs) are enzymes involved in the epigenetic control of gene expression. A handful of HDAC inhibitors have been approved for the treatment of cancer, and HDAC inhibition has also been proposed as a novel therapeutic strategy for neurodegenerative disorders. These disorders include progranulin (PGRN)-deficient forms of frontotemporal dementia caused by mutations in the GRN gene that lead to haploinsufficiency. Hydroxamic acid-based inhibitors of HDACs 1–3, reported to have fast-on/fast-off binding kinetics, induce increased expression of PGRN in human neuronal models, while the benzamide class of slow-binding HDAC inhibitors does not produce this effect. These observations indicate that the kinetics of HDAC inhibitor binding can be tuned for optimal induction of human PGRN expression in neurons. Here, we further expand on these findings using human cortical-like, glutamatergic neurons. We provide evidence that two prototypical, potent hydroxamic acid HDAC inhibitors that induce PGRN (panobinostat and trichostatin A) exhibit an initial fast-binding step followed by a second, slower step, referred to as mechanism B of slow binding, rather than simpler fast-on/fast-off binding kinetics. In addition, we show that trapoxin A, a macrocyclic, epoxyketone-containing class I HDAC inhibitor, exhibits slow binding with high, picomolar potency and also induces PGRN expression in human neurons. Finally, we demonstrate induction of PGRN expression by fast-on/fast-off, highly potent, macrocyclic HDAC inhibitors with ethyl ketone or ethyl ester Zn²⁺ binding groups. Taken together, these data expand our understanding of HDAC1–3 inhibitor binding kinetics, and further delineate the specific combinations of structural and kinetic features of HDAC inhibitors that are optimal for upregulating PGRN expression in human neurons and thus may have translational relevance in neurodegenerative disease.

Graphical Abstract

INTRODUCTION

Epigenetic processes are responsible for gene activation and repression¹. A complex and highly dynamic epigenetic mechanism is the reversible decoration of the side chains of histone proteins through covalent post-translational modifications (PTMs). These PTMs affect the interaction of histones with DNA and the recruitment of transcriptional machinery, and they can lead to changes between different chromatin condensation states. Residues in the N-terminal histone tail segments are continuously modified by enzymes such as lysine or arginine methyltransferases and lysine acetyltransferases, and reversely, the resulting PTMs are removed by demethylases and deacetylases. In addition, a subset of post-translational histone modifications are recognized by protein domains that recruit transcription factors among other effectors¹. Lysine acetylation disrupts electrostatic interactions between histones and DNA, favoring transcription, and it is regulated by histone acetyltransferases (HATs) and histone deacetylases (HDACs). Moreover, this modification is specifically recognized by bromodomain-containing proteins, which generally serve as transcriptional co-activators¹. By altering the acetylation state of histones, as well as non-histone proteins, HDAC inhibitors can thus affect gene expression². In particular, inhibitors of the Zn²⁺-dependent HDACs 1–3 (members of class I) have attracted interest since these are the most active deacetylases of histones both in vitro and in cells^{2, 3}. HDACs 1–3 play a critical role in a number of nuclear multiprotein transcriptional repressor complexes with distinct biological roles³. Through affecting the deacetylase activity and the overall epigenetic state, inhibitors of HDACs 1–3 induce growth arrest and differentiation in tumor cells, and they reverse the effects of gene silencing and haploinsufficiency in neurodegenerative disorders^{1, 2}.

