Discovery of PROTAC BCL-XL degraders as potent anticancer agents with low on-target platelet toxicity

Article abstract of DOI:10.1016/j.ejmech.2020.112186

Anti-apoptotic protein BCL-X_L plays a key role in tumorigenesis and cancer chemotherapy resistance, rendering it an attractive target for cancer treatment. However, BCL-X_L inhibitors such as ABT-263 cannot be safely used in the clinic because platelets solely depend on BCL-X_L to maintain their viability. To reduce the on-target platelet toxicity associated with the inhibition of BCL-X_L, we designed and synthesized PROTAC BCL-X_L degraders that recruit CRBN or VHL E3 ligase because both of these enzymes are poorly expressed in human platelets compared to various cancer cell lines. We confirmed that platelet-toxic BCL-X_L/2 dual inhibitor ABT-263 can be converted into platelet-sparing CRBN/VHL-based BCL-X_L specific degraders. A number of BCL-X_L degraders are more potent in killing cancer cells than their parent compound ABT-263. Specifically, XZ739, a CRBN-dependent BCL-X_L degrader, is 20-fold more potent than ABT-263 against MOLT-4 T-ALL cells and has >100-fold selectivity for MOLT-4 cells over human platelets. Our findings further demonstrated the utility of PROTAC technology to achieve tissue selectivity through recruiting differentially expressed E3 ligases.

Full text of DOI:10.1016/j.ejmech.2020.112186

Journal Pre-proof
Discovery of PROTAC BCL-X degraders as potent anticancer agents with low on-
L
target platelet toxicity
Xuan Zhang, Dinesh Thummuri, Xingui Liu, Wanyi Hu, Peiyi Zhang, Sajid Khan, Yaxia
Yuan, Daohong Zhou, Guangrong Zheng
PII:
S0223-5234(20)30153-7
DOI:
https://doi.org/10.1016/j.ejmech.2020.112186
EJMECH 112186
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 13 January 2020
Revised Date: 21 February 2020
Accepted Date: 23 February 2020
Please cite this article as: X. Zhang, D. Thummuri, X. Liu, W. Hu, P. Zhang, S. Khan, Y. Yuan, D.
Zhou, G. Zheng, Discovery of PROTAC BCL-X degraders as potent anticancer agents with low on-
L
target platelet toxicity, European Journal of Medicinal Chemistry (2020), doi: https://doi.org/10.1016/
j.ejmech.2020.112186.
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.
©
2020 Published by Elsevier Masson SAS.

Graphical Abstract
1

Discovery of PROTAC BCL-X Degraders as Potent Anticancer Agents with
L
Low On-target Platelet Toxicity
a, 1
b, 1
b
a
a
b
Xuan Zhang, Dinesh Thummuri, Xingui Liu, Wanyi Hu, Peiyi Zhang, Sajid Khan,
b
b,
a,
Yaxia Yuan, Daohong Zhou, * Guangrong Zheng *
a
b
Department of Medicinal Chemistry and Department of Pharmacodynamics, College of
Pharmacy, University of Florida, 1333 Center Drive, Gainesville, FL, 32610, United States
*
Corresponding Authors.
E-mail addresses: zhengg@cop.ufl.edu (G.Z.); zhoudaohong@cop.ufl.edu (D.Z.).
1
Xuan Zhang and Dinesh Thummuri contributed equally.
Abstract: Anti-apoptotic protein BCL-XL plays a key role in tumorigenesis and cancer
chemotherapy resistance, rendering it an attractive target for cancer treatment. However, BCL-
X
L
inhibitors such as ABT-263 cannot be safely used in the clinic because platelets solely
depend on BCL-X to maintain their viability. To reduce the on-target platelet toxicity associated
L
L L
with the inhibition of BCL-X , we designed and synthesized PROTAC BCL-X degraders that
recruit CRBN or VHL E3 ligase because both of these enzymes are poorly expressed in human
platelets compared to various cancer cell lines. We confirmed that platelet-toxic BCL-X /2 dual
L
inhibitor ABT-263 can be converted into platelet-sparing CRBN/VHL-based BCL-X
L
specific
degraders. A number of BCL-X degraders are more potent in killing cancer cells than their
L
L
parent compound ABT-263. Specifically, XZ739, a CRBN-dependent BCL-X degrader, is 20-
fold more potent than ABT-263 against MOLT-4 T-ALL cells and has >100-fold selectivity for
MOLT-4 cells over human platelets. Our findings further demonstrated the utility of PROTAC
technology to achieve tissue selectivity through recruiting differentially expressed E3 ligases.
L
Keywords: BCL-X ; PROTAC; apoptosis; platelet; degradation
1

