Identification of potent inhibitors of the sortilin-progranulin interaction

Article abstract of DOI:10.1016/j.bmcl.2020.127403

High-throughput screening methods have been used to identify two novel series of inhibitors that disrupt progranulin binding to sortilin. Exploration of structure-activity relationships (SAR) resulted in compounds with sufficient potency and physicochemic

Full text of DOI:10.1016/j.bmcl.2020.127403

Bioorganic&MedicinalChemistryLetters30(2020)127403
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Identification of potent inhibitors of the sortilin-progranulin interaction
⁎
Shawn J. Stachela, , Anthony T. Ginnettia, Scott A. Johnsonc, Paige Cramerd, Yi Wangb,
Marina Bukhtiyarovab, Daniel Kroskyb, Craig Stumpa, Danielle M. Hurzya, Kelly-Ann Schlegela,
Andrew J. Cookea, Samantha Allenb, Gregory O'Donnelle, Michael Ziebelle, Gopal Parthasarathyc,
Krista L. Gettye, Thu Hoe, Yangsi Oub, Aneta Jovanovskab, Steve S. Carrollb, Mark Pauschb,
Kevin Lumbe, Scott D. Mosserb, Bhavya Voletid, Daniel J. Kleinc, Stephen M. Soissonc,
Celina Zerbinattid, Paul J. Colemana
a Department of Medicinal Chemistry, Merck & Co. Inc., PO Box 4, West Point, PA 19486, USA
b Department of Pharmacology, Merck & Co. Inc., PO Box 4, West Point, PA 19486, USA
c Department of Chemistry and Structural Chemistry, Merck & Co. Inc., PO Box 4, West Point, PA 19486, USA
d Department of Neuroscience, Merck & Co. Inc., PO Box 4, West Point, PA 19486, USA
e Screening and Protein Science, Merck & Co. Inc., PO Box 4, West Point, PA 19486, USA
A R T I C L E I N F O
A B S T R A C T
Keywords:
High-throughput screening methods have been used to identify two novel series of inhibitors that disrupt pro-
granulin binding to sortilin. Exploration of structure-activity relationships (SAR) resulted in compounds with
sufficient potency and physicochemical properties to enable co-crystallization with sortilin. These co-crystal
structures supported observed SAR trends and provided guidance for additional avenues for designing com-
pounds with additional interactions within the binding site.
Sortilin
Progranulin
Protein-protein interaction inhibitor
Structure activity relationship
Progranulin (PGRN) is a 593 amino acid secreted glycoprotein in-
volved in development, inflammation, cell proliferation and protein
homeostasis. Mutations in the granulin (GRN) gene, that encodes for
(pro)granulin result in haploinsufficiency and reduced progranulin le-
vels leading to the most common inherited form of frontotemporal
dementia (FTD).1,2 Individuals affected by this autosomal dominant
trait carry loss-of-function mutations resulting in a 50% reduction in
progranulin levels. In addition, GRN-null mutations have since been
associated with other neurodegenerative phenotypes such as Hunting-
ton’s, Parkinson’s, and Alzheimer’s Disease. Aged GRN knock-out (KO)
mice show reduced survival, age-dependent gliosis, and increased levels
of phosphorylated TDP-43 and other markers of cellular aging.3 Fur-
thermore, GRN KO mice display behavioral abnormalities including
reduced social engagement and learning/memory deficits.4 The above
data suggests that reduction in progranulin levels could trigger me-
chanisms that are common in several neurodegenerative diseases, and
that intervention which increases progranulin levels may therefore
have a substantial therapeutic benefit.
system and periphery.5 SORT functions by shuttling proteins between
the cell surface and various intracellular compartments, directing tar-
geted proteins to distinct fates including cell surface exposure, signal
transduction, regulated secretion, endocytic uptake and anterograde/
retrograde sorting. PGRN binds with high affinity to SORT, resulting in
its cellular uptake and eventual degradation in the lysosome.6 Sup-
portive evidence for SORT’s role in controlling PGRN levels has also
been demonstrated in SORT KO mice where brain and plasma PGRN
levels are increased by 2.5 and 5-fold respectively.3 These results in-
dicate that inhibition of the SORT-PGRN interaction has the potential to
increase PGRN levels 2.5-fold thus restoring the 50% deficit in PGRN
levels observed in loss-of-function human genetic mutations causative
of FTD. As such, disruption of the SORT-PGRN interaction maybe a
viable therapeutic pathway for increasing progranulin levels in the CNS
thereby protecting against neurodegenerative diseases including Alz-
heimer’s Disease.