Haploinsufficiency of the progranulin gene (GRN) has been identified as a leading cause of familial forms of frontotemporal dementia (FTD), the second most common type of presenile dementia after Alzheimer’s disease⁴. Even though the molecular pathological mechanisms are still unknown, loss-of-function mutations in one copy of the GRN gene leading to reduced progranulin (PGRN) expression have been shown to lead to ubiquitinated TAR DNA-binding protein 43 (TDP-43) accumulation that is characteristic of ~50% of neuropathologically confirmed cases of FTD⁴. Furthermore, PGRN restoration in FTD mouse models has proven beneficial for correcting abnormalities associated to the disease^5–7. A library screen employing mouse neuronal cells and following studies in induced pluripotent stem cell (iPSC)-derived human neurons identified the broadly acting HDAC inhibitor SAHA (vorinostat, 1, Scheme 1) as an epigenetic promoter of PGRN expression^{8, 9}. More detailed studies employing selective HDAC inhibitors have suggested HDACs 1–3 to be the target of SAHA (1) in relation to this phenotype, after ruling out potential contributions of HDAC6 and HDAC8 that are also inhibited by SAHA (1)¹⁰. These studies also gave rise to a proposed mechanism of action for HDAC inhibitor-induced PGRN expression involving enhanced H3K27 acetylation at the GRN promoter region and recruitment of the transcription factor EB (TFEB)¹⁰, which has also been identified as an acetylation target¹¹. Most strikingly, not all inhibitors of HDACs 1–3 effectively led to increase in PGRN protein levels. In particular, the ortho-aminoanilides CI-994 (tacedinaline, 3) and Cpd-60 (4) were shown to fail to induce PGRN expression, despite their demonstrated effect on overall histone hyperacetylation in the same human neuronal cell culture system¹⁰. Intriguingly, this observation concerning o-aminoanilides was the opposite to that observed for the regulation of the FXN gene encoding frataxin^{10, 12}, which was only enhanced by o-aminoanilides and not by hydroxamic acid-based HDAC inhibitors, thereby providing evidence for locus-selective epigenetic regulation of genes relevant to neurological disorders.

Scheme 1. Chemical structure of selected HDAC inhibitors.

(A) Small molecule and (B) peptide-based macrocyclic HDAC inhibitors, with numbers representing different scaffolds and letters indicating ethyl ketone (a), ethyl ester (b), hydroxamic acid (c) or excised (d) Zn²⁺-binding group.

HDAC inhibitors based on an o-aminoanilide chemotype have attracted significant attention due to potential selectivity towards a subset of HDAC-containing multiprotein complexes^{3, 13–15} and due to a general exhibition of slow-binding inhibition kinetics^{14, 16–19}. In-depth characterization of the mechanism of action of traditional and recent drugs has brought to light the importance of such time dependency on inhibitory effects²⁰. Most enzyme inhibitors have been thought to compete for the substrate at the time scale of enzymatic catalysis, reaching equilibrium and full inhibition rapidly in a fast-on/fast-off fashion^{20, 21}. However, some enzyme–inhibitor systems, referred to as slow-binding, approach equilibrium gradually during minutes or even hours due to slow release of the inhibitor, which can be attractive for enhancing pharmacological effects through improved residence time^22–25. A less commonly studied and often harder to detect mechanism referred to as mechanism B of slow binding is a hybrid between these two scenarios, with inhibitors exhibiting an initial fast-binding step followed by a second, slower binding step and slow release^{21, 22}.

Kinetics of HDAC inhibition appears to play a role in the regulation of GRN, and the o-aminoanilide inhibitors tested so far were unable to induce PGRN production¹⁰. Here, we interrogate this observation using a focused collection of macrocyclic HDAC inhibitors with distinct and well-characterized inhibition kinetics, and we describe the ability of macrocycle-based inhibitors with slow-binding properties to promote PGRN expression. Furthermore, we report the surprising finding that the HDAC probes panobinostat (LBH-589, 5) and trichostatin A (TSA, S1, Scheme S1) exhibit slow, tight-binding kinetics, which is unprecedented amongst hydroxamic acid-containing small molecules.

RESULTS

SAHA (1, Scheme 1) is the most widely recognized example of a fast-on/fast-off HDAC inhibitor^{16, 18, 26}, a mechanism tentatively extended to all structurally-related hydroxamic acids. In contrast, the o-aminoanilide Zn²⁺-binding group has been shown to endow small molecules with longer residence times against class I HDACs^{14, 16–19}. This has also been observed for the trifluoromethyl ketone version of SAHA (SATK, 2)²⁶. We recently reported on the kinetic effect of incorporating various Zn²⁺-binding groups onto the scaffold of naturally occurring macrocycles such as trapoxin B (7) and apicidin A (9a)²⁷. Surprisingly, hydroxamic acid analogues TpxB^Asuha (7c) and ApiA^Asuha (9c) exhibited slow-binding kinetics against HDACs 1–3, 6, and 11 with K_i inhibitor constants similar or lower than the enzyme concentration employed (slow, tight-binding)^27–29. Intrigued by these observations, we decided to explore the binding kinetics of hydroxamic acids smaller than tetrapeptide macrocycles, but with potencies higher than that of SAHA (1).