1
. Introduction
Apoptosis is a conserved and highly regulated biological process that plays a critical role in
maintaining cellular homeostasis [1]. Resistance to apoptosis is a common feature in human
malignancies and considered a key hallmark of cancer [2]. Therefore, targeting specific anti-
apoptotic pathways is a promising cancer therapeutic strategy. Members of B-cell lymphoma 2
(BCL-2) protein family, consisting of both pro- and anti-apoptotic proteins, regulates the
intrinsic apoptotic pathway. Several anti-apoptotic BCL-2 family proteins such as BCL-2, BCL-
X and MCL-1 have been validated as anticancer targets [3-5]. Inhibition of these proteins with
L,
small-molecule inhibitors promotes Bax/Bak oligomerization and ultimately induces
mitochondrial outer membrane permeabilization, followed by cytochrome c release and
activation of caspases to execute apoptosis [1].
Figure 1. Chemical structures of representative BCl-2/BCL-X inhibitors and BCL-2 family
L
protein degraders.
2

ABT-737 (Figure 1), the first potent BCL-2/BCL-X
fragment-based drug discovery using NMR [6,7]. Optimization of ABT-737 afforded navitoclax
ABT-263) (Figure 1) [8,9], an orally bioavailable BCL-2/BCL-X dual inhibitor that has been in
L
dual inhibitor, was developed via
(
L
Phase II clinical trials for hematological malignancies and small cell lung cancer (SCLC).
However, ABT-263 treatment leads to rapid and dose-dependent thrombocytopenia when dosed
as a single agent [10], consistent with the studies showing the dependency of platelets on BCL-
L
X to maintain their viability [11,12]. To overcome this on-target, dose-limiting toxicity, a
selective BCL-2 inhibitor, venetoclax (ABT-199) (Figure 1), was developed. Venetoclax has
shown antileukemic activity without the induction of thrombocytopenia [13]. Subsequently, it
was approved by the FDA for the treatment of chronic lymphocytic leukemia (CLL) and small
lymphocytic lymphoma (SLL) as a single agent, and for acute myeloid leukemia (AML) in
combination with low-intensity chemotherapy [14].
The overall response rate of CLL patients to venetoclax is 71-79% but the complete remission
rate (20%) is relatively low [15]. Upregulation of BCL-X by microenvironmental survival
L
signals has been identified as the major component accountable for the resistance, consistent
with the high efficacy of BCL-2/BCL-X
CLL cells [16]. In addition, venetoclax has limited utility for the treatment of solid tumors
17,18] because BCL-2 is mainly associated with the survival of hematological malignancies and
L
dual inhibitor ABT-263 in killing venetoclax resistant
[
BCL-X is the most common BCL-2 family member overexpressed in solid tumors, as well as in
L
a subset of leukemia and lymphoma cells [19]. Bioinformatics analyses also reveal a strong
correlation between the levels of BCL-X expression and resistance to chemotherapies [20].
L
L
Further, it has been well established that inhibition of BCL-X by ABT-263 is primarily
3

