Hetero-bifunctional PROteolysis TArgeting Chimeras (PROTACs) represent a new emerging class of small molecules designed to induce polyubiquitylation and proteasomal-dependent degradation of a target protein. Despite the increasing number of publications about the synthesis, biological evaluation, and mechanism of action of PROTACs, the characterization of the pharmacokinetic properties of this class of compounds is still minimal. Here, we report a study on the metabolism of a series of 40 PROTACs in cryopreserved human hepatocytes at multiple time points. Our results indicated that the metabolism of PROTACs could not be predicted from that of their constituent ligands. Their linkers' chemical nature and length resulted in playing a major role in the PROTACs' liability. A subset of compounds was also tested for metabolism by human cytochrome P450 3A4 (CYP3A4) and human aldehyde oxidase (hAOX) for more in-depth data interpretation, and both enzymes resulted in active PROTAC metabolism.
Hetero-bifunctional PROteolysis TArgeting Chimeras (PROTACs) represent a new emerging class of small molecules designed to induce polyubiquitylation and proteasomal-dependent degradation of a target protein. Despite the increasing number of publications about the synthesis, biological evaluation, and mechanism of action of PROTACs, the characterization of the pharmacokinetic properties of this class of compounds is still minimal. Here, we report a study on the metabolism of a series of 40 PROTACs in cryopreserved human hepatocytes at multiple time points. Our results indicated that the metabolism of PROTACs could not be predicted from that of their constituent ligands. Their linkers' chemical nature and length resulted in playing a major role in the PROTACs' liability. A subset of compounds was also tested for metabolism by humancytochrome P450 3A4 (CYP3A4) and humanaldehyde oxidase (hAOX) for more in-depth data interpretation, and both enzymes resulted in active PROTAC metabolism.
Rational
drug design represents an essential approach to optimize
time and cost in drug discovery and development,[1] but it remains a challenging task. Indeed, not only is
drug potency a critical feature, but also absorption, distribution,
metabolism, and excretion (ADME) properties require optimization by
modulating the chemical structure of the candidate. Drugs undergo
biotransformations, and thus the optimization of the drug structure per se could be useless when significant metabolic liability,
generating novel compounds (metabolites), occurs. In the last two
decades, many efforts have been made to decode and predict the metabolic
fate of drugs,[2] and in silico models,[3−6]in vitro assays,[2] and
hybrid approaches (i.e., innovative assays associated with software-assisted
data processing)[7,8] have been developed to identify
the “soft spots” of drugs. Despite signs of progress
in the field, all available ADME tools have been calibrated mainly
using the chemical space of small molecules, witnessing the outstanding
impact that the Lipinski rules[9] have had
in pharmaceutical research in the past. In the comfortable space of
small molecules, the accuracy and sensitivity of the models are usually
very good.[10−14] Nowadays, the chemical space of the drugs is quickly expanding,
ranging from peptides or peptidomimetics[15] to Proteolysis Targeting Chimeras (PROTACs)[16−26] and their analogues.[27−29] PROTACs can be defined as hetero-bifunctional molecules
that induce a ligand to bind with the protein of interest (POI), another
ligand to recruit an E3 ubiquitin ligase, and a linker to concatenate
the two ligands.[17] The formation of the
ternary complex composed of the POI, the PROTAC, and the E3 ligase
allows the E2 ubiquitin-conjugating enzyme to transfer ubiquitin to
the surface of the POI, inducing its proteasomal-dependent degradation.[30] One of the main advantages of PROTACs is that
they can degrade proteins regardless of their function, thus turning
into druggable also the “undruggable”, due to their
innovative mechanism of action.[16] Degradation
by PROTACs is a catalytic process, due to the dissociation of the
complex after polyubiquitination of the POI, indicating that PROTACs
can be recycled for successive rounds of degradation and thus used
at reduced doses.[31] Therefore, PROTACs
represent an innovative class of compounds that overcome traditional
limitations, opening a new therapeutic modality and, at the same time,
breaking the rules used so far with the potential to revolutionize
drug discovery. As extensively reviewed,[32−36] hundreds of PROTAC molecules have been developed
so far, targeting a wide range of different disease-related protein
targets. The entry in phase I clinical trial in 2019 of the first
two oral PROTACs (ARV-110 and ARV-471) for the treatment of metastatic
castration-resistant cancer and metastatic breast cancer (NCT0888612
and NCT04072952) has focused attention even more on this innovative
therapeutic paradigm.[37,38] Despite their intriguing capabilities,
PROTACs are characterized by a high molecular weight (600–1400
Da),[39] making the delivery and bioavailability
of PROTACs the most significant hurdles to overcome on the way to
the clinic.[40] Thus, better understanding
and prediction of the ADME properties of PROTACs represent an urgent
need for their rational design. To date, the evaluation of the ADME
properties of this class of compounds is still minimal, with only
a few studies on their experimental physical–chemical properties
available[41−43] and only one paper about PROTAC metabolism has been
published.[35] Nevertheless, preliminary
studies on small subsets of PROTACs, whose log P was experimentally measured, indicate that traditional in
silico tools for property prediction may fail,[41] likely due to their peculiar structural features
compared with traditional druglike compounds used for generation of
predictive models. Therefore, there is an urgent need to collect experimental
physicochemical and ADMEdata on PROTACs to shed light on their peculiar
behavior and to be used for modeling purposes. Here, focusing on the
human metabolism, a collection of 40 PROTACs (compounds 1–40, Supporting Information Table S1) was studied, assessing their metabolism in cryopreserved
human hepatocytes at multiple time points. Their enzymatic biotransformations
were also compared with those of the constituent ligands (compounds 41–46, Supporting Information Table S2). Both metabolic rate (half-life value)
and soft spot identification were investigated. In addition, a subset
of compounds was also tested for metabolism by humanCYP3A4 and hAOX for deeper data interpretation, representing the principal
isoenzyme involved in liver metabolism (including large substrates)[44,45] and one of the emerging enzymes in metabolism studies,[46−49] respectively. The complete data set with more experimental details
is provided in the Supporting Information.
Results and Discussion
Chemistry
The synthesis of JQ1-based
PROTACs 4–6, 8, 9, and 11 was accomplished according to Scheme . Briefly, derivative 43 was
coupled by
amidation reaction with the appropriated E3 ligase ligand properly
functionalized with linkers of different lengths in the presence of
1-[bis(dimethylamino)methylene]-1H-1,2,3-triazolo[4,5-b]pyridinium 3-oxid hexafluorophosphate (HATU) and N,N-diisopropylethylamine (DIPEA) at room temperature in
dimethylformamide (DMF).
Scheme 1
Syntheses of JQ1-Based PROTACs 4–6, 8,
9, and 11
Reagents and conditions: (a)
HATU, DIPEA, dry DMF, room temperature (rt).
Syntheses of JQ1-Based PROTACs 4–6, 8,
9, and 11
Reagents and conditions: (a)
HATU, DIPEA, dry DMF, room temperature (rt).Analogously, as shown in Scheme , CX4945-based PROTACs 12–20 were obtained by HATU-mediated amidation reaction between
derivative 44 and the appropriated E3 ligase-linker intermediate.
Scheme 2
Syntheses of CX4945-Based PROTACs 12–20
Reagents and conditions: (a)
HATU, DIPEA, dry DMF, rt.
Syntheses of CX4945-Based PROTACs 12–20
Reagents and conditions: (a)
HATU, DIPEA, dry DMF, rt.The olaparib-based
PROTACs 21–25 were synthesized as
depicted in Schemes and . For cereblon
(CRBN)-addressing PROTAC 21, derivative 45(50) was first reacted with 11-((tert-butoxycarbonyl)amino)undecanoic acid by N,N,N′,N′-tetramethyl-O-(1H-benzotriazol-1-yl)uronium
hexafluorophosphate (HBTU)-mediated amidation reaction leading to
Boc-protected intermediate 60, which after Boc-deprotection
reaction gave the key intermediate 61. Then, compound 62 was reacted with fluorothalidomide 62(51) in the presence of DIPEA at 70 °C in DMF
(Scheme ). For the
von Hippel-Lindau (VHL)-addressing PROTACs 22 and 23, derivative 45(50) was first reacted with the appropriate dicarboxylic acid monomethyl
ester linker by amidation reaction in the presence of HBTU and Et3N at room temperature in dry DMF to furnish intermediates 63–64. The successive basic hydrolysis
of methyl esters 63–64 gave intermediates 65–66, which in turn were finally coupled
with derivative 42(52) by HATU-mediated
amidation reaction to afford PROTACs 22 and 23, respectively (Scheme ).