In brief, we considered the three kinetic models outlined in Figure 1, namely: 1) fast-on/fast-off inhibition (reaching equilibrium instantly), 2) mechanism A of slow-binding inhibition (with k_obs rates of enzyme–inhibitor complex formation proportional to the concentration of inhibitor), and 3) mechanism B of slow-binding inhibition (with k_obs fitting a hyperbolic function)^{22, 27}. Slow-binding inhibitors with calculated K_i under the free enzyme concentration range were described as slow, tight-binding inhibitors²⁹. First, we tested tacedinaline (3) as a slow-binding control compound. Unexpectedly, this revealed a mechanism B type of slow-binding inhibition with tight binding against HDACs 1 and 2, and nanomolar potency against HDAC3 (Figure 1C and Table 1). Interestingly, we then found panobinostat (5) and TSA (S1), widely used as HDAC probes³⁰, to also follow slow, tight-binding inhibition profiles (mechanism B) against HDACs 1–3, and effectively irreversible behavior for panobinostat (5) against HDAC2 and for TSA (S1) against HDAC3 (see Tables 1 and S1 for numerical data and Figure S1 for graphical representation). The hydroxamic acid TSA (S1) has been suggested to exhibit slower dissociation rates than SAHA (1) in a reporter displacement assay, but the kinetics were not investigated in detail and mechanism B was not considered in this previous investigation¹⁸.

Figure 1. Representative examples of different kinetic mechanisms of HDAC inhibition.

Assay progression curves and secondary plots fitted to the relevant equation (see Supporting Information for more detail). (A) Fast-on/fast-off binding kinetics, exemplified by apicidin (6a), (B) mechanism A of slow-binding kinetics, exemplified by trapoxin A (4), and (C) mechanism B of slow-binding inhibition, exemplified by tacedinaline (3).

Table 1. Summary of the potency and mechanism of HDAC inhibitors, and of their effect on human PGRN and H3K9ac levels at 1 μM concentration.

Color shadings relate to kinetic mechanisms (cyan: fast- on/fast-off, orange: mechanism A, red: mechanism B of slow-binding kinetics).