Only two small molecule protein-protein interaction inhibitors
(PPIs) of the SORT-PGRN interaction have been reported in the litera-
ture to date. The Lundbeck inhibitor, AF40431 was the first such re-
ported small molecule in the public domain, but its use a tool molecule
was limited due to low solubility and membrane permeability.7
Sortilin (SORT) is a type I membrane receptor, belonging to the
family of vacuolar protein sorting 10 protein (VPS10P) domain re-
ceptors, that is ubiquitously expressed in both the central nervous
⁎ Corresponding author.
E-mail address: shawn_stachel@merck.com (S.J. Stachel).
https://doi.org/10.1016/j.bmcl.2020.127403
Received 1 June 2020; Received in revised form 6 July 2020; Accepted 7 July 2020
Availableonline15July2020
0960-894X/©2020ElsevierLtd.Allrightsreserved.

S.J. Stachel, et al.
Bioorganic&MedicinalChemistryLetters30(2020)127403
a widely diverse array of functionality with very little effect on activity.
Variation of functional groups and positions on the aryl ring as shown
in compounds 3–5 resulted in little change in potency. Likewise, ex-
tending the aryl group out by installation of a two carbon-linker as seen
in compound 6 produced little effect. Introduction of charged or polar
residues (compounds 7 and 8) as well as replacement of the aromatic
ring with a saturated hydrocarbon (compound 9) also result in little
change in activity. Most notably, even a large, saturated, polar amide
such as compound 10 was within 3-fold of the simple benzamide hit 1.
Given this binding data, it was the team’s hypothesis that this region of
the compound most likely does not interact directly with the protein,
but more likely extends into solvent. As such, it was assumed that en-
hancement in binding affinity or potency in a ligand efficient manner
would be difficult with continued exploration of this region. It was
thought, however, that this region could eventually be exploited to
modify the physicochemical properties of an inhibitor if additional
binding potency could be realized elsewhere.
Fig. 1. Published sortilin inhibitors.
With the knowledge that we could not meaningfully increase po-
tency through structural change in the amide region, we shifted our
attention to exploration of the amino acid side chain. Keeping our most
potent amide from the initial scan constant (the 3,5-dichloroamide from
compound 3), we screened this region resulting in a much wider range
of SAR (Table 2). Here the large cyclohexyl sidechain in 13 displayed
approximately the same potency as the isopropyl sidechain, 3, whereas
more subtle changes to the isopropyl group such as isopropenyl (11),
cyclopropyl (12) or 1.1.1 bicyclopentane (15) lost several fold in po-
tency. Replacement with a phenyl group (16) abrogated all activity.
Interestingly, swapping from an isopropyl to a t-butyl group (14) re-
sulted in an 8-fold increase in potency. Homologation to the t-butyl
ethyl sidechain afforded an addition enhancement in potency, resulting
in 17, the most potent compound in the series (IC50 = 0.17 µM).
However, exchanging one of the carbons for an oxygen (18) resulted in
a 10-fold loss of potency and additional branching (19) led to a more
dramatic loss in activity. This data led the team to the hypothesis that
the amino acid sidechain is likely interacting with a more discrete
pocket within the enzyme as compared with the amide region, which
showed relatively flat SAR. It should be noted that activity was found to
reside in the S enantiomer and the R enantiomers are inactive for all
Fig. 2. HTS lead 2 and subsequent benchmark compound 1.
