Engineered peptidic constructs metabolize amyloid β by self-assembly-driven reactions

Article abstract of DOI:10.1039/c9cc01582e

Inspired by the unique mechanism of proteolytic maturation of host cell factor-1, we designed small peptide-based constructs that selectively recognized amyloid β (Aβ) and cleaved it in a non-catalytic manner initially around the α-secretase cleavage-site, thus termed as "artificial α-secretases". "Artificial α-secretases" also cleaved Aβ on the cleavage-sites of other Aβ processing enzymes by prolonged treatment, as evidenced by time-resolved MALDI-TOF-mass analyses.

Full text of DOI:10.1039/c9cc01582e

ChemComm
View Article Online
View Journal
COMMUNICATION
Engineered peptidic constructs metabolize
amyloid b by self-assembly-driven reactions†
Cite this:DOI: 10.1039/c9cc01582e
Tanmay Mondal
and Bhubaneswar Mandal
*
Received 25th February 2019,
Accepted 29th March 2019
DOI: 10.1039/c9cc01582e
rsc.li/chemcomm
Inspired by the unique mechanism of proteolytic maturation of host cell cleavage site first, thus mimicking its Ab-specific proteolytic
factor-1, we designed small peptide-based constructs that selectively function. Interestingly, with prolonged treatment at physiological
recognized amyloid b (Ab) and cleaved it in a non-catalytic manner pH and temperature, ‘‘artificial a-secretases’’ also cleave Ab at the
initially around the a-secretase cleavage-site, thus termed as ‘‘artificial cleavage sites of other metabolic enzymes, finally dissolving the
a-secretases’’. ‘‘Artificial a-secretases’’ also cleaved Ab on the cleavage- amyloid and rendering it non-toxic.
sites of other Ab processing enzymes by prolonged treatment, as
To the best of our knowledge, only one protein reported to date
follows a solely Glu aided cleavage mechanism in vivo. Host cell
factor-1 (HCF-1), a transcriptional co-regulator of human cell-cycle
evidenced by time-resolved MALDI-TOF-mass analyses.
Proteases cleave peptide linkages either by a catalytic triad-based progression, undergoes proteolytic maturation in which any of its six
mechanism or assisted by a metal.1 Mimicking enzyme action has repeated sequences is cleaved by the O-linked N-acetylglucosamine
been achieved by mimicking the triad-based catalytic site by protein (O-GlcNAc) transferase (OGT).9 OGT usually glycosylates Ser or Thr
engineering2 that essentially required the specific supramolecular side chains. Therefore, OGT driven proteolysis seemed mysterious
arrangement of large protein constructs3 making them more bulky, initially. However, recently Walker et al. demonstrated that it occurs
immunogenic, and cost intensive. To bypass these challenges, we via an unusual glycosylation of a glutamate side chain by OGT
planned to achieve the proteolytic function of a protease explicitly for followed by an internal pyroglutamate formation that undergoes
a specific target. Therefore, we would be sacrificing all other functions, spontaneous peptide bond hydrolysis.10 We wanted to mimic the
even the catalytic activity of the protease, to reduce the possibility of Glu-based site-selective proteolysis mechanism of HCF-1. However,
unrelated effects keeping the total mass minimum. Thus a function nucleophilic attack of the backbone nitrogen at the glycosylated side
specific mimic of a protease or proteases would be generated.
chain carboxylate of Glu is feasible for the favorable formation of the
To demonstrate the utility of the concept we selected Alzheimer’s intramolecular five-membered cyclic pyroglutamate in vivo, but
amyloid b peptide (Ab), its precursor protein (APP) and its oligomers replicating it in an intermolecular fashion is another challenge.