responsible for the observed synergy with chemotherapies in solid tumors [21,22]. Taken
together, BCL-X is one of the most important validated cancer targets.
L
More recently, we and others discovered that ABT-263 and other BCL-X inhibitors, such as
L
A-1331852 and A-1155463 are potent senolytics, referring to small-molecules that can
selectively kill senescent cells [23-26]. This is because BCL-X is a key anti-apoptotic protein in
L
many types of senescent cells. Subsequent studies on ABT-263 in mouse models have
demonstrated that clearance of chemotherapy-induced senescent cells reduces several short- and
long-term adverse effects of the chemotherapy, as well as cancer relapse and metastasis [27].
L
These findings suggest an added benefit of targeting BCL-X for cancer treatment. Therefore, it
is highly desirable to develop a strategy that can retain the versatility and efficacy of BCL-XL
inhibitors, while reducing their on-target platelet toxicity.
Several strategies have been devised to minimize the on-target platelet toxicity associated with
the inhibition of BCL-X . One strategy is to combine ABT-263 with chemotherapy or targeted
L
therapies to reduce the dose regimen of ABT-263 so that thrombocytopenia can be manageable
L
[22]. In another strategy, prodrugs of BCL-2/BCL-X dual inhibitors have been prepared to limit
drug exposure to platelets [28], and the lead candidate, APG-1252, is currently in Phase I clinical
trial [29]. Further, antibody-drug conjugates (ADCs) of BCL-X inhibitors have been designed to
L
deliver the inhibitors to cancer cells more specifically. For example, ABBV-155, a BCL-XL
inhibitor ADC that targets B7H3 expressing cancer cells, is currently in Phase I clinical trial
[
30]. We have recently demonstrated that the proteolysis targeting chimera (PROTAC), an
emerging therapeutic modality [31-33], is also a feasible solution to reduce platelet toxicity
associated with BCL-X inhibition [34]. This was based on the hypothesis that because
PROTACs engage E3 ligases to induce protein degradation, they can achieve cell/tissue
L
4

selectivity if the E3 ligase they recruit is differentially expressed in cells or tissues [34]. Von
Hippel–Lindau (VHL) cullin-2 and cereblon (CRBN) cullin-4A RING E3 ligases, of which
small-molecule ligands have been successfully employed in PROTAC design, were found to be
minimally expressed in human platelets [35,36]. In a proof of concept study, we have shown that
DT2216 (Figure 1), a VHL-based PROTAC derived from ABT-263, is more potent to a variety
of BCL-X dependent cancer cells with significantly less platelet toxicity than its parent
L
compound ABT-263 [34]. In another study, through XZ424 (Figure 1), which is constructed by
L
tethering BCL-X specific inhibitor A-1155463 to CRBN ligand pomalidomide, we have
demonstrated that CRBN could be an alternative E3 ligase to be recruited to build BCL-X
degraders with low platelet toxicity [37].
L
The discovery of PROTAC MCL-1 and BCL-2 degraders (Figure 1) have been reported
recently [38,39]. Herein, we describe our efforts in the discovery of potent PROTAC BCL-XL
degraders with low platelet toxicity. Emerging evidence suggests that it could be beneficial to
parallelly develop two or more PROTAC series hijacking different E3 ligases [40-42]. Thus, we
L
have evaluated both CRBN and VHL E3 ligases based BCL-X degraders. The critical role of
linkerology in degradation potency and efficacy of PROTACs is well estabolished [43-45].
Therefore, we have also systematically varied the length and composition of the linker unit.
5

2
. Results and discussion
Figure 2. X-ray crystal structures of ABT-263 (black) bound to (A) BCL-X , available from
L
PDB, code: 4QNQ, and (B) BCL-2, available from PDB, code: 4LVT. The local view of the key
hydrogen bond between morpholine ring and protein is shown in the additional box.
Cytotoxicity of the designed PROTACs and their ability in inducing BCL-X degradation were
L
primarily evaluated in MOLT-4, a T-cell acute lymphoblastic leukemia (T-ALL) cell line
dependent on BCL-X for survival. We focused on ABT-263 as the ‘warhead’ in our PROTAC
L
design. The co-crystal structures of ABT-263 in complex with BCL-2/BCL-X (PDB code 4LVT
L
and 4QNQ, Figure 2) [13] indicate that the morpholine ring in ABT-263 is solvent-exposed,
making it a suitable position for tethering to an E3 ligase binding unit through a linker.
Accordingly, we initially converted ABT-263 to a ‘PROTAC-ready’ derivative by replacing the
morpholine ring with a piperazine ring, followed by tethering it to a previously reported VHL
ligand [46] through various linkers to yield compounds 1a-f, 2a-d, 3a-c, and 4a-b (Scheme 1,
Table 1).
6