Reagents and conditions: (a)
3-bromoprop-1-yne, K2CO3, KI, ACN, reflux; (b)
HBTU, Et3N, dry DMF, rt; (c) CuSO4, sodium ascorbate,
DMF/tBuOH/H2O (1:1:1), rt.For the VHL-addressing PROTACs 24 and 25 (Scheme ), an alkynyl
group was introduced by reacting derivative 45(50) with propargyl bromide, generating intermediate 67, or by coupling it with hex-5-ynoic acid, generating intermediate 68. Thus, final PROTACs 24 and 25 were obtained through the copper-assisted click reaction coupling
the alkynyl derivatives 67 and 68 with azide-containing
VHL derivative 69.The syntheses of AR ligand-based
CRBN-addressing PROTACs 26–32 are
shown in Scheme .
A substitution reaction between derivative 46(53) and tert-butyl
2-bromoacetate in the presence of K2CO3 in acetonitrile
at room temperature gave intermediate 70, which after
acidic hydrolysis afforded the key intermediate 71. The
successive amidation reaction in the presence of HATU and DIPEA at
room temperature in DMFbetween compound 71 and the appropriated
thalidomide-linker intermediate furnished PROTACs 26–31. For the synthesis of PROTAC 32, a first Mitsunobu
reaction between derivative 46(53) and the hydroxyl poly(ethylene glycol) (PEG)-linker gave the Boc-protected
intermediate 73. The Boc-deprotection reaction of 73 furnished intermediate 74, which was then
reacted with fluothalidomide 62(51) in the presence of DIPEA at 70 °C in DMF to afford PROTAC 32.
With the aim of covering a large
chemical diversity, various combinations of ligands for four target
proteins, ligands for two E3 ligases, and nineteen linkers were selected
to give a final data set of 40 PROTACs (Figure ). In particular, concerning the selection
of ligands for target proteins, the bromodomain and extra-terminal
(BET) inhibitor (+)-JQ1,[54] the casein kinase
2 (CK2) inhibitor CX4945,[55] the FDA-approved
poly(ADP-ribose) polymerase (PARP) inhibitor olaparib,[50] and an androgen receptor DNA-binding domain
binder[53] were used. Concerning the ligands
for E3 ligases, binders for cereblon (CRBN) and von Hippel-Lindau
(VHL) were selected, with these being two of the four most commonly
used E3 ligases in PROTAC synthesis, together with cell inhibitor
of apoptosis protein (cIAP) and mouse double minute 2 homolog (MDM2).[56] Finally, aliphatic, polyethylene glycol (PEG)-based,
and cyclic linkers were variably combined, modulating their length
and anchor point (Figure ). The chemical structures of the entire data set are provided
in the Supporting Information (Table S1). Among them, five compounds were commercially available (Supporting
Information Table S1), entries 1
(dBet1),[57]2 (dBet6),[58]3 (ARV-825),[59]7 (MZ1),[60] and 10 (ARV-771),[52] 15 were
kindly provided by Montelino Therapeutics Inc. (Supporting Information Table S1, entries 26–40), while the others were designed and synthesized in house
to increase chemical and structural variability.
Figure 1
Scheme of building blocks
characterizing the data set of tested
PROTACs.
Scheme of building blocks
characterizing the data set of tested
PROTACs.
Optimization of the Metabolic
Stability Assay for PROTACs
Metabolic stability of PROTACs
has been rarely discussed in the
literature, with only one paper published to date.[35] In that paper, Zhou et al.[35] evaluated the metabolism in mouse liver microsomes (phase I metabolism
only) of one PROTAC, known to degrade the BET proteins with thalidomide
as the CRBN ligand, after 20 and 40 min incubation times. The major
metabolite detected was proposed to be a hydroxylated product, with
the site of transformation occurring in the alkyl linker. In the present
study, the aim was to test a diverse data set of PROTACs for their
metabolic stability in cryopreserved human hepatocytes at multiple
time points within a time range of 4 h. Differently from liver microsomes,
cryopreserved human hepatocytes contain all phase I and II metabolic
enzymes, with all necessary cofactors, and are compatible with longer
incubation times.[61] In commonly used protocols
for metabolic stability assays, enzymatic reactions are quenched at
the desired time of incubation by adding an organic solvent (e.g.,
acetonitrile) to the enzyme-containing solution to induce protein
precipitation[62−64] and, after centrifugation, the water-containing supernatant
is collected and analyzed by liquid chromatography–mass spectrometry
(LC–MS). However, such a protocol would not prevent the nonenzymatic
degradation of PROTACs in the autosampler during LC–MS analysis.[65] Therefore, it is especially critical for high-throughput
screenings, in which a large number of samples are collected in the
autosampler simultaneously and analyzed in a long sequence of analysis.[65] In our studies on PROTAC metabolism, we reasoned
that this could be a critical point to be addressed also taking into
account that the rapid degradation of thalidomide[66] and thalidomide-containing PROTACs in aqueous solution
has already been reported elsewhere.[43] Therefore,
studies on the potential nonenzymatic degradation of substrates in
the autosampler during LC–MS analysis were conducted on the
commercial PROTAC 1 (dBet1), a potent Bromodomain-containing
protein 4 (BRD4) protein degrader that is composed of (+)-JQ1 linked
to thalidomide through an aliphatic linker. Compound 1 (dBet1) was incubated in three different solvents:
(1) in pure phosphate buffer at pH = 7.4 (PBS); (2) in a mixture of
PBS/acetonitrile (1:1 v/v), which is the most common composition of
the supernatant injected in the LC–MS instrument in metabolism
assays (named here PBS/ACN); and (3) in pure dimethyl sulfoxide (DMSO).
A fourth condition entailed the incubation of 1 (dBet1) in PBS/acetonitrile (1:1, v/v) and immediately after
the sample was dried under a nitrogen stream (to remove the solvent)
and redissolved in DMSO before injection in the LC–MS system.
The latter protocol is here named as PBS/ACN-DMSO and was designed
to evaluate whether the removal of the first solvent and the resuspension
in DMSO could prevent further degradation in the autosampler. Thus,
the nonenzymatic stability of 1 (dBet1)
in the four solvents and solvent mixtures was analyzed by LC–MS
for 12 h at 37 °C, with injections at 0, 3, 6, and 12 h. Figure illustrates the
results of the stability of 1 (dBet1). As
expected, compound 1 (dBet1) rapidly degrades
when stored in the autosampler in pure PBS (Figure A). Degradation was also observed in the
presence of the PBS/ACN solution, although to a lower extent, and
it occurred during the first three hours, becoming constant with time.
In DMSO, the solution of 1 (dBet1) became
very stable over time. When the PBS/ACN solution was removed by a
nitrogen stream and then replaced with an equal volume of DMSO (protocol
PBS/ACN-DMSO), the substrate degradation within the first three hours
was reduced to about 10%, remaining rather constant with time. The
formation of two degradation products was also monitored (Figure A), corresponding
to the hydrolysis of the phthalimide (degradation product D1) or glutarimide
groups (degradation product D2) in the thalidomide moiety (Figure A,B). When the PBS/ACN-DMSO
protocol was used, the formation of the degradation product was observed
in a limited amount, and their concentration was constant over time,
indicating that they formed during the PBS/ACN solution removal and
not during storage in the autosampler.
Figure 2
Nonenzymatic stability
of 1 (dBet1) in
the autosampler during LC–MS analysis acquisitions. (A) Percentage
of the remaining 1 and percentage of formation of the
degradation products resulted by the hydrolysis of the phthalimide
moiety (D1, orange) or of the glutarimide moiety (D2, green) at the
four time points (0, 3, 6, 12 h) in the different solutions. (B) Chemical
structure of 1 (dBet1), with highlighted
phthalimide (orange) and glutarimide (green) rings. (C) Comparison
of the metabolic profiles of 1 (dBet1) in
cryopreserved human hepatocytes for 4 h using the PBS/ACN (filled
circles) or the PBS/ACN-DMSO (empty circles) protocol.