as the targets. Alzheimer’s disease (AD), caused by the deposition of Therefore, we replaced O-GlcNAc by a more reactive benzyl (O-Bn)
insoluble Ab plaques outside neurons4 and hyper-phosphorylated group. Furthermore, we replaced Glu with a Lys attached to a benzyl
tau protein in the neuron along with other mechanisms, has no adipate (BnAdp). Adipic acid is a six carbon-containing linear
cure to date despite enormous efforts in diverse directions.5,6 In dicarboxylic acid. While one end was attached to the Lys side chain,
spite of the prevailing controversy, it is usually accepted that in order for it to be long enough to reach the peptide backbone of
inhibition of formation and metabolism of these peptide aggregates the target Ab or APP, the other end was benzylated. Finally, we
have therapeutic relevance.7 However, the inhibition of formation attached such constructs at the N-terminus of a self-assembling
and metabolism of disease- causing peptide aggregates cannot be partial sequence of Ab, KLVFF (Ab16–20), which is also a part of APP
achieved yet by externally added compounds or medicines. We (APP687–691).11 These peptide conjugates (‘‘artificial a-secretases’’,
engineered a small peptide to selectively recognize and cleave Ab Fig. 1) are intended to self-assemble with homologous parts of Ab
and its toxic oligomers by supramolecular interactions.8 We termed and APP by supramolecular interactions and cleave at or around the
them as ‘‘artificial a-secretases’’ as they function at the a-secretase a-secretase cleavage site, primarily. As ‘‘artificial a-secretases’’ are
designed to act by a self-assembly-driven proteolysis mechanism,
Department of Chemistry, Indian Institute of Technology Guwahati, Assam-781039,
they are also able to truncate and dissolve aggregated Ab, mimicking
the proteolytic function of NEP, ECE, IDE, P, ACE, and MMPs. Thus,
India. E-mail: bmandal@iitg.ac.in
† Electronic supplementary information (ESI) available: Experimental and char-
‘‘artificial a-secretases’’ are supposed to act as a function-specific
mimic of a-secretase and these Ab degrading enzymes at once.
acterization details, biophysical studies and additional figures as described in the
text. See DOI: 10.1039/c9cc01582e
This journal is ©The Royal Society of Chemistry 2019
Chem. Commun.

View Article Online
Communication
ChemComm
were stable for even as long as 15 days; however, debenzylated
products started appearing after only one day. After three days, peaks
corresponding to the mass of various intermediates and fragments
could be observed. Plausible routes of proteolysis of mAb by
‘‘artificial a-secretases’’ were drawn (Fig. S23–S50, ESI†) based on
the mass values. Each of the ‘‘artificial a-secretases’’ followed a
similar mechanism. Routes followed by ‘‘artificial a-secretase’’2 are
depicted in Fig. 2 as a representative example. At first, the amine
group of the side chain of Lys of mAb became attached to the
nucleophilic carbonyl of the pendant adipate of ‘‘artificial
a-secretase’’2 by the expulsion of the good leaving group,
O-Bn, generating the intermediate C00 (C0, C00 and C00 0 for ‘‘artificial
a-secretase’’1, 2, and 3, respectively) via intermolecular O to N acyl
migration (Fig. 2a and b inset). The intermediate C00 then followed
either route 1mAb–AaS2 to cleave itself between Leu17 and Val18 of
mAb to generate segment D00 and intermediate E00 that was again
sliced to form I00 or followed route 2mAb–AaS2 to cut between Lys16
and Leu17 of mAb to generate J00 and another intermediate K00. K00
might further convert to O00 via L00 and N00. The time-dependent
persistence map of the intermediates and fragments (Fig. 2c–e)
revealed that while ‘‘artificial a-secretase’’1 and 2 started frag-
Fig. 1 (a) Mechanism of HCF-1 self-cleavage, and (b) anticipated mode of
proteolysis of Ab by ‘‘artificial a-secretases’’.
Four ‘‘artificial a-secretases’’ were designed altering the positions menting mAb as early as after three days, this took ‘‘artificial
of Lys(BnAdp) (Table 1, AaS1–4 stands for ‘‘artificial a-secretase’’1–4). a-secretase’’3 B13 days. Also, while ‘‘artificial a-secretase’’2
‘‘Artificial a-secretase’’4 was designed to achieve a synergistic effect. generated most of the fragments by seven days, ‘‘artificial
To obtain an initial idea of their proteolytic activity, a partial a-secretase’’1 took 413 days. Thus, the order of reactivity is
sequence of Ab (Ab12–20, which is also APP683–691), which contains ‘‘artificial a-secretase’’2 4 1 4 3, and this depends on the
the a-secretase cleavage site, the peptide bond between Lys16 and position of the pendant group.