Table 1. BCL-X
L
Degraders Designed with Various Linkers Tethering through a Piperazine
Ring to a VHL Ligand
a
a
Compd
Linker
IC50 (nM)
Compd
2b
Linker
IC50 (nM)
ABT-263
-
230
82.0
1
1
1
a
b
c
>2000
>2000
2c
2d
3a
94.5
181
324
>2000
467
1
1
d
e
3b
3c
4a
4b
512
232
151
914
7
7.1
(
DT2216)
1
f
108
130
2
a
a
IC50 values are the means of at least three independent experiments; reduction of MOLT-4 cell viability after 48 h treatment.
We first evaluated the effects of these compounds on the viability of MOLT-4 T-ALL cells,
with ABT-263 as the positive control (Table 1), followed by examining their ability to BCL-
2
/BCL-X degradation in the same cell line using Western blotting (Figure 3). In the series with
L
7

linkers containing an amide linkage and an alkane chain, we synthesized compounds 1a-f with 1-
methylene groups in the alkane chain (Table 1) and determined the optimal linker length by
6
comparing their potencies in the cell viability assay. DT2216 (1e) with five methylene groups in
the linker was the most potent in this series and 3 times more potent than ABT-263. Shortening
the linker in 1e afforded 1a-d, which exhibited largely decreased cellular potencies in the same
assay. However, the analog (1f) with one methylene unit longer than that in 1e showed slightly
reduced cytotoxicity against MOLT-4 cells. These results indicate that the optimal linker length
is six carbon atoms. The cell viability assay data of 1a-f were well correlated with the protein
degradation data (Figure 3), indicating that the cytotoxicity is mainly derived from BCL-X
L
protein degradation. Replacing the amide linkage in 1a-f with a C-N linkage yielded analogs 2a-
d. The optimal linker length in this series is five carbon atoms (compound 2b) and appears to be
one carbon atom shorter than the amide linkage containing series, which could be attributed to
the increased flexibility of the C-N bond in comparison to the amide bond. Replacement of one
methylene group in the linker of 1e with an oxygen atom afforded 3a, which was ~4 times less
potent than 1e. In general, introduction of oxygen atoms into the linker unit is detrimental to
cellular potency. None of the oxygen-containing PROTACs we synthesized showed better
cytotoxicity to MOLT-4 cells in comparison to their corresponding alkane linker analogs (e.g. 3a
vs 1e, 3b vs 1f, and 4a vs 2b). Interestingly, although 4b was slightly more potent in inducing
BCL-X degradation than 4a at 100 nM (Figure 3B), it was ~6 times less potent in reducing
L
MOLT-4 cell viability. The cytotoxicity of the PROTACs is derived from both direct inhibition
of BCL-X and degradation of this protein. The relative high cellular potency of 4a could be due
L
L
to its higher cell permeability than 4b despite its low efficiency in inducing BCL-X protein
degradation.
8

Figure 3. Western blot analysis of the effect of VHL-based BCL-X degraders on BCL-X and
L
L
BCL-2 protein levels in MOLT-4 T-ALL cells. Cells were treated with 100 nM of each
compound for 16 h before harvesting. Degradation activity is reported as % of total protein
remaining after compound treatment relative to the vehicle after normalization with β-actin as
quantified by Image J software.
Table 2. BCL-XL Degraders Designed with Various Linkers Tethering through an N-
Methylamino Group to a VHL Ligand
O
O O
OH
S
S
N
H
O
N
N
N
N
H
N
N
O S
O
H
S
CF
3
O
N
O
H
Linker
N
Cl
a
a
Compd
5
Linker
IC50 (nM)
Compd
6e
Linker
IC50 (nM)
>2000
139
9

6
6
6
a
b
c
1243
460
6f
1058
97.0
200
7a
7b
7c
81.3
147
6
d
206
a
IC50 values are the means of at least three independent experiments; reduction of MOLT-4 cell viability after 48 h treatment.
The salt bridge between Glu96 of BCL-XL and the protonated nitrogen atom on the
morpholine ring of ABT-263 is one of the crucial interactions for high BCL-X binding (Figure
L
2
). However, the ring system is not required to maintain high binding affinity as the morpholine
ring in ABT-263 can be replaced by a dimethylamino group as shown in ABT-737. Thus, we
replaced the morpholine ring in ABT-263 with an N-methylamino group and tethered it to the
same VHL binding moiety as compounds in Table 1 through an alkane chain to generate
compounds 6a-f (Table 2). Interestingly, despite the smaller linkage unit (a single N atom vs. a
piperazine ring), the optimal linker length for this series of compounds was also six carbon atoms;
compound 6c was the most potent degrader among its analogs (Figure 3B) and exhibited a
comparable cellular potency to 1e in reducing MOLT-4 cell viability. Compound 5, which was
generated by replacing the C-N linkage in 6c with an amide bond, had no effect on MOLT-4 cell
viability (IC > 2 µM) and weak BCL-X degradation in MOLT-4 cells at 100 nM, confirming
5
0
L
L
the importance of the salt bridge in maintaining high BCL-X protein binding and the subsequent
protein degradation. Incorporation of oxygen atoms into the linker unit of 6d caused small
increases in both protein degradation and cellular potency (7a vs. 6d). In addition, PEG-linker
10