Compound	Mechanism of binding and potency (Ki(,1): nM, dis. t½: min)			Effect on PGRN (1 μM, NPCs)	Effect on H3K9ac (1 μM, NPCs)
	HDAC1	HDAC2	HDAC3*
1 (SAHA)	Ki = 1.3 (± 0.05)30	Ki = 1.6 (± 0.05)30	Ki = 5.0 (± 0.2)30	>2 fold incr.‡, 10	>20 fold incr.‡, 10
2 (SATK)	Ki = 9800 (± 6700)26 Dis. t½ = 41 (± 17)	Ki = 11000 (± 4000)26 Dis. t½ = 87 (± 22)	Ki, 1 = 570 (± 260)26 Ki ~1126 Dis. t½ ≥ 139	No change	No change
3 (tacedinaline)	Ki, 1 = 1450 (± 320) Ki ~1 Dis. t½ ≥ 1386	Ki, 1 = 2190 (± 510) Ki ~14 Dis. t½ ≥ 139	Ki, 1 = 3560 (± 400) Ki = 105 (± 86) Dis. t½ ≥ 50	Decrease‡	~18 fold incr‡
5 (panobinostat)	Ki, 1 = 33 (± 26) Ki ~0.1 Dis. t½ ≥ 41	Ki, 1 = 4.7 (± 2.5) Dis. t½ > 105	Ki, 1 = 15 (± 5) Ki ~0.1 Dis. t½ ≥ 63	~2.3 fold incr.	~36 fold incr.
6 (trapoxin A)	Ki ~0.003 Dis. t½ ~6931	Dis. t½ > 105	Ki, 1 = 0.59 (± 0.29)27 Dis. t½ > 105	~2.4 fold incr.	~5.0 fold incr.
7a (TpxBAoda)	Ki = 6.6 (± 0.5)27	Ki = 3.9 (± 0.4)27	Ki = 0.78 (± 0.05)27	No change	~2.8 fold incr.
7b (TpxBAsu(Et))†	Ki = 500 (± 90)27	Ki = 400 (± 20)27	Ki = 500 (± 80)27	No change	~2.1 fold incr.
7c (TpxBAsuha)	Ki, 1 = 0.60 (±0.12)27 Dis. t½ > 105	Ki, 1 = 0.13 (±0.11 )27 Dis. t½ > 105	Ki, 1 = 1.4 (± 0.4)27 Ki ~0.0227 Dis. t½ ≥ 173	~2.2 fold incr.	~6.3 fold incr.
8a (apicidin)	Ki = 0.824 (± 0.003)	Ki = 0.619 (± 0.004)	Ki = 0.051 (± 0.002)27	~1.7 fold incr.	~3.3 fold incr.
9b (ApiAAsu(Et))†	Ki = 10 (± 1)27	Ki = 11 (± 2)27	Ki = 15 (± 3)27	~1.4 fold incr.	~2.1 fold incr.
9c (ApiAAsuha)	Ki, 1 = 0.32 (± 0.06)27 Dis. t½ > 105	Ki, 1 = 0.35 (±0.10)27 Dis. t½ > 105	Ki, 1 = 1.8 (± 0.6)27 Ki ~0.0427 Dis. t½ ≥ 77	~2.5 fold incr.	~6.4 fold incr.
10 (romidepsin)	Ki, 1 = 0.88 (± 0.18) Ki ~0.02 Dis. t½ ≥ 26	Ki, 1 = 0.47 (± 0.09) Ki ~0.009 Dis. t½ ≥ 41	Ki, 1 = 1.3 (± 0.4)27 Dis. t½ > 105	~4.1 fold incr.	~2.3 fold incr.
10d (RomiAla, Dha)	Ki = 690 (± 3)27	Ki = 370 (± 50)27	Ki = 800 (± 100)27	No change	Decrease

^*HDAC3 in combination with the DAD domain of NCoR2.

^†Inhibition data corresponding to the carboxylic acid hydrolyzed species (TpxB^Asu and ApiA^Asu, respectively)²⁷.

^‡Treatment performed at 10 μM concentration.

HDAC inhibition assay progression curves were recorded as in previous studies by addition of a small amount of trypsin developer (10 μg/mL) to the reaction mixture^{16, 26, 27}, which does not affect the steady-state of the reaction (Figure S1). Here, it came to our attention that trypsin addition resulted in the cleavage of small C- and/or N-terminal fragments of the enzyme without affecting the catalytic domain (see Figure S2 and Methods)³¹. Gratifyingly, end-point experiments performed without trypsin in the reaction mixture afforded estimated progression curves similar to those recorded in continuous assays (Figure S3), which indicated that these terminal fragments are not required for catalysis or inhibitor binding.

In our previous findings on hydroxamic acid-containing macrocycles, irreversible behavior had only been observed for the natural product trapoxin A (6)³². This compound bears an epoxyketone Zn²⁺-binding group speculated to form a covalent bond with nucleophilic residues in the HDAC active site^{32, 33}. However, recent co-crystal structures of trapoxin A (6) and of the related natural product HC-toxin (S2, Scheme S1) show that the epoxyketone moiety is intact upon binding to HDAC8³⁴ and to the C-terminal domain (CD2) of HDAC6³⁵, respectively. Moreover, washout experiments to look for reversibility have not generally been performed for extended time periods. We have previously observed that inhibition of HDAC3 by trapoxin A (6), as well as by HC-toxin (S2), follow a slow, tight-binding profile (mechanism B)²⁷. We further reported mechanism B type of inhibition for HC-toxin (S2) against HDACs 1 and 2²⁷, and fast-on/fast-off kinetics against HDAC6 in agreement with the structural findings reported by Christianson and co-workers³⁵. Now, we also report here observing mechanism A kinetics for trapoxin A (6) against HDACs 1 and 2. Interestingly, this distinct kinetic behavior of compounds comparing HDACs 1 and 2 vs HDAC3 has also been seen with SATK (3), and with the o-aminoanilide-containing inhibitor 106 (S3, Scheme S1) developed by Gottesfeld and co-workers^{16, 26}. Lastly, apicidin (8a) and romidepsin (10) were tested against HDACs 1 and 2, confirming fast-on/fast-off binding kinetics for the former²⁷ and slow, tight-binding (mechanism B) for the latter^{27, 36}.