Lundbeck has since described AF38469 as an improved SORT-PGRN PPI
with improved physicochemical properties (Fig. 1).8 As part of our in-
ternal effort to identify inhibitors of the SORT-PGRN interaction, a
high-throughput screen (HTS) was performed on the internal Merck
compound collection using a homogeneous time-resolved fluorescence
(HTRF) assay format9 and subsequently confirmed as SORT binders
using surface plasmon resonance (SPR). Two series, exemplified by
compounds 1 and 2, emerged as promising chemotypes from a ligand
binding efficiency perspective to warrant additional investigation
(Fig. 2).
Compound 1 was identified as a singleton hit from HTS, which was
notable given its relatively simple chemical structure. We began in-
vestigating the SAR by focusing on the amide as an easy handle to
quickly explore this region (Table 1). Surprisingly, the amide tolerated
Table 1
Amide SAR.
Table 2
Amino acid SAR.
2

S.J. Stachel, et al.
Bioorganic&MedicinalChemistryLetters30(2020)127403
Fig. 3. Properties of Compound 17.
compounds in this series.
Because of its superior potency, we further profiled 17 to assess its
utility as a potential in vivo tool molecule (Fig. 3). In addition to its
activity in the biochemical assay, compound 17 also displayed sub-
micromolar activity in a functional-based progranulin uptake assay9
(0.65 µM). This assay measures the progranulin levels remaining in
culture supernatant as measured by ELISA. Though highly protein
bound in rat (99.78%), 17 displayed a good pharmacokinetic profile
overall with a low unbound clearance in rat of 215 mL/min/kg, t1/
= 5.9 h and 100% F. Compound 17 also possessed good solubility at
2
pH 2 and 7, good apparent permeability, was a borderline P-glyco-
protein (Pgp) substrate in rat but a non-substrate against the human
MDR1 cell-line, and had a clean ancillary profile with no significant
responses < 10 µM in a counter-screen panel consisting of a variety of
46 enzymes and receptors.
As part of our exploratory SAR we also sought to investigate con-
tributions of the amide and carboxylic acid to binding (Figure 4). While
transformation of the acid to a primary amide or amine resulted in loss
of inhibitory activity, we found that we were able to replace the car-
boxylic acid using tetrazole as a bioisostere (racemic compound 20)
however the modification effected a 10-fold potency loss. Additionally,
we demonstrated that the carbonyl of the amide bond was not involved
in a critical hydrogen bonding interaction with sortilin as replacement
of the amide carbonyl with a CF3 amine (21) resulted in only a slight
loss in activity.
Fig. 5. NTS binding interactions with SORT receptor.
importance of these terminal amino acids for binding has also been
demonstrated experimentally where truncation of the five c-terminal
(Leu-Arg-Glu-Leu-Leu) residues from full-length progranulin peptide
completely eliminated its ability to be endocytosed by SORT.11
We co-crystallized compound 17 with SORT and showed that it also
binds to the NTS/PGRN binding site (Fig. 6). As seen in the co-crystal
with neurotensin, compound 17 is also anchored through a salt bridge
with Arg292 and additional interaction with Ser283 and Tyr318. In
addition, the t-butylethyl sidechain occupies the same hydrophobic
PGRN binds SORT in the 10-bladed β-propeller domain, resulting in
cellular uptake and degradation in the lysosomes.6 Neurotensin (NTS),
a 13-aa neuropeptide has also been reported as a SORT ligand, binding
to the same location as PGRN. The SORT-NTS complex has been de-
termined by X-ray crystallography at 2 Å resolution.10 The structure
shows neurotensin binding in the 10-bladed β-propeller domain, but
only the 4-C-terminal residues (10-Pro-Tyr-Ile-Leu-13) are resolved in
the x-ray structure. The well-defined C-terminal region indicates its
importance in SORT binding (Fig. 5).5 Analysis of this structure showed
a buried C-terminal salt bridge interaction with Arg292 with additional
bifurcating hydrogen-bond interactions with Tyr318 and Ser283, the
hallmark salt bridge. In addition to the salt bridge interaction, the
leucine sidechain can be seen nestled into a small hydrophobic pocket.