Leu17,12 and the self-assembling Ab16–20 part was synthesized and
However, mAb remained after as long as 30 days in all three
acted as a model aggregating peptide (mAb). One of the C-terminal cases, but it disappeared completely with two-fold molar excess
Gly mimicked the highly disease-prone Flemish Ala21 Gly of ‘‘artificial a-secretase’’4 by ten days. ‘‘Artificial a-secretase’’4
mutation;13 the other served as a spacer for the synthetic advantage. retained the site-selectivity and followed similar mechanistic
Proteolytic activity of the ‘‘artificial a-secretases’’ was finally con- pathways (routes 1–5mAb–AaS4, Fig. S51–S56, ESI†), but produced
firmed by commercial Ab1–40. To examine the selective nature of more intermediates and fragments having a range of lifetimes at a
‘‘artificial a-secretases’’, we also synthesized a 28 residue-long, faster rate than the other analogues. ‘‘Artificial a-secretases’’ also
unrelated negative control peptide (NC) based on the mutant generated self-degradation products, e.g., Q00 and R00 by ‘‘artificial
diphtheria toxin (DTP28)14 by replacing Val8 and Lys23 by Pro. a-secretase’’2, following route 1SD–AaS2 and route 2SD–AaS2, respec-
Molecular docking studies (Fig. S21 and S22, ESI†) in AutodockVina tively, confirmed by similar experiments in the absence of mAb
1.1.215 suggested that ‘‘artificial a-secretases’’ fitted well in the (Fig. 2a and Fig. S57–S63, ESI†). The self-degradation products of
16
binding pockets of both helical and fibrillar Ab1–40
.
As ‘‘artificial mAb (control, Fig. S64–S67, ESI†) are different from the ‘‘artificial
a-secretase’’4 exhibited the lowest binding affinity, it was also docked a-secretase’’-mediated proteolysis products. This result, along
into the helical C-terminal domain of APP. The interacting amino with the observation of the corresponding mass of C00 and
acids were Leu17, Val18, Phe19, and Phe20, as desired.
consistency of appearance of the fragments, collectively supports
To check the proteolytic efficiency, first ‘‘artificial a-secretases’’ the proteolytic role of ‘‘artificial a-secretases’’ excluding the
1–3 were co-incubated with mAb in 1 : 1 molar ratio at 37 1C and pH probability of fragmentation by the spectrometer.
7.4 (PBS) on a water bath. Aliquots were tested by a MALDI-TOF mass
DFT calculation revealed the most stable conformation of C00
spectrometer in a time-dependent manner. ‘‘Artificial a-secretases’’ (Fig. 2f) and C0000 (Fig. S68, ESI†). The bond distances between
Table 1 Peptides and their roles in the current study
Name
Peptide sequence
Role of the peptide
AaS1
AaS2
AaS3
AaS4
mAb
Ab1–40
Mutant DTP28
Ac-K(BnAdp)LVFF-NH2
Ac-K(BnAdp)GLVFF-NH2
BnAdp-AALVFF-NH2
BnAdp-K(BnAdp)K(BnAdp)LVFF-NH2
Ac-VHHQKLVFFGG-NH2
DAEFRHDSGYEVHHQKLVFFAEDVGSNKGAIIGLMVGGVV-NH2
GSSDSIGPLGYGKTVDHTKVNSPLSLFG-NH2
Site-selective cleavage
Site-selective cleavage
Site-selective cleavage
Site-selective cleavage
Prototype of Ab & APP
Target aggregating peptide
Negative control
Chem. Commun.