Nonenzymatic stability
of 1 (dBet1) in
the autosampler during LC–MS analysis acquisitions. (A) Percentage
of the remaining 1 and percentage of formation of the
degradation products resulted by the hydrolysis of the phthalimide
moiety (D1, orange) or of the glutarimide moiety (D2, green) at the
four time points (0, 3, 6, 12 h) in the different solutions. (B) Chemical
structure of 1 (dBet1), with highlighted
phthalimide (orange) and glutarimide (green) rings. (C) Comparison
of the metabolic profiles of 1 (dBet1) in
cryopreserved human hepatocytes for 4 h using the PBS/ACN (filled
circles) or the PBS/ACN-DMSO (empty circles) protocol.Finally, protocols PBS/ACN and PBS/ACN-DMSO were compared
in a
real metabolism assay of 1 (dBet1). The
study of the metabolic stability of this PROTAC in cryopreserved human
hepatocytes was conducted at five time points (0, 30, 60, 120, and
240 min). Monitoring metabolic stability by a kinetic approach allows
not only the half-life calculation for the substrate but also reduction
in false positives in the characterization of metabolites. Samples
for each time point were analyzed by LC–MS/MS, and raw data
were analyzed using Mass-MetaSite software[7,67] in
the WebMetabase platform.[68−70]Figure C shows that the PBS/ACN-DMSO protocol significantly
increases the stability of 1 (dBet1) during
analysis. Therefore, the final method for the metabolic stability
assay used in this study included the PBS/ACN-DMSO protocol (see the
Methods section for the whole procedure), to reduce the risk of further
degradation of the substrate in the autosampler during analysis.
Metabolic Stability of Constituent Ligands for Tested PROTACs
Before performing an extensive study of the metabolic stability
of the 34 PROTACs, the ligands used in PROTACs’ design and
synthesis were tested for their metabolism in cryopreserved human
hepatocytes, to have a reference on the behavior of the units connected
in the final PROTAC structures. The same experimental protocol was
used for both PROTACs and ligands (see the methods section). Metabolic
stability was first expressed in terms of the half-life (t1/2) of
the parent compound, as it represents a commonly used parameter to
judge the intrinsic stability of a compound (Supporting Information Table S2).[61] Concerning
the structure of the tested ligands, an olaparib analogue lacking
the carboxycyclopropyl moiety (compound 45, Supporting
Information Table S2) was used a reference
for the PROTACs (21–25, Supporting
Information Table S1) targeting PARP. This
carboxycyclopropyl moiety, which represents a solution to improve
oral absorption,[71] was assumed to only
slightly affect the binding with the target[72] and was removed to allow the use of the unbound nitrogen of the
piperazine ring as the anchor point for the linker. As free compounds,
ligands used in this study for PROTAC design (41–46, Supporting Information Table S2) were characterized by good metabolic stability (t1/2 higher than
90 min), with the exception of the AR ligand (compound 46) showing a t1/2 of less than 20 min (Supporting Information Table S2).
Metabolic Stability of
PROTACs in Cryopreserved Human Hepatocytes
As for the free
ligands, metabolic stability of the whole set of
PROTACs in cryopreserved human hepatocytes was studied over a four-hour
incubation period (see the Method section). An example of the complete
kinetic behavior of the disappearance of the substrate and the appearance
of metabolites with time is shown for compound 7 (MZ1) in Figure S1 (Supporting Information).
Based on observed kinetic data for the entire data set, the half-life
for each PROTAC was calculated, and results are shown in the Supporting
Information Table S1.The analysis
of the half-lives led to a number of observations. First, for PROTACs
bearing ligands targeting BET, CK2, and PARP (compounds 1–25 in the Supporting Information Table S1), the use of the thalidomide moiety
as the E3 ligase binder led to lower t1/2 values compared to PROTACs
bearing the VHL ligand (i.e., 5 versus 10, 16 versus 18, 21 versus 22 in the Supporting Information Table S1). PROTAC 1 (dBet1) was the only
exception to this trend, possibly due to the very short linker, which
could hamper the interaction with metabolism-devoted enzymes. This
point will be further discussed in the following paragraphs. The lower
half-life values of thalidomide-containing PROTACs suggest that these
compounds, in addition to enzymatic transformation, might also undergo
partial nonenzymatic degradation during incubation time (the instability
of thalidomide was previously shown in Figure ), but a further analysis of this phenomenon
was beyond the scope of this paper. Concerning the PROTAC series containing
the AR ligand 46 (compounds 26–40 in the Supporting Information Table S1), they generally showed a higher susceptibility to metabolism
in cryopreserved human hepatocytes, with all t1/2 values lower than
100 min independent of the linker or the E3 ligase binder used. This
trend, associated with the low metabolic stability of AR ligand 46 as previously discussed (Supporting Information Table S2), suggested that the primary site(s)
of metabolism in this series is probably related to metabolic liabilities
in the compound 46 moiety rather than in the E3 ligase
part of the molecules. Therefore, for a deeper understanding of the
t1/2 values, the soft spot analysis was performed.
PROTAC Soft
Spot Identification
While metabolic stability
data expressed as the half-life of the parent compound represent a
valuable parameter to judge the intrinsic stability of a compound,[61] the identification of soft spots in a molecule
is crucial for the rational design of new and more stable compounds.[61] Although it is now always possible to identify
the exact site of metabolism by LC–MS/MS analysis, our study
allowed us to devise some general indications, which we believe will
be useful in the design of new PROTACs. Soft-spot analysis data are
provided in the Supporting Information Table S3.
Whole Differs from the Sum of Its Parts
Due to their
composed structure, the first natural comparison to make was whether
soft spots in a PROTAC could be predicted from the soft spots of the
free ligands. As an example, in Table , the observed metabolism of compounds 4 and 37, which are composed of totally different building
blocks, is compared with that of the corresponding constituent ligands.
Table 1
Half-Life for PROTACs and Single Ligand
Components upon Incubation in Cryopreserved Human Hepatocytes
Concerning PROTAC 4, enzymatic
degradation of the
thalidomide moiety was not observed, although a low metabolic stability
of pomalidomide 41 was detected, due to the opening of
the phthalimide and glutarimide rings. Similarly, the JQ1 moiety in
PROTAC 4 did not undergo the aliphatic hydroxylation
observed for ligand 43, probably due to an increase of
the steric hindrance of the JQ1 moiety site of metabolism (however,
for other JQ1-based PROTACs, traces of this metabolic route were observed,
see the Supporting Information Table S3). Nevertheless, 4 was highly metabolized in cryopreserved
human hepatocytes, with the soft spots being identified in the linker
and especially in its connection points with the ligands. Concerning
PROTAC 37, the liability of the morpholine ring present
in the AR ligand 46 was confirmed, as well as that of
two out of four points of the VHL ligand structure 42. However, the PEG-like linker also played an important role in the
enzymatic degradation of 37, being subjected to O-dealkylation and amide hydrolysis reactions. The preserved
liability of the AR ligand 46 moiety is also in agreement
with what was discussed in the previous paragraph for the AR degrader
series (26–40). In conclusion, translating
the well-known Aristotelian concept (“The Whole is Greater
than the Sum of its Parts”) to PROTACs’ metabolism,
one can derive that “the whole differs from the sum of its
parts”. Although this statement may seem trivial, it has a
strong impact on medicinal chemists because it confirms that PROTACs
represent totally independent chemical entities and that their metabolism
cannot be predicted from the one of the ligands used for their design
and synthesis. Table also shows that in PROTACs the most labile soft spots are represented
by the linker and the chemical connections used to join it to the
ligands.