These kinetic data add to the understanding of HDAC inhibitor binding dynamics. Panobinostat (5), TpxB^Asuha (7c), ApiA^Asuha (9c), romidepsin (10), and TSA (S1) display dissociative half-life (dis. t_½)²² over 20 min, even though they lack the steric characteristics of the o-aminoanilide Zn²⁺-binding group (see Tables 1 and S1). Thus, the structural basis for this extended residence time should be different from conformational changes in the ‘foot pocket’ of the active site, as observed for other substituted o-aminoanilides^{31, 37, 38}. On the other hand, there is a marked difference between tacedinaline (3), as well as 106 (S3)¹⁶, versus the remaining mechanism B type of inhibitors in terms of the order of magnitude of the first and faster step equilibrium constant (K_i,1, Table 1). Both o-aminoanilides exhibit constants in the micromolar range¹⁶, whereas all hydroxamic acids and the thiol-containing depsipeptide romidepsin (10) reach nano- and sub-nanomolar K_i,1 values. This still discriminates o-aminoanilides from other Zn²⁺-binding groups, and it leaves open the possibility for kinetic constants k₁ and k_-1 to be orders of magnitude different for this compound class. In this regard, more complex and time resolved in vitro systems would be necessary for measuring k₁ and k_-1 rates for mechanism B type of inhibitors.

We then decided to investigate further our previous observations regarding the effect of binding kinetics of class I HDAC inhibitors on the epigenetic modulation of PGRN production. For this purpose, we adapted culture and compound treatment of human iPSC-derived neural progenitor cells (NPCs) and neurons in 96-well plates. This allowed the performance of simultaneous experiments in high-throughput and optimization of the use of commercially available PGRN ELISA to complement our Western blotting and qRT-PCR assays. Since HDAC inhibition was shown to furnish similar effects on both intracellular and excreted PGRN, only the former was considered for the current study¹⁰. While our previous studies on the effect of HDAC inhibitors on human iPSC-derived neurons used a method of neural differentiation involving growth factor withdrawal¹⁰, this is only one of many ways to differentiate human neurons and is known to produce a diversity of neuronal subtypes and glia. Here, we used an entirely different method of neuronal differentiation with directed differentiation to an enriched population of cortical-like, excitatory, glutamatergic neurons using induced expression of the bHLH proneural transcription factor Ngn2 (Figure S4)³⁹. Since cortical neurons are the predominant cell type that degenerates in subjects with PGRN-deficient FTD, it renders our current studies, which demonstrate the ability of class I selective HDAC inhibitors to enhance PGRN expression, highly relevant as an ex vivo surrogate of human disease biology.