This journal is ©The Royal Society of Chemistry 2019

View Article Online
ChemComm
Communication
Fig. 3 (a) Some of the probable routes of proteolytic cleavage and (b) MALDI-
TOF mass spectra of Ab1–40 in the presence of ‘‘artificial a-secretases’’4 (1 : 2) at
ten days in PBS pH 7.4 at 37 1C. (c) Time-dependent persistence map of various
fragments. (d) Multiple proteolytic enzyme activities of ‘‘artificial a-secretases’’4.
(e) Plausible routes of self-degradation of NC into various fragments (NCFs).
Interestingly, ‘‘artificial a-secretase’’4 also cleaved the full-
length Ab1–40 site-selectively. Various plausible routes of proteolytic
cleavage (routes 1–10Ab–AaS4, Fig. 3a, b and Fig. S77–S83, ESI†) were
delineated from MALDI-TOF mass analyses, as before. A time-
dependent persistence map of the Ab fragments (Fig. 3c) demon-
strates the time-resolved manner of the programmed proteolysis.
‘‘Artificial a-secretase’’4 cleaved Ab1–40 between Lys16 and Leu17,
which is the cleavage site of not only a-secretase but also of ECE,
MMP2, MMP9 and P; and between Leu17 and Val18, the same as
ECE by the first ten days. After prolonged incubation (421 days),
‘‘artificial a-secretase’’4 cleaved Ab1–40 between Val12 and His13 (the
cleavage site of NEP and IDE), Asp7 and Ser8 (ACE cleavage site), and
Glu3 and Phe4 (NEP cleavage site). Thus, ‘‘artificial a-secretase’’4
Fig. 2 (a) Plausible routes of proteolytic cleavage of mAb into various frag-
ments. (b) MALDI-TOF mass spectra of mAb in the presence of ‘‘artificial
a-secretase’’2 (1 : 1) after ten days [inset: after three days] of incubation in PBS
pH 7.4 at 37 1C. (c–e) Time-dependent persistence map of various fragments of
mAb in the presence of ‘‘artificial a-secretases’’. (f) The most stable conformation
of the intermediate C00 (image generated by PyMOL) obtained from DFT
calculation [B3LYP, basis set: 6-31G]. AaS stands for ‘‘artificial a-secretase’’.
the carbonyl oxygen of the in situ generated amide and the nitrogens itself can cleave Ab1–40, mimicking the proteolytic action of multiple
of the backbone amides of Leu and Val are 7.9 Å and 9.2 Å, proteases. ‘‘Artificial a-secretase’’4 also cleaved Ab1–40 between Glu11
respectively (8.0 Å and 10.5 Å for C0000), which may be small enough and Val12 and between Val24 and Gly25, unlike any known proteo-
to facilitate the nucleophilic attack by the mentioned nitrogens on lytic enzymes. None of the fragments mentioned above were
the carbonyl groups that in turn become even closer and start observed by the time-dependent MALDI-TOF mass analysis of
forming a tetrahedral intermediate; this probably explains the site- Ab1–40 in the absence of ‘‘artificial a-secretase’’4 (Fig. S84–S87, ESI†);
selectivity of the proteolysis. It also indicates that the driving force of rather aggregation was noted, unlike the ‘‘artificial a-secretase’’4
such a nucleophilic attack could be the release of the strain of the treated samples, indicating its amyloid-solubilizing nature.
cyclic structure generated by the inter-strand aromatic interactions. None of the expected fragments possibly generated by ‘‘artificial
Furthermore, the intensity of the peak corresponding to ‘‘artificial a-secretase’’4 were present in the co-incubated (1 : 1) sample of
a-secretase’’2 diminished but did not disappear until ten days in a the mutant DTP28 (NC, Fig. S88–S97, ESI†) even after ten days.