Linker Effect
The linker commonly
plays an important
role in the biological activities and physicochemical properties of
PROTACs. With time, the chemical nature of linkers has been variably
modified, changing from initial peptide linkers[73] to (un)saturated alkane or PEG-like chains, to variably
functionalized linkers.[74] In particular,
PROTACs containing PEG-based linkers usually display a better solubility
profile when compared to those bearing alkyl linkers or even triazole-containing
linkers.[75] Indeed, introducing PEG moieties
(possessing a good safety profile)[76] represents
a commonly used strategy for improving pharmaceutical properties of
small molecules. Recently, piperidine-containing linkers have also
proved to be a good option to improve solubility.[74] Various bonds have been used to connect a linker to the
two ligands, including amide bonds, ether bonds, alkylamines, carbon–carbon
bonds, and click-chemistry products.[73] Finally,
the length of the linker has been extensively changed to adapt the
PROTAC to its biological function. In fact, on the one hand, if a
linker is too short, the simultaneous binding of the two ligands with
their targets will be hampered, and the formation of the ternary complex
will not occur. On the other hand, if the linker is too long, the
PROTAC will not efficiently move the target and the E3 ligase closer
to each other, and thus the target protein will not be ubiquitinated.[73,75,77,78] In the previous section, we discussed an example showing a significant
metabolic liability of the linkers (Table ). Therefore, one may ask how the tailoring
of the linker may affect the metabolic stability. We already mentioned
that the short linker in PROTAC 1 (dBet1) could be a reason for its high metabolic stability. Indeed, the
comparison of the t1/2 value of 1 (dBet1) with that for 2 (dBet6) indicated that
the extension of the linker from four to eight methylene units reduces
the t1/2 value from 135 to 18.2 min, respectively (Figure ).
Figure 3
Effect of the length
of linear linkers on metabolic stability.
The half-life values associated with each compound and expressed in
minutes are reported in brackets. Bond cleavages are illustrated as
dotted lines, while circles represent atoms subjected to oxidation
in a well-defined position by MS/MS fragmentation (blue) or in a defined
moiety by MS/MS fragmentation (pink). In the case of pink circles,
the displayed position was suggested by MetaSite predictions. Finally,
ellipses indicate dehydrogenation reactions, and the same color code
for circles was used.
Effect of the length
of linear linkers on metabolic stability.
The half-life values associated with each compound and expressed in
minutes are reported in brackets. Bond cleavages are illustrated as
dotted lines, while circles represent atoms subjected to oxidation
in a well-defined position by MS/MS fragmentation (blue) or in a defined
moiety by MS/MS fragmentation (pink). In the case of pink circles,
the displayed position was suggested by MetaSite predictions. Finally,
ellipses indicate dehydrogenation reactions, and the same color code
for circles was used.This finding could suggest
generalizing that, in the case of linear
linkers, a shorter aliphatic linker can be responsible for increased
metabolic stability, probably due to steric hindrance of the PROTAC
entering into the catalytic site of the metabolic enzymes. The hypothesis
was verified comparing two additional PROTACs with a linker composed
of four methylene units to long-linker analogues (Figure ). In particular, the AR-based
PROTACs (compounds 33 and 35, Figure ) confirmed the expected trend,
although with a reduced difference in half-life values possibly due
to the reduced length difference between the two linkers (from four
to six methylene units). However, replacing the AR ligand unit in
compounds 33 and 35 with the CX4945 (44) moiety, the increased length of the linear linker did
not show the hypothesized effect, with compounds 17 and 18 having equal t1/2 values. Soft spots suggest that, while
for the pairs 1/2 and 33/35 the longer linker seems more prone to the N-dealkylation reaction, in the case of 17/18 metabolism is very similar. Thus, this and other possible comparisons
of data in Table S1 (Supporting Information)
suggest that a very short linker can be commonly associated with improved
metabolic stability. However, our data show that, in some cases, PROTACs
composed of longer linkers could also be metabolically stable.Similarly, the nature of the linear linker and its binding moiety
seem not to heavily affect the metabolic stability. In Table , four PROTACs are displayed,
containing the AR ligand and pomalidomide moieties as ligands connected
by either an aliphatic or a PEG linker of the same length. In compounds 26 and 27, the hydroxyl thalidomide is linked
by an acetamide moiety, while in compounds in 30 and 31, the linker is directly connected through an amine group.
Despite this, compounds 27 and 31 share
a similar metabolic stability, compound 30 appears only
slightly more stable than the two PEG-containing PROTACs, and compound 26 resulted in being very unstable, with a t1/2 of only 8.4
min. Intriguingly, the greater instability of 26 is not
related to different soft spots, if compared to 30 (Table ), suggesting that
the different anchor point between the linker and the thalidomide
might modify the affinity with the metabolic enzyme(s). As revealed
by the comparison of the two PROTACs with an aliphatic linker, compounds 27 and 31, both possessing a PEG-like linker,
also share a similar soft spot pattern (Table ), which is mainly due to O-dealkylation reactions. It is noteworthy that, although in PEG-like
PROTACs the number of soft spots is higher due to the multiple O-dealkylation reactions, the overall metabolic rate is
not necessarily negatively affected. However, from the medicinal chemistry
perspective, a higher number of soft spots make the design of more
stable compounds by protection strategies more challenging.
Table 2
Soft Spots for PROTACs 26, 27, 30, and 31
Finally, six PROTACs with
various target and E3 ligase binders
and bearing cyclic linkers were synthesized to evaluate their effect
on metabolic stability. Among them, compounds 6, 11, and 20 were characterized by the presence
of a piperazine moiety in the linker, while compounds 24, 25, and 39 were endowed with a triazole
ring. A comparison with similarPROTACs bearing linear linkers (Table S1, Supporting Information) suggested that,
with the exception of 39, the presence of cyclic linkers
resulted in a higher metabolic stability. Soft spots for compounds 6, 11, 20, 24, 25, and 39 are illustrated in Figure . In the BET series (Table S1, entries 1–11 in the Supporting
Information), compounds 6 and 11 show higher
t1/2 values in the pomalidomide-containing PROTACs and VHL-containing
PROTACs, respectively. In addition, the significant increase of stability
observed for compound 11 can be explained by the short
length of its linker and by the attachment of piperazine to JQ1 through
an amide group, which hampers a second N-dealkylation
reaction (Figure ).
In the CK2 series (Table S1, entries 12–20
in the Supporting Information), VHL-containing compounds with linear
linkers 17–19 were all endowed with
short linkers and were characterized by a high metabolic stability.
Therefore, compound 20, bearing a short piperazine-containing
linker, was synthesized and tested, and again this compound turned
out to be the most stable in the series, although the t1/2 value was
slightly higher than the one for the linear analogue 17 (218 and 207 min, respectively). In the PARP series (Table S1, entries 21–25 in the Supporting
Information), a click-chemistry approach was applied to connect the
linker to the target binder through a triazole ring to give compounds 24 and 25. The triazole-containing PROTACs displayed
a much greater metabolic stability when compared to their linear analogues
(23 and 22, respectively). Indeed, as shown
in Table S1, t1/2 values for compounds 24 and 25 were greater than 240 min, with about
84% substrate left. Interestingly, an N-dealkylation
reaction occurred at the triazole, similar to that observed for the
piperazine-containing linkers (Figure ). Finally, as anticipated, compound 39 turned out to be the only PROTAC endowed with a cyclic moiety with
a lower metabolic stability when compared with its linear analogue 33, considering the triazole ring as a bioisosteric substitution
of the amide linkage (Table S1, Supporting
Information). Among the detected soft spots (Figure ), the main site of metabolism was localized
at the attachment point of the AR ligand, where the occurrence of
an O-dealkylation reaction is likely favored by the
presence of two aromatic moieties nearby.
Figure 4
Effect of the cyclic
linkers on metabolic stability. The half-life
values associated with each compound and expressed in minutes are
reported in brackets. Bond cleavages are illustrated as dotted lines,
while circles represent atoms subjected to oxidation in a well-defined
position by MS/MS fragmentation (blue) or in a defined moiety by MS/MS
fragmentation (pink). In the case of pink circles, the displayed position
was suggested by MetaSite predictions. Finally, ellipses indicate
dehydrogenation reactions, and the same color code for circles was
used.
Effect of the cyclic
linkers on metabolic stability. The half-life
values associated with each compound and expressed in minutes are
reported in brackets. Bond cleavages are illustrated as dotted lines,
while circles represent atoms subjected to oxidation in a well-defined
position by MS/MS fragmentation (blue) or in a defined moiety by MS/MS
fragmentation (pink). In the case of pink circles, the displayed position
was suggested by MetaSite predictions. Finally, ellipses indicate
dehydrogenation reactions, and the same color code for circles was
used.