Previous data on tacedinaline (3), panobinostat (5) and apicidin (8a) were recapitulated (Figure 2A and B, and Table 1) although with a slight decrease in signal which we attribute to a shorter compound treatment (18 h vs 24 h) and the high density, 96-well plate culture format which may have impacted growth rates¹⁰. Macrocyclic ethyl esters TpxB^Asu(Et) (7b) and ApiA^Asu(Et) (9b), which follow fast-on/fast-off kinetic profiles, were able to induce PGRN in both NPCs and neurons as predicted based on their HDAC binding kinetics and potency²⁷. Surprisingly, based upon the inability of the o-aminoanilides to enhance PGRN, we found that all mechanism A (trapoxin A, 6) and mechanism B (TpxB^Asuha, 7c; ApiA^Asuha, 9c; and romidepsin, 10) slow-binding macrocycles generated a substantial increase in PGRN, even at concentrations below 100 nM (compounds 6 and 9c, Figure 2C). Together with our finding that panobinostat (5) is able to bind HDACs 1–3 in a slow-binding manner, this encourages future development of PGRN enhancers with long resident times that do not rely on an o-aminoanilide moiety as the basis for the prolonged kinetics. A selection of structurally diverse compounds representative of each kinetic binding mode was also tested on HEK293 cells, as this human cell type is known to have HDAC inhibitor-inducible PGRN expression¹⁰. This afforded similar results for each of the six compounds as in NPCs, which further corroborates our observations (Figure S5).

Figure 2. Intracellular PGRN levels in response to HDAC inhibitor treatment.

Fold-change in PGRN protein concentration after 18 h compound treatment in (A) human iPSC-derived NPCs and (B) Ngn2 neurons (representation scale was adapted to maximum signal, data represent mean ± SD, see Figure S6 for alternative refined analysis). Dotted lines represent PGRN concentration upon DMSO treatment, and colors represent fast-on/fast-off (cyan), mechanism A (orange) and mechanism B (red) of slow-binding against HDAC1 and HDAC2. ^†Romidepsin (10) impaired cellular viability at 10 μM concentration. (C) Dose-response curves of tacedinaline (3, EC₅₀ >20 μM), panobinostat (5, EC₅₀ = 0.20±0.01 μM), trapoxin A (6, EC₅₀ = 0.022±0.001 μM), TpxB^Asu(Et) (7b, EC₅₀ ≈ 6.0±0.8 μM), TpxB^Asuha (7c, EC₅₀ = 0.45±0.01 μM), apicidin (8a, EC₅₀ = 0.79±0.02 μM), ApiA^Asu(Et) (9b, EC₅₀ ≈ 9±2 μM), and ApiA^Asuha (9c, EC₅₀ = 0.038±0.001 μM).

Since SATK (2, mechanism A) and tacedinaline (3, mechanism B) were the only slow-binding inhibitors that were not able to affect PGRN production (an effect also reported for Cpd-60, 4)¹⁰, we next evaluated the influence of the tested compounds on overall histone acetylation in cells to ensure cellular target engagement. In particular, we measured histone 3 (H3) acetylation at lysines (K) 9 and 27, which have been shown to be enriched upon HDAC inhibition¹⁰. The intensity of histone marks were increased to different extents by all inhibitors (Figure 3 and S7), with the exception of SATK (2) and Romi^Ala,Dha (10d), which may suffer from poor potency, stability and/or ability to penetrate the cell membrane^{26, 27}. Overall, mechanism A and mechanism B slow-binding inhibitors outperformed fast-on/fast-off inhibitors in terms of the magnitude of induction in cellular assays, highlighting the functional advantage of an extended enzyme residence time^{22, 23}. Tacedinaline (3) induced H3 acetylation to levels comparable to panobinostat (5) at concentrations where PGRN levels remained unchanged, in agreement with previous observations¹⁰. We thus found o-aminoanilides to be the only case in which effects on PGRN expression and HDAC inhibition do not correlate, independent of the mechanism of binding.

Figure 3. Relationship of PGRN expression to histone acetylation after HDAC inhibitor treatment.

Two-dimensional plots of changes in intracellular PGRN vs. changes in (A) H3K9 acetylation or (B) H3K27 acetylation upon 18 h compound treatment at 1 μM concentration. Colors relate to kinetic mechanisms against HDACs 1 and 2 (cyan: fast-on/fast-off, orange: mechanism A, red: mechanism B of slow-binding kinetics). Histone acetylation Western Blot data and images are included in Figure S7 and S8. *Tacedinaline (3) was dosed at 10 μM concentration.