time-dependent HPLC experiment (Fig. S69–S76, rt 19 min, ESI†). Instead, the self-degradation products of ‘‘artificial a-secretase’’4
The same is true of mAb (rt 13.4 min), which did not shift but and NC appeared. NC self-degraded at its Ser2 and Ser5 positions,
changed its shape. ESI-MS analysis of the eluent revealed the subsequently (Fig. 3e), which was confirmed by a parallel
presence of D00, I00, J00, O00 and mAb (B00). All of these results experiment in the absence of ‘‘artificial a-secretase’’4. Such
collectively indicate that the ‘‘artificial a-secretases’’ cleaved cleavage may proceed via route 11NC, which is unrelated to
mAb in a site-selective manner (both amide bonds Lys16–Leu17 ‘‘artificial a-secretase’’4. This result indicates the substrate-
and Leu17–Val18) first and then at the other positions over time. selective nature of ‘‘artificial a-secretase’’4. All of the mentioned
This journal is ©The Royal Society of Chemistry 2019
Chem. Commun.

View Article Online
Communication
ChemComm
function-specific mimic of these enzymes may ease degradation and
excretion of plaques.
In summary, a general method of de novo design of function-
specific protease-mimics is disclosed. The engineered peptides,
termed as ‘‘artificial a-secretases’’, indeed first sliced Ab peptide at
the a-secretase cleavage site followed by slicing at the cleavage sites
of the other Ab-degrading enzymes by a self-association driven
repetition of an imidation–hydrolysis–transamidation cycle.
One interesting feature of such ‘‘artificial a-secretases’’ is that
the cleavage sites can be manipulated by altering the position and
length of the pendant groups. Most importantly, the structurally
unrelated peptide was unaffected, indicating excellent substrate-
selectivity. Thus, we could achieve the proteolytic action of Ab
formation preventing enzyme (a-secretase) as well as Ab degrading
enzymes, all by ‘‘artificial a-secretases’’, presumably without affecting
other substrates of these enzymes in a non-catalytic manner. While
the amyloid-degrading enzymes cannot attack solvent inaccessible
insoluble amyloid, the self-association initiated mechanism favours
‘‘artificial a-secretases’’ to recognize and disrupt amyloid-generating
non-toxic Ab fragments. Therefore, design and development of such
constructs may be a promising approach for drug design against not
only AD but also other amyloidoses. However, further optimization
of the structure to improve the reactivity and pharmacokinetic
profile is required to establish their physiological relevance. Such a
strategy may also be extended to mimic other enzyme functions.
Fig. 4 Time-dependent (a and b) TEM images, and (c–e) Congo red stained
birefringence images of Ab1–40 (a and c) in the absence and (b, d and e) in
the presence of 5-fold ‘‘artificial a-secretase’’4, while (b and d) show
inhibition of amyloidosis, and (e) disruption of pre-existing amyloid (PBS
pH 7.4 at 37 1C).
routes of cascades of chemical reactions are logically-drawn,
probable routes. Many other modes of bond formation and
cleavage may be operating in the sample. All of the intermediates
and fragments may not be identified, but most of them could be
assigned, except those in parentheses in the figures.
Finally, the kinetics of amyloid accumulation of Ab1–40, and
the disruption of their preformed fibrils was monitored in the
absence and presence of ‘‘artificial a-secretase’’4 (Fig. 4 and
Fig. S98–S106, ESI†). The comparative thioflavin T aided fluores-
cence, CD, TEM, AFM, Congo-red birefringence, and DLS analyses
suggested that ‘‘artificial a-secretase’’4 not only inhibited fibrillation
of Ab1–40 but also dissolved preformed fibrils in a dose-dependent
manner. Such destruction of the fibril and amyloid is probably due
to the pre-programmed site-selective proteolysis by ‘‘artificial
a-secretase’’4. To the best of our knowledge this is the first report
of amyloid-digestion by an externally added chemical. Smaller,
soluble Ab oligomers produced by fibril disruption were unable to
rupture carboxyfluorescein dye-entrapped large unilamellar vesicles
(Fig. S106, ESI†), which is usually related to the non-toxic nature of
the disrupted smaller fibrils or oligomers.6,17
Conflicts of interest
There are no conflicts to declare.