Effect of the Linker’s
Site of Attachment
In
the PROTAC design, the site of attachment of the linker to the ligands
is typically selected by analyzing the solvent-exposed areas on ligand-protein
binding structures.[73]Figure shows that the selection of
the site of attachment might have an impact on the overall metabolic
degradation of the PROTAC. Indeed, compound 35 was less
stable than 40, although the identified soft spots were
comparable.
Figure 5
Effect of the linker’s site of attachment on PROTAC stability.
Half-life values expressed in minutes are shown in brackets. Bond
cleavages are illustrated as dotted lines, while the pink ellipse
indicates a dehydrogenation reaction occurring in the linker moiety
(the exact position of the soft spot was not elucidated by MS/MS fragmentation,
and the displayed position reflects the most probable soft spot according
to MetaSite predictions).
Effect of the linker’s site of attachment on PROTAC stability.
Half-life values expressed in minutes are shown in brackets. Bond
cleavages are illustrated as dotted lines, while the pink ellipse
indicates a dehydrogenation reaction occurring in the linker moiety
(the exact position of the soft spot was not elucidated by MS/MS fragmentation,
and the displayed position reflects the most probable soft spot according
to MetaSite predictions).
PROTAC Degradation by CYP3A4
CYP3A4 represents a major
isozyme in the human liver and is known to metabolize a larger variety
of xenobiotics.[44] An important feature
of CYP3A4 is its plasticity, which allows it to accommodate an extensive
substrate in its binding site.[45] Based
on these considerations, we assumed that CYP3A4 could be responsible
for most of the phase I metabolism observed in the cryopreserved human
hepatocyte data. Therefore, six PROTACs with variable combinations
of ligands and linkers were selected, and their metabolism by incubation
in the presence of recombinant CYP3A4 for 60 min was studied. Figure shows the soft spots
detected, while t1/2 values are provided in brackets. In the experimental
conditions used (see the Methods section), all tested PROTACs were
significantly metabolized, with half-lives well below 30 min. In addition,
the already discussed instability at the level of the linker was proved
to be caused by CYP3A4.
Figure 6
Soft-spot identification of six PROTACs tested
for metabolism by
CYP3A4. Half-life values expressed in minutes are shown in brackets.
Bond cleavages are illustrated as dotted lines. The pink ellipse indicates
that a dehydrogenation reaction occurred in the linker. Since the
MS/MS fragmentation was not enough to define the exact position of
the dehydrogenation’s soft spot, its most probable position
was suggested by MetaSite predictions.
Soft-spot identification of six PROTACs tested
for metabolism by
CYP3A4. Half-life values expressed in minutes are shown in brackets.
Bond cleavages are illustrated as dotted lines. The pink ellipse indicates
that a dehydrogenation reaction occurred in the linker. Since the
MS/MS fragmentation was not enough to define the exact position of
the dehydrogenation’s soft spot, its most probable position
was suggested by MetaSite predictions.
PROTAC Degradation by hAOX
In addition
to CYP-mediated metabolism, increasing importance has been attributed
to humanaldehyde oxidase (hAOX), a cytosolic drug-metabolizing
enzyme expressed in the human liver.[12,79−81] Indeed, strategies designed to reduce CYP-mediated metabolism have
resulted in increasing drug reactivity toward AOX.[46,47] As a consequence, several compounds have failed due to undetected hAOX-mediated oxidation (e.g., BIBX1382, RO-1, FK3453, carbazeran).[48,49] Two reactions are reported to be catalyzed by hAOX: (1) the oxidation of a wide range of azaaromatic scaffolds at
the unsubstituted carbon in ortho to the nitrogen
(usually the most electron-deficient);[82] and (2) the hydrolysis of amides although a few examples have been
reported so far.[80,81,83] Since the PROTACs commonly contain amide groups and heteroaromatic
rings, two PROTACs were selected to be screened for hAOX metabolism. Compounds 33 and 34 were
selected as each one contained three amide groups that might be liability
spots for hAOX metabolism. In addition, they also
contain one 4-aryl or 5-aryl substituted thiazole ring (in the AR
ligand moiety (46) and the VHL ligand moiety (42), respectively). Although five-term moieties are commonly considered
not prone to be metabolized by hAOX unless it is
fused with a phenyl ring to give a benzothiazole,[84−86] one exception
has been reported by Arora et al.,[87] showing
that 2H-oxazoles substituted at the C-4 or C-5 position with variably
decorated phenyl rings can undergo oxidation by mouse cytosolic AOX
to give the corresponding 2-oxazolones. Therefore, we hypothesized
that a similar oxidation pattern could occur in the selected PROTACs
although this reaction has not been reported for substituted thiazoles
to date. Compounds 33 and 34 were therefore
incubated in human liver cytosol for 30, 60, and 90 min in the absence
and presence of hydralazine, a selective inhibitor of hAOX,[88] and the kinetics data are illustrated
in Figure S2 (Supporting Information).
Two reactions occurred for both compounds, the hydrolysis of an amide
and an oxidation, both in the VHL ligand moiety (Figure A).
Figure 7
Metabolism of 33 and 34 in human liver
cytosol. (A) Soft-spot identification, with filled circles indicating
metabolism by hAOX and empty circles indicating metabolism
by other enzymes. Half-life values expressed in minutes are shown
in brackets. (B) Pose of 33 in the hAOX cavity according to MetaSite prediction with protein in the surface
mode. (C) Pose of 33 in the hAOX cavity
according to MetaSite prediction highlighting the main interacting
residues and the molybdenum pyranopterin cofactor (MoCo).
Metabolism of 33 and 34 in human liver
cytosol. (A) Soft-spot identification, with filled circles indicating
metabolism by hAOX and empty circles indicating metabolism
by other enzymes. Half-life values expressed in minutes are shown
in brackets. (B) Pose of 33 in the hAOX cavity according to MetaSite prediction with protein in the surface
mode. (C) Pose of 33 in the hAOX cavity
according to MetaSite prediction highlighting the main interacting
residues and the molybdenum pyranopterin cofactor (MoCo).While the hydrolysis of the amide was observed also in the
presence
of hydralazine, indicating that an enzyme other than hAOX is responsible for this cleavage, the hydroxylated product was
formed only in the absence of the selective inhibitor (Supporting
Information Figures S2 and S3). In addition,
the MS/MS fragment ion with m/z 218.0637
revealed that the soft spot for oxidation is located in the 4-methyl-5-phenyl-thiazole
moiety that, in its unoxidized form, shows a fragment ion with m/z of 202.0685 (Supporting Information, Figure S2). Based on the findings by Arora et
al.,[87] it is likely to assume that the
metabolites formed for both 33 and 34 are
the corresponding 2-thiazolones on the VHL moiety. This finding is
noteworthy since VHL is one of the most common E3 ligases exploited
for the PROTAC strategy and, therefore, additional larger-scale studies
are currently in progress. Finally, MetaSite software was used to
generate the most probable binding mode of 33 in the hAOX cavity. Figure B illustrates that 33 nicely fits the hAOX cavity and that the pose exposing the 4-methyl-5-phenyl-thiazole
toward the molybdenum pyranopterin cofactor (MoCo) is stabilized by
several favored interactions, including a H-bond between the hydroxyl
group of the pyrrolidine moiety and Glu882 and a π–π
stacking between the thiazole ring in the AR ligand moiety and Phe885.
These two residues were recently hypothesized to stabilize other hAOX substrates (Figure C).[80,82]
Conclusions
This
study represents the first analysis of the metabolic stability
of PROTACs applied to a collection of compounds with large chemical
variability. Metabolism assays were first performed in cryopreserved
human hepatocytes that, containing all hepatic drug-metabolizing enzymes
and cofactors at physiological levels, represent the “gold
standard” even for the early screening of metabolic stability.
Additional studies were conducted to evaluate whetherCYP3A4 and hAOX could be involved in the observed metabolic biotransformations.