Hydrogen bond interactions are also present between the pocket and
the terminal amide of NTS, as well as hydrogen bonding interactions
between the hydroxyl group of Tyr318 and Lys227 of SORT. The
Fig. 6. Co-crystals structure of compound 17 in the SORT binding pocket. (A)
indicates key hydrogen bonding interactions (black dashes) between compound
17 (grey) and the protein (pink). (B) Protein molecular surface representation.
PDB ID: 6X48.
Fig. 4. Compound 17 bioisostere replacements.
3

S.J. Stachel, et al.
Bioorganic&MedicinalChemistryLetters30(2020)127403
Table 3
SAR for pyrrazole benzyl region.
Fig. 7. Optimized compounds in amide series.
and 5.32 for compounds 22 and 23 respectively vs. 4.02 for compound
17. These findings are particularly noteworthy given that the increased
potency is comparable with the intrinsic activity of the 18-mer PGRN
peptide, IC50 = 0.014 µM, which issued as the positive control in the
assay. In addition to the improvements in activity, compounds 22 and
23 also possessed favorable pharmacokinetic profiles with low unbound
clearance in rat, good solubility at pH 2 and 7, and Pgp efflux ratio
suitable for CNS penetration although the apparent permeability for
these analogs was somewhat reduced compared to 17.
The improvement in potency observed in compound 23 can be ra-
tionalized by analyzing the co-crystal of a similar compound 24, in the
SORT binding site where the homo t-butyl group is replaced by a t-butyl
leucine. Fig. 8 shows that additional hydrophobic interactions are
present in an edge-to-face π-interaction of the terminal aryl ring with
Tyr362, as well as a face-to-face π-interaction between the pyridyl ring
and Phe317. Two additional interactions are also observed: a direct
hydrogen-bonding interaction between the aryl ether oxygen and the
Tyr362 phenol as well as a bidentate water-mediated hydrogen-
Fig. 8. Co-crystals structure of compound 24 in the SORT binding pocket. (A)
indicates key hydrogen bonding interactions (black dashes) between compound
24 (grey) and the protein (yellow). (B) Protein molecular surface representa-
tion. PDB ID: 6X4H.
pocket as the leucine sidechain in neurotensin. The aryl ring, however,
is orientated into the central solvent exposed cavity of the sortilin
channel. The lack of hydrophobic interactions with the protein here
potentially explain the promiscuity observed in the flat SAR for this
region.
Table 4
SAR in 3-position of pyrrazole series.
The crystallographic data also supported the result obtained with
compound 21, where the amide was replaced with α-CF3 amine, con-
firming that the carbonyl oxygen was not involved a direct hydrogen-
bonding interaction with the protein. This is in contrast to the amide
carbonyl of NTS, which is observed to hydrogen bond to Ser272 in the
SORT-NTS x-ray structure. These structural results spurred a re-
investigation of the amide with the hypothesis that binding might be
enhanced through more distal interaction with the progranulin binding
site as observed with neurotensin. As such, we re-explored the amide
region again, this time using the optimized t-butylethyl amino acid side
chain.
Upon a second iteration of amide screening, we were able to affect
additional improvements in activity whilst retaining the favorable
physicochemical properties present in compound 17 as shown with
compounds 22 and 23 (Fig. 7). These improvements in potency are also
commensurate with increases in lipophilic ligand efficiency (LLE), 5.57
4

S.J. Stachel, et al.
Bioorganic&MedicinalChemistryLetters30(2020)127403
initially hypothesized that these structurally distinct HTS hits that ap-
peared to share common functional groups might bind in similar lo-
cations, it is reassuring nonetheless that the convergent SAR and
binding validation by co-crystallization confirmed these assumptions.