Panobinostat (5), and not tacedinaline (3), has been reported to increase histone acetylation at the promoter of the GRN gene¹⁰. This could indicate that the SIN3 family of HDAC1/2-containing multiprotein complexes is responsible for the regulation of GRN transcription, as tacedinaline (3) and other o-aminoanilides are not able to inhibit them^{13, 15}. This possibility might be testable by genetic or chemical-biological perturbation of SIN3 complex components. On the other hand, and although GRN mRNA levels are upregulated by SAHA (1), panobinostat (5), apicidin (8a) and other HDAC inhibitors to quantitatively similar levels as compared to PGRN protein levels^{8, 10}, enhanced protein stability could also contribute to PGRN accumulation. PGRN contains two lysine residues that can be targeted for ubiquitination⁴⁰, and future studies will be needed to assess whether HDAC inhibition could lead to acetylation of these lysines and stabilization of PGRN protein.

In conclusion, we provide data to support the hypothesis that slow-binding HDAC inhibitors that are not based upon an o-aminoanilide scaffold can provide valid candidates for restoring PGRN levels in GRN-haploinsufficient FTD cases. This feature may be beneficial for maximizing biological effects and can be found repeatedly amongst clinically effective, approved drugs^{20, 25}. In addition, we report previously unknown mechanism B slow, tight-binding properties of certain hydroxamic acid-containing HDAC inhibitors, which we envision might influence future analysis of their behavior. The PGRN enhancement was achieved not only with hydroxamic acids, but also with compounds bearing ethyl ketone (7a and 8a) and ethyl ester (7c and 9c) Zn²⁺-binding groups. These are attractive due to better theoretical membrane penetrating properties and as potential prodrugs, respectively²⁷. Tacedinaline (3) with its mechanism B slow, tight-binding HDAC inhibitory properties, but inability to enhance PGRN expression, remains an intriguing exception in the epigenetic regulation of PGRN¹⁰. Further investigations into inhibitor kinetics and selectivity towards different physiological HDAC-containing complexes^13–15 will be necessary to fully understand the modes of action in HDAC inhibitor upregulation of PGRN.

Methods

More detailed experimental methods are included in the Supporting Information.

HDAC Inhibition Continuous Assays.

Kinetic in vitro assays were performed in Tris buffer pH 8.0 composed of 50 mM Tris/Cl, 137 mM NaCl, 2.7 mM KCl, 1 mM MgCl₂ and 0.5 mg/mL bovine serum albumin (BSA), with additional 0.2 mM tris(2-carboxyethyl)phosphine (TCEP) for assays including romidepsin (10). The assay was performed in 96-well plate format in a final volume of 50 μL per well, where dilution series of the inhibitor (2-fold dilutions) were incubated with peptide substrate (Ac-LGKac-AMC, 20 μM), trypsin (10 μg/mL) and HDAC enzyme (3.5 nM HDAC1, 1.8 nM HDAC2, or 1.2 nM HDAC3/NCoR2). HDAC integrity was assessed via steady-state substrate conversion (Figure S1) and gel electrophoresis (Figure S2). HDACs 1–3 were found to be ~7 kDa smaller than the full length protein (HDAC1_assay: 49 kDa, HDAC2_assay: 48 kDa, HDAC3_assay: 43 kDa, Figure S2) under the experimental conditions employed here and in previous studies^{16, 26, 27}. This would correspond to cleavage of the C-terminal His tag (and C-terminal HDAC1 FLAG tag) as well as of small C- and/or N-terminal fragments by trypsin without affecting the catalytic domain (36 kDa), as supported by the linear conversion rates observed in control experiments (Figure S1). Moreover, inhibitor kinetic profiling resulted similar in end-point assays without trypsin in the reaction mixture (Figure S3). In situ fluorophore release was monitored immediately by fluorescence readings recorded continuously every 30 s for 30–60 min at 25 °C. Data were fitted to the relevant equations as reported in the Supporting Information.

Human Neural Progenitor Cell Culture.