A comprehensive analysis of the data in terms of half-life values
and soft spot identification allowed us to highlight general trends
in PROTAC metabolism. The linker resulted in being the most liable
moiety in a PROTAC molecule. Its instability is mainly localized at
the attachment points to ligands, involving N-dealkylation
and amide hydrolysis reactions. Such reactions also occurred in CYP3A4
incubation, indicating that this isoform can play an essential role
in PROTAC degradation. In the case of PEG-like linkers, a large number
of O-dealkylation reactions was observed, indicating
that multiple fragmentation points are possible. Nevertheless, the
most significant number of soft spots in PEG-like linker-based PROTACs
compared to aliphatic-based ones seems not to negatively affect the
overall metabolic stability of a compound, with the half-life values
being comparable or even better. From the medicinal chemistry perspective,
however, the soft spot protection strategies might be more challenging
for PEG-like-based PROTACs. The length of the linker also played a
role in metabolic stability, with 4-unit linkers being very stable
compared to longer ones. Unfortunately, longer linkers are mostly
used in PROTAC design for activity optimization, thus limiting the
application of this evidence. The use of linkers endowed with cyclic
moieties may represent a good strategy to increase metabolic stability;
however, we demonstrated that exceptions are possible, as in the case
of compound 39. Concerning the E3 ligands, thalidomide-based
PROTACs suffered from nonenzymatic degradation in aqueous solutions.
Nevertheless, the protocol developed in this study reduces this degradation
during storage in the autosampler for LC–MS analysis. This
will not prevent the eventually occurring nonenzymatic hydrolysis
of the thalidomide moiety during incubation in cryopreserved human
hepatocytes, but it will improve reproducibility of the results. When
the VHL ligand is used, we discovered that PROTACs could undergo hAOX metabolism at the 5-phenyl-thiazole moiety. This finding
not only represents the first evidence of the metabolism on a substituted
thiazole by hAOX based on our knowledge but also
indicates that further studies are required to verify the affinity
of the hydroxylated metabolism with VHL aimed at evaluating the impact
on PROTAC efficiency. The metabolic degradation of PROTACs by hAOX on a large scale is currently under investigation.
To conclude, we believe that the study herein reported represents
a solid base to start considering metabolism in rational PROTAC design.
Experimental Section
General
Unless otherwise noted, starting materials,
reagents, and solvents were purchased from commercial suppliers and
were used as received without further purification.Compound 41 was purchased from Fluorochem, compound 42 was prepared according to the reported procedure,[52] while compounds 43 and 44 were
purchased from Ambeed and Fluorochem, respectively. PROTACs 26–40 were kindly provided by Montelino
Therapeutics Inc.Reactions were routinely monitored by thin-layer
chromatography
(TLC) performed on a silica gel 60 F254 (layer 0.2 mm)
precoated aluminum foil (with a fluorescent indicator UV254) (Sigma-Aldrich).
Developed plates were air-dried and visualized under UV light (254/365
nm) or using KMnO4, ninhydrin, or phosphomolybdic acid
stain solutions. Flash column chromatography was performed on Merck
silica gel 60 (mesh 230–400). 1H NMR and 13C NMR spectra were recorded at room temperature at 400 and 101 MHz,
respectively, on a Bruker Avance 400 spectrometer using tetramethylsilane
(TMS) or residual solvent peak as the internal standard. Chemical
shifts are reported in ppm (δ), and the coupling constants (J) are given in Hertz (Hz). Peak multiplicities are abbreviated
as follows: s (singlet), bs (broad singlet), d (doublet), dd (double
doublet), t (triplet), dt (double triplet), q (quartet), p (pentet),
and m (multiplet). High-resolution mass spectroscopy (HRMS) analyses
were carried out on the Agilent Technologies 6540 UHD Accurate Mass
Q-TOF LC–MS system. The purity of all final compounds was confirmed
to be >95% by ultraperformance liquid chromatography-tandem mass
spectrometry
(UPLC-MS). The analyses were carried out according to the method listed
below. The mobile phase was a mixture of water (solvent A) and acetonitrile
(solvent B), both containing formic acid at 0.1%. Method: Acquity
UPLC BEH C18 1.7 μm (C18, 150 mm × 2.1 mm) column at 40
°C using a flow rate of 0.65 mL/min in a 10 min gradient elution.
Gradient elution was as follows: 99.5:0.5 (A/B) to 5:95 (A/B) over
8 min, 5:95 (A/B) for 2 min, and then reversion back to 99.5:0.5 (A/B)
over 0.1 min. The UV detection is an averaged signal from a wavelength
of 190–640 nm, and mass spectra are recorded on a mass spectrometer
using positive-mode electrospray ionization. The chemical names were
generated using ChemBioDraw 12.0 from CambridgeSoft.
General Procedure
A: HATU-Mediated Amidation
Under
a nitrogen atmosphere, to a stirred solution of the appropriate carboxylic
acid (1.0 equiv), suitable amine (1.0 equiv), and DIPEA (4.0 equiv)
in dry DMF, HATU (1.25 equiv) was added and the reaction mixture was
stirred at room temperature (1–18 h). The mixture was poured
in ice-water, yielding a precipitate collected by filtration. When
no precipitate formed, the mixture was extracted with EA (×3)
and the reunited organic phases were washed with water (×3) and
brine (×3), dried over Na2SO4, and evaporated
to dryness. The crude was purified as described below.
Under a nitrogen atmosphere,
a solution of 42 (0.200 g, 0.416 mmol), 3-(1-tert-butoxycarbonylpiperazin-4-yl)propionic acid (0.124
g, 0.458 mmol), HATU (0.209 g, 0.520 mmol), and DIPEA (0.3 mL, 1.664
mmol) in dry DMF (2.0 mL) was stirred at room temperature for 1 h.
Then, the reaction mixture was poured in ice-water and extracted with
EA (×3). The reunited organic phases were washed with water (×2)
and brine, dried over Na2SO4, and evaporated
to dryness. The crude residue was purified by flash column chromatography
on SiO2 (DCM/acetone/MeOH, 75:20:5) to give tert-butyl-4-(3-(((S)-1-((2S,4R)-4-hydroxy-2-(((S)-1-(4-(4-methylthiazol-5-yl)phenyl)ethyl)carbamoyl)pyrrolidin-1-yl)-3,3-dimethyl-1-oxobutan-2-yl)amino)-3-oxopropyl)piperazine-1-carboxylate
(0.145 g, 51% yield) as a colorless oil. 1H NMR (400 MHz,
CDCl3) δ 9.02 (bs, 1H), 8.67 (s, 1H), 7.57 (d, J = 7.9 Hz, 1H), 7.56–7.30 (m, 4H), 5.13–5.01
(m, 1H), 4.78 (t, J = 8.0 Hz, 1H), 4.47 (s, 1H),
4.44–4.36 (m, 1H), 4.21 (d, J = 11.2 Hz, 1H),
3.74–3.14 (m, 6H), 2.91–2.30 (m, 11H), 2.14–2.04
(m, 1H), 1.78–1.63 (m, 1H), 1.53–1.41 (m, 12H), 1.07
(s, 9H). HRMS (ESI) m/z [M + H]+
calcd for C35H52N6O6S
685.37473, found 685.37551. UPLC retention time: 3.716 min.Then, the solution of the obtained compound (0.140 g, 0.204 mmol)
in 4.0N HCl in dioxane (1.5 mL) was stirred at room temperature for
2 h. The solvent was evaporated to dryness, and the solid was tritured
with DEE and collected by filtration, yielding 52 as
a white solid (0.124 g, 98% yield). 1H NMR (400 MHz, MeOD)
δ 9.86 (s, 1H), 7.68–7.46 (m, 4H), 5.09–4.99 (m,
1H), 4.66–4.53 (m, 2H), 4.50–4.38 (m, 1H), 3.96 (d, J = 10.9 Hz, 1H), 3.82–3.47 (m, 12H), 2.93 (t, J = 6.5 Hz, 2H), 2.61 (s, 3H), 2.25 (dd, J = 12.8, 7.8 Hz, 1H), 2.01–1.90 (m, 1H), 1.67–1.47
(m, 3H), 1.18–0.95 (m, 9H). HRMS (ESI) m/z [M + H]+ calcd for C30H44N6O4S 585.32230, found 585.32540. UPLC retention time: 2.852
min.