Notes and references
1 M. F. Browner, W. W. Smith and A. L. Castelhano, Biochemistry,
1995, 34, 6602–6610.
While APP processing by b- and g-secretases generate disease-
causing Ab,18 the natural mechanism of AD prevention lies on the
site-specific APP cleavage by a-secretases to release the non-
amyloidogenic neuroprotective APP fragment, known as soluble
APP-a (sAPPa), preventing generation of neurotoxic and patho-
genic Ab.19 On the other hand, enzymes such as NEP, ECE, IDE,
P, ACE, and MMPs are known to degrade Ab aggregates to
facilitate their excretion.20 When these enzymes fail, plaques
form. Promoting a-secretase-mediated APP processing and Ab
2 W. J. Song and F. A. Tezcan, Science, 2014, 346, 1525–1528.
3 A. J. Burton, et al., Nat. Chem., 2016, 8, 837–844.
4 J. A. Hardy and G. A. Higgins, Science, 1992, 256, 184–185.
5 C. Soto, E. M. Sigurdsson, L. Morelli, A. R. Kumar, E. M. Castano and
B. Frangione, Nat. Med., 1998, 4, 822–826.
6 A. Paul, K. C. Nadimpally, T. Mondal, K. Thalluri and B. Mandal,
Chem. Commun., 2015, 51, 2245–2248.
7 N. Sehgal, et al., Proc. Natl. Acad. Sci. U. S. A., 2012, 109, 3510–3515.
8 C. G. Glabe, Neurobiol. Aging, 2006, 27, 570–575.
9 M. B. Lazarus, et al., Science, 2013, 342, 1235–1239.
10 J. Janetzko, S. A. Trauger, M. B. Lazarus and S. Walker, Nat. Chem.
Biol., 2016, 12, 899–901.
metabolism is, therefore, a promising therapeutic option for AD 11 L. O. Tjernberg, et al., J. Biol. Chem., 1996, 271, 8545–8548.
management.7,21 ADAM10 (a disintegrin and metalloproteinase), a
transmembrane protein which is approximately 750 amino acids
12 K. S. Kosik, Proc. Natl. Acad. Sci. U. S. A., 1999, 96, 2574–2576.
13 Y. Tian, B. Bassit, D. Chau and Y. M. Li, Nat. Struct. Mol. Biol., 2010,
17, 151–158.
long, is known as one of the main a-secretases.22 Its over-expression 14 S. Choe, et al., Nature, 1992, 357, 216–222.
15 O. Trott and A. J. Olson, J. Comput. Chem., 2010, 31, 455–461.
16 S. Vivekanandan, et al., Biochem. Biophys. Res. Commun., 2011, 411,
increased sAPPa, reduced plaque load, and improved learning and
memory in the AD mouse model.23 However, a wide variety of
312–316.
substrates of ADAM10 are linked to many physiological, as well 17 M. F. M. Sciacca, et al., Biophys. J., 2012, 103, 702–710.
18 G. He, et al., Nature, 2010, 467, 95–99.
19 F. S. Esch, et al., Science, 1990, 248, 1122–1124.
20 R. J. Baranello, et al., Curr. Alzheimer Res., 2015, 12, 32–46.
as pathological conditions, including not only the immune and
nervous systems but carcinogenesis has also been reported.24
Therefore, addressing AD through the up-regulation of a-secretases 21 K. Endres and T. Deller, Front. Mol. Neurosci., 2017, 10, 56.
22 R. Peron, et al., Pharmaceuticals, 2018, 11, 12.
23 R. Postina, et al., J. Clin. Invest., 2004, 113, 1456–1464.
24 B. Vincent and F. Checler, Curr. Alzheimer Res., 2012, 9, 140–156.
may improve AD but this also increases the risk of cancer and other
diseases.25 Similarly, up-regulation of other Ab-degrading enzymes
also may cause side effects. Therefore, developing the proposed 25 Y. Pan, et al., J. Mol. Endocrinol., 2012, 49, 21–33.
Chem. Commun.
This journal is ©The Royal Society of Chemistry 2019