In summary, we have identified two novel inhibitor series of the
SORT-PGRN interaction via high-throughput screening. Initial SAR re-
sulted in compounds with improved potency and physicochemical
properties enabling co-crystallization with SORT. These co-crystal
structures supported the SAR and provided guidance for additional
avenues for increased interactions within the binding site. These two
series were found to share a common binding motif anchored by a
carboxylate-arginine salt bridge that is also critical in PGRN binding to
the SORT receptor. Further optimization of these series and in vivo
pharmacodynamic studies will be the subject of a future publication.
Supporting Information: Experimental procedures and compound
characterization data for lead compounds. The Supporting Information
is available free of charge on the ACS Publications website. Structure
coordinates have been deposited in the Protein Data Bank (PDB): codes
6X48, 6X4H, 6X3L.
Fig. 9. Co-crystals structure of compound 2 in the SORT binding pocket. (A)
indicates key hydrogen bonding interactions (black dashes) between compound
2 (grey) and the protein (cyan). (B) Protein molecular surface representation.
PDB ID: 6X3L.
bonding interaction between the aryl ether and the Tyr362 carbonyl
and phenol.
Declaration of Competing Interest
Compound 2 represents a second series that was identified in our
HTS campaign. We began our investigation of this second series by
focusing on the benzyl portion of the molecule. As was observed in the
amide portion of the first series, the SAR was very flat with a wide
diversity of size and polarity resulting in little effect on potency as
shown in Table 3. By contrast the t-butyl group showed much less
tolerability for modification (Table 4) as compared to the previous
amide series. In light of the structural similarities between the two
series (t-butyl, carboxylic acid moieties) we sought to determine if
homologation of the t-butyl group would prove beneficial as it had
previously. Unfortunately, these modifications resulted in reduced po-
tency.
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influ-
ence the work reported in this paper.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://
doi.org/10.1016/j.bmcl.2020.127403.
References
Again, co-crystallographic analysis of the inhibitor-SORT complex
with 2.7 Å resolution provided additional insight. The crystal structure
showed, unsurprisingly, that this series indeed binds in the same site on
SORT as the previous series with similar interactions (Fig. 9). The
carboxylic acid was involved in a hydrogen bonding network with
Arg292, Ser283, and Tyr318, and the t-butyl group occupied the hy-
drophobic binding pocket occupied by the leucine in NTS and com-
pound 17. Again, the benzyl group was oriented into the solvent ex-
posed central cavity of the SORT channel thereby explaining the flat
SAR, as was observed in the amide series. However, in this instance, the
benzyl substituent was orientated upwards, thereby limiting access to
potentially engaging additional binding interactions as realized in
compound 23 as compared to 27. The most active analog obtained in
this series was compound 31 with and IC50 = 0.49 µM. While it was
1. Eriksen JL, Mackenzie RA. J. Neurochem. 2008;104:287–297.
2. Riedl L, Mackenzie IR, Forstl H, Kurz A, Diehl-Schmid J. Neuropsychiatr Dis Treat.
2014;10:297–310.
3. Kumar-Singh S. J. Mol. Neurosci. 2011;45:561–573.
4. Arrant AE, Filiano AJ, Warmus BA, Hall AM, Roberson ED. Genes Brain Behavior.
2016;15:588–603.
5. Petersen CM, Nielsen MS, Nykjaer A, et al. J. Biol. Chem. 1997;272:3599–3605.
6. Hu F, Padukkavidana T, Vægter CB, et al. Neuron. 2010;68:654–657.
7. Andersen JL, Schrøder TJ, Christensen S, et al. Acta Cryst. 2014;D70:451–460.
8. Schroder TJ, Christensen S, Lindeberg S, et al. Bioorg. Med. Chem. Lett.
2014;24:177–180.
9. Experimental conditions and procedures for the HTRF and functional-based pro-
granulin uptake assays are described in the Supporting Information section.
10. Quistgaard, E. M.; Madsen, P.; Grøftehauge, M. K.; Nissen, P.; Claus M Petersen, C. M.
;Søren S Thirup, S. S. Nat. Struct. Mol. Biol. 2009, 16, 96-98.
11. Lee WC, Almeida S, Prudencio M, et al. Hum Mol Genet. 2014;23:1467–1478.
5