Human neural progenitor cells (NPCs) were derived from reprogrammed induced pluripotent stem cells (iPSCs), as described in the literature^{39, 41}. This cell line was transfected with an inducible Neurogenin-2 (Ngn2) construct for differentiation purposes as reported in Cheng et al., 2017³⁹. Neural proliferation medium (NPM) was composed of 70% Dulbecco’s Modified Eagle’s Medium, 30% Ham’s F12 with L-glutamine, with 1% Pen/Strep and 2% B-27 Supplement. NPM was then supplemented with Epidermal Growth Factor (EGF, 20 ng/mL), basic Fibroblast Growth Factor (bFGF, 20 ng/mL) and heparin (5 μg/mL) before use (complete NPM). NPCs were maintained in 6-well plates coated with poly-ornithine and laminin, with complete NPM, and passaged every 3 days.

Human Neuron Culture.

Human iPSC-derived neurons were obtained from NPCs by neurogenin-2 (Ngn2)-induced differentiation as described in Cheng et al., 2017³⁹. iNgn2 neural medium (N3aM) was composed of 48% Neurobasal medium, 48% DMEM/F12, 1% B-27 Supplement, 0.5% N-2 Supplement, 0.75% GlutaMax, 1% Pen/Strep, 0.5% MEM non-essential amino acids solution and 50 μM 2-mercaptoethanol. N3aM was then supplemented with 0.2% BSA, doxycycline (2 μg/mL), Brain-Derived Neurotrophic Factor (BDNF, 10 ng/mL) and Neurotrophin-3 (NT3, 10 ng/mL) just before use (complete N3aM). NPCs were plated in coated 96-well tissue culture plates in complete N3aM at 30,000 cells/well density (200 μL/well), and fed every 2 days with half media change. On day 2 and day 4, cells were fed with complete N3aM containing puromycin (1 μg/mL); on day 6, with complete N3aM containing cytosine arabinoside (AraC, 5 μM); and on days 8–12, with fresh complete N3aM. On day 13 cells were treated by adding compounds at 2× concentration in N3aM.

Enzyme-Linked ImmunoSorbent Assay (ELISA).

ELISA was performed with a commercial Progranulin (human) ELISA kit (AdipoGen, #AG-45A-0018YPP-KI01) following the guidelines provided. For NPC experiments, cells were transferred to coated 96-well plates in complete NPM at 30,000 or 60,000 cells/well density, allowed to grow to 95–100% confluency, and then treated by half media change with compounds at 2× concentration. For neuron experiments, treatment was performed on day 13 of differentiation in N3aM. Treatment was allowed to proceed for 18 h, followed by media removal and cell lysis at 4 °C for 30 min. Then, 12.5 μL/well (NPCs) or 25 μL/well (neurons) of lysate were transferred to ELISA strips and analyzed. Three internal replicates of each treatment were included, and each experiment was performed at least twice. Data were analyzed using GraphPad Prism software, and reported as mean ± SD of the biological replicates.

Western Blot.

Cell lysates were prepared for Western blot analysis by adding 4× NuPAGE LDS Sample Buffer and 30× dithiothreitol (DTT), heating and sonicating on PCR plates. 15 μL/sample were separated by gel electrophoresis in MOPS SDS Running buffer and transferred onto activated 0.45 μm PVDF membranes. Membranes were then blocked with 5% non-fat dry milk and probed overnight at 4 °C with H3K9ac (rabbit polyclonal, MilliPore, #07–352, Lot 2203126) or H3K27ac (rabbit polyclonal, Cell Signaling, #4353T, Lot 1) antibodies at 1:5000 dilution in TBST with 5% BSA and 0.02% sodium azide. Secondary HRP-conjugated antibody incubation was performed at 1:5000 dilution in TBST containing 5% milk for 1 h at room temperature, and blots were developed with chemiluminiscence reagents. Thereafter, membranes were probed at room temperature for 4 h with GAPDH antibody (mouse monoclonal, abcam, #ab8245, Lot GR3207992–1) at 1:5000 dilution in TBST with 5% milk and 0.02% sodium azide, incubated with secondary HRP-conjugated antibody and developed. Western Blot images were obtained by exposing the membranes to autoradiography films in the dark. The three internal replicates from a single experiment were analyzed in the same membrane, and relative histone acetylation fold increase is reported as mean ± SD of the three.