Under a nitrogen atmosphere, a solution
of 45(50) (0.267 g, 0.663 mmol),
methyl 11-((tert-butoxycarbonyl)amino)undecanoate[92] (0.200 g, 0.663 mmol), HBTU (0.314 g, 0.829
mmol), and Et3N (0.185 mL, 0.1.327 mmol) in dry DMF (2.5
mL) was stirred at room temperature for 3 h. The reaction mixture
was poured in ice-water and extracted with EA (×3). The reunited
organic phases were washed with water (×2) and brine, dried over
Na2SO4, and evaporated to dryness. The crude
residue was purified by flash column chromatography on SiO2 (DCM/MeOH, 97:3) to give 60 (0.266 g, 62% yield) as
a white solid. 1H NMR (400 MHz, CDCl3) δ
10.66 (s, 1H), 8.47 (d, J = 6.2 Hz, 1H), 7.85–7.67
(m, 3H), 7.37–7.29 (m, 2H), 7.04 (t, J = 8.6
Hz, 1H), 4.54 (bs, 1H), 4.29 (s, 2H), 3.51 (dd, J = 125.0, 74.0 Hz, 8H), 3.09 (t, J = 7.0 Hz, 2H),
2.41–2.23 (m, 2H), 1.44 (s, 12H), 1.27 (s, 13H).
A solution of 60 (0.260 g,
0.400 mmol) in 4.0N HCl in dioxane was stirred at room temperature
for 3 h. The solvent was evaporated to dryness, and the solid was
tritured with DEE and collected by filtration, yielding 61 as a white solid (0.218 g, 93% yield).Under a nitrogen atmosphere,
a mixture of 61 (0.117 g, 0.199 mmol), 62(51) (0.055 g, 0.199 mmol), and DIPEA (87
μL, 0.498 mmol) in dry DMF (1.0 mL) was stirred at 70 °C
for 2 h. After cooling to room temperature, the reaction mixture was
poured in ice-water, yielding a yellow precipitate, which was collected
by filtration and purified by flash column chromatography on SiO2 (DCM/acetone, 7:3 to 6:4) followed by preparative TLC purification
on SiO2 (DCM/MeOH, 97:3) to give 21 as a fluorescent-yellow
solid (0.005 g, 3% yield). 1H NMR (400 MHz, CDCl3) δ 10.30 (s, 1H), 10.12 (s, 1H), 9.02 (s, 1H), 8.63 (s, 1H),
8.46 (s, 1H), 7.87–7.67 (m, 3H), 7.55–7.43 (m, 1H),
7.34–7.30 (m, 1H), 7.09 (d, J = 7.1 Hz, 1H),
6.88 (d, J = 8.5 Hz, 1H), 6.23 (s, 1H), 4.93 (dd, J = 12.0, 5.2 Hz, 1H), 4.28 (s, 2H), 3.86–3.52 (m,
6H), 3.44–3.30 (m, 2H), 3.30–3.21 (m, 2H), 2.94–2.69
(m, 3H), 2.39–2.25 (m, 2H), 2.15–2.07 (m, 1H), 1.71–1.59
(m, 4H), 1.27 (d, J = 11.5 Hz, 12H). HRMS (ESI) m/z [M + H]+ calcd for C44H48FN7O7 806.367752, found 806.36739.
UPLC retention time: 5.535 min.
Under a nitrogen atmosphere, a solution
of 45(50) (0.150 g, 0.372 mmol),
12-methoxy-12-oxododecanoic acid (0.091 g, 0.372 mmol), HBTU (0.176
g, 0.465 mmol), and Et3N (0.1 mL, 0.745 mmol) in dry DMF
(1.5 mL) was stirred at room temperature for 3 h. The reaction mixture
was poured in ice-water and extracted with EA (×3). The reunited
organic phases were washed with water (×2) and brine, dried over
Na2SO4, and evaporated to dryness. The crude
residue was purified by flash column chromatography on SiO2 (DCM/MeOH, 97:3) to give 63 (0.113 g, 51% yield) as
a yellow oil, which solidified upon standing. 1H NMR (400
MHz, CDCl3) δ 10.68 (bs, 1H), 8.52–8.40 (m,
1H), 7.83–7.67 (m, 3H), 7.40–7.28 (m, 2H), 7.04 (t, J = 8.8 Hz, 1H), 4.29 (s, 2H), 3.89–3.69 (m, 2H),
3.66 (s, 3H), 3.61–3.22 (m, 6H), 2.39–2.26 (m, 4H),
1.67–1.55 (m, 4H), 1.33–1.21 (m, 12H). HRMS m/z [M + Na]+ calcd for C33H41FN4O5 615.29532, found
615.29651. UPLC retention time: 5.408 min.
To a stirring solution of 63 (0.097 g, 0.164 mmol) in THF (1.0 mL) at 0 °C was added a solution
of LiOH monohydrate (0.069 g, 1.636 mmol) in water (0.5 mL). The reaction
mixture was stirred at room temperature for 6 h. The organic solvent
was evaporated under reduced pressure, the residue was diluted with
water (10 mL), and at 0 °C it was acidified with 2 N HCl (pH
= 3) to afford 65 as a white precipitate, collected by
filtration and dried (0.084 g, 88% yield). 1H NMR (400
MHz, CDCl3) δ 11.79 (bs, 1H), 8.44 (d, J = 6.8 Hz, 1H), 7.89–7.63 (m, 3H), 7.41–7.28 (m, 2H),
7.03 (t, J = 8.5 Hz, 1H), 4.29 (s, 2H), 3.90–3.18
(m, 8H), 2.46–2.18 (m, 4H), 1.67–1.53 (m, 4H), 1.26
(s, 12H). HRMS m/z [M + H]+ calcd for C32H39FN4O5 579.29773, found 579.29912. UPLC retention time: 4.676 min.
Under a nitrogen atmosphere,
a solution of 45 (0.150 g, 0.372 mmol), 5-hexynoic acid
(0.042 g, 0.372 mmol), HBTU (0.175 g, 0.465 mmol), and Et3N (0.1 mL, 0.744 mmol) in dry DMF (2.0 mL) was stirred at room temperature
for 2 h. Then, the reaction mixture was poured in ice-water and extracted
with EA (×3). The reunited organic phases were washed with water
(×2) and brine, dried over Na2SO4, and
evaporated to dryness. The crude residue was purified by flash column
chromatography on SiO2 (DCM/MeOH, 96:4) to afford the title
compound (0.112 g, 66% yield) as a colorless oil. 1H NMR
(400 MHz, CDCl3) δ 11.70–11.51 (m, 1H), 8.46
(d, J = 6.2 Hz, 1H), 7.83–7.65 (m, 3H), 7.40–7.28
(m, 2H), 7.02 (t, J = 9.0 Hz, 1H), 4.29 (s, 2H),
3.91–3.66 (m, 3H), 3.67–3.50 (m, 3H), 3.45 (s, 1H),
3.39–3.16 (m, 2H), 2.50 (t, J = 7.3 Hz, 2H),
2.33–2.22 (m, 2H), 1.91–1.78 (m, 2H). HRMS (ESI) m/z [M + H]+ calcd for C26H25FN4O3 461.19889, found 461.19983. UPLC
retention time: 3.763 min.
A solution of 4.0N HCl in
dioxane (2.0 mL) was added to 73 (0.100 g, 0.189 mmol),
and the resulting suspension was stirred at room temperature overnight.
The solvent was evaporated to dryness, and the residue was tritured
with DEE and collected by filtration to afford 74 (0.080
g, 91% yield) as a white solid. 1H NMR (400 MHz, MeOD)
δ 7.49–7.40 (m, 1H), 7.28 (s, 1H), 7.04 (dd, J = 16.9, 9.0 Hz, 1H), 4.25 (s, 1H), 3.86–3.50 (m,
18H), 3.02 (s, 2H).
General Procedure A (6 h) was followed
using 74 (0.044 g, 0.095 mmol) and 62(51) (0.024 g, 0.086 mmol), to afford the title compound
as a yellow solid (0.06 g, 54% yield) following purification by flash
column chromatography on SiO2 (DCM/acetone, 9:1). 1H NMR (400 MHz, CDCl3) δ 8.21 (s, 1H), 7.89
(ddd, J = 8.7, 6.3, 2.1 Hz, 1H), 7.58 (s, 1H), 7.52–7.44
(m, 1H), 7.11 (d, J = 7.1 Hz, 1H), 6.99–6.88
(m, 2H), 6.51 (t, J = 5.3 Hz, 1H), 4.90 (dd, J = 12.1, 5.3 Hz, 1H), 4.40–4.24 (m, 2H), 3.98–3.81
(m, 6H), 3.79–3.62 (m, 6H), 3.59–3.38 (m, 6H), 2.96–2.65
(m, 3H), 2.19–2.03 (m, 1H). HRMS (ESI) m/z [M + H]+ calcd for C32H33F2N5O8S 686.20907, found 686.20855. UPLC retention
time: 5.749 min.
Metabolic Stability in Cryopreserved Human
Hepatocytes
Cryopreserved human hepatocytes (pooled suspension
hepatocytes, Gibco)
were thawed placing in a 37 °C shaking water bath according to
the manufacturer’s specifications and resuspended in Williams
E medium (WEM) to have 1 × 106 cells/mL. Samples with
the test compound at 1 μM were incubated at 37 °C, and
aliquots of 50 μL were collected at 0, 10, 30, 60, 120, and
240 min. The incubations were quenched 1:3 with ice-cold acetonitrile
(containing 1 μM labetalol as the internal standard). Samples
were then centrifuged at 12000 rpm for 5 min at 4 °C. The supernatant
was concentrated by evaporation under a nitrogen stream and reconstituted
with DMSO. Blank was prepared similarly but in the absence of the
investigated compounds.
Metabolic Stability in CYP3A4 and Human Liver
Cytosol
Metabolism of the selected compounds was evaluated
upon incubation
with human liver cytosol (Gibco) and supersome recombinant CYP isoforms
3A4 expressed in baculovirus-infected insect cells (Gibco). For metabolism
in CYP3A4, test compounds (1 μM) were incubated in 0.1 M potassium
phosphate buffer (pH 7.4) and allowed to equilibrate at 37 °C.
Freshly prepared 25 mM nicotinamide adenine dinucleotide phosphate
(NADPH, 1 mM final concentration) was added to the incubation mixture
to start the reaction. For metabolism in human liver cytosol, reactions
were initiated by adding test compounds (1 μM) to the mix reaction
constituted by the human liver cytosol (1 mg/mL) prewarmed in potassium
phosphate buffer (pH 7.4). Similarly, in the inhibition experiments
of the AOX1, test compounds were incubated in the presence or absence
of the selective inhibitor hydralazine (50 μM). In both procedures,
aliquots (50 μL) of the incubation mix were removed at 0, 10,
20, 30, and 60 min (CYP3A4) or 0, 30, 60, and 90 min (human liver
cytosol) and added to 50 μL of ice-cold acetonitrile quench
solution (internal standard, 1 μM labetalol) to stop the reaction.
Samples were then centrifuged at 12 000 rpm for 5 min at 4
°C. The supernatant was concentrated by evaporation under a nitrogen
stream and reconstituted in DMSO. The blank was prepared similarly
but in the absence of the investigated compounds. Samples were then
analyzed by LC–MS/MS (see the Analytical procedure).
Stability
of Compound 1 (dBet1) in
Four Different Solutions during LC–MS Acquisitions
dBet1 (1) (10 mM in DMSO) was diluted 1:1000 in PBS/ACN
(50:50, vol/vol) to afford a final concentration of 10 μM; 100
μL was drawn and dried under a gentle nitrogen stream at 40
°C for 30 min. After that evaporation was completed, the dry
residue was reconstituted with 100 μL of DMSO. Additional three
solutions of dBet1 (1) (10 μM) were prepared in
PBS, PBS/ACN (50:50, vol/vol), and DMSO. The four solutions were stored
into the autosampler of the LC apparatus at 37 °C for 12 h. Injections
(2 μL) were programmed every 3 h. At each time point (0, 3,
6, and 12 h), the samples underwent analyses with a Thermo Q-exactive
mass spectrometer (Thermo Fisher Scientific, Waltham, MA) as described
in the dedicated section.
UHPLC-MS Analysis
A Thermo Q-exactive
mass spectrometer
(Thermo Fisher Scientific, Waltham, MA) was used. The LC system, governed
by Chromeleon X-press software, consists of a Binary pump, a thermostated
autosampler, and a column compartment, all Dionex Ultimate 3000 series
modules (Thermo Fisher Scientific, Waltham, MA). A volume of 2 μL
was injected for each sample. Chromatographic separation of analytes
was conducted in reverse-phase chromatography. In brief, a Luna Omega
1.6 μm Polar (C18, 2.1 mm × 150 mm) was used, and the mobile
phases consisted of water (A) and acetonitrile (B), both containing
formic acid at 0.1%. The LC flow was set at 0.400 mL/min in a 12 min
gradient elution as follows: 99.5:0.5 (A/B) to 5:95 (A/B) over 10
min, 5:95 (A/B) for 2 min, and then reversion back to 99.5:0.5 (A/B)
over 2.5 min. The column was operating at a constant temperature of
40 °C. The LC effluents were introduced into the Q-Exactive mass
spectrometer by an H-ESI source that operated in the positive mode
with a sheath gas flow rate of 45; an auxiliary gas flow rate of 15;
a spray voltage of 3.5 kV; capillary temperature and auxiliary gas
heater temperature, respectively, of 320 and 350 °C; and S-lens
RF level 50. The Q-Exactive mass spectrometer operates in the data-dependent
scan (DDS) mode, with a resolution of 70.000 in full mass and 17.500
in MS/MS, in the scan mass range of 100–1500 at collision energies
of 15, 60, and 120 V. The MS/MS data were processed using “MetaSite
5.1.8 Mass 3.3.5” and “WebMetabase release-4.0.4”
(MolecularDiscovery, Ltd.).
Authors: Keith A Menear; Claire Adcock; Robert Boulter; Xiao-ling Cockcroft; Louise Copsey; Aaron Cranston; Krystyna J Dillon; Jan Drzewiecki; Sheila Garman; Sylvie Gomez; Hashim Javaid; Frank Kerrigan; Charlotte Knights; Alan Lau; Vincent M Loh; Ian T W Matthews; Stephen Moore; Mark J O'Connor; Graeme C M Smith; Niall M B Martin Journal: J Med Chem Date: 2008-09-19 Impact factor: 7.446
Authors: Anthony A Paiva; Cheryl Klakouski; Shu Li; Benjamin M Johnson; Yue-Zhong Shu; Jonathan Josephs; Tatyana Zvyaga; Ismael Zamora; Wilson Z Shou Journal: Bioanalysis Date: 2017-03-24 Impact factor: 2.681
Authors: Ch Dittrich; G Greim; M Borner; K Weigang-Köhler; H Huisman; A Amelsberg; A Ehret; J Wanders; A Hanauske; P Fumoleau Journal: Eur J Cancer Date: 2002-05 Impact factor: 9.162
Authors: Georg E Winter; Andreas Mayer; Dennis L Buckley; Michael A Erb; Justine E Roderick; Sarah Vittori; Jaime M Reyes; Julia di Iulio; Amanda Souza; Christopher J Ott; Justin M Roberts; Rhamy Zeid; Thomas G Scott; Joshiawa Paulk; Kate Lachance; Calla M Olson; Shiva Dastjerdi; Sophie Bauer; Charles Y Lin; Nathanael S Gray; Michelle A Kelliher; L Stirling Churchman; James E Bradner Journal: Mol Cell Date: 2017-06-29 Impact factor: 17.970
Authors: Mingxing Teng; Jie Jiang; Zhixiang He; Nicholas P Kwiatkowski; Katherine A Donovan; Caitlin E Mills; Chiara Victor; John M Hatcher; Eric S Fischer; Peter K Sorger; Tinghu Zhang; Nathanael S Gray Journal: Angew Chem Int Ed Engl Date: 2020-06-03 Impact factor: 15.336
Authors: S Cyrus Khojasteh; Upendra A Argikar; James P Driscoll; Carley J S Heck; Lloyd King; Klarissa D Jackson; Wenying Jian; Amit S Kalgutkar; Grover P Miller; Valerie Kramlinger; Ivonne M C M Rietjens; Aaron M Teitelbaum; Kai Wang; Cong Wei Journal: Drug Metab Rev Date: 2021-06-24 Impact factor: 6.984
Scheme 1
Scheme 2
Scheme 3
Scheme 4
Scheme 5
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7