L
containing PROTACs, i.e. 7a-c, are linker length independent in inducing BCL-X protein
degradation and cell death.
Similar to DT2216 (1e) [34], none of these VHL-based PROTACs were able to degrade BCL-
2
in MOLT-4 cells (Figure 3), further demonstrated the feasibility of achieving target specificity
through conversion of non-selective inhibitors to PROTACs. Our studies with the most potent
analog in this series, DT2216, have shown that BCL-X , but not BCL-2, can form stable ternary
L
complexes with DT2216 and VHL in live cells as determined by nanoBRET assay [47], which
L
could contribute to the specific BCL-X degradation [34].
Table 3. BCL-X Degraders Designed with Various Linkers Tethering through a Piperazine
L
Ring to Pomalidomide
a
a
Compd
Linker
IC50 (nM)
Compd
10a
Linker
IC50 (nM)
8
a
82.0
161
8
b
346
518
10b
10c
211
126
8
c
8
d
>2000
>2000
11a
11b
118
722
8
e
11

9
9
a
28.3
43.0
100
12a
12b
12c
60.5
41.4
17.3
b
9
c
a
IC50 values are the means of at least three independent experiments; reduction of MOLT-4 cell viability after 48 h treatment.
We next replaced the VHL binding moiety in compounds 1-4 with CRBN binding moiety
pomalidomide. It is well established that different E3 ligases require different linker length
and/or composition for optimal cellular potency. Thus, we also conducted a systematic
investigation of linker length and composition in the CRBN-based PROTACs. As shown in
Table 3, compounds with an amide linkage were generally less potent in reducing MOLT-4 cell
viability when compared with their corresponding C-N linkage containing analogs with a similar
linker length. Among the alkane-linker containing PROTACs (8a-e and 9a-c), compounds with
the shortest linkers, i.e. 8a and 9a, were the most potent. Compound 8a, which contains an amide
linkage and three methylene groups, was 3 times more potent than the parent compound ABT-
2
63. Switching the amide linkage to a C-N linkage resulted in a significant improvement on
cellular potency; compound 9a, which contains three methylene groups in the linker, was 8 times
more potent than ABT-263. Compounds with an amide linkage and a PEG-chain, 10a-c,
appeared to have little dependence on linker length for their cellular potency. Replacing the
amide linkage in 10a-c with a urea linkage, exemplified by 11a, had little effect on the ability to
inhibit MOLT-4 growth. However, the thiourea analog 11b exhibited a ~6-fold decrease in
potency compared to 11a, in the same assay. In contrast to 9a-c that contain a C-N linkage and
an alkane-chain in their linkers, PEG-linker containing analogs 12a-c preferred longer linker
length for optimal cellular potency. Compound 12c, which has the longest linker in this series
12

with 11 atoms, had the highest cellular potency that was ~12 times more potent than ABT-263 in
reducing MOLT-4 cell viability. These results further demonstrate the empirical nature of
PROTAC design, which is at least partially attributable to the fact that potencies of reducing cell
viability for inhibitor-derived PROTACs are a combination of protein degradation and direct
protein inhibition, both of which are dependent on cell permeability. As a result, analogs with
similar intrinsic protein degradation efficiency may exhibit completely different ability to induce
protein degradation in cells.
L
We evaluated all the compounds in Table 3 for their ability to induce BCL-X /BCL-2 protein
degradation in MOLT-4 cells at 50 nM for 16 h. As shown in Figure 4, a good correlation
between the ability to reduce MOLT-4 cell viability and inducing BCL-X degradation was
L
observed within each analog series. However, there was no obvious correlation among
compounds from different series. For example, compounds 9a-c were more potent in the cell
viability assay but much less effective in inducing protein degradation than 10a-c, indicating the
cellular activity of 9a-c was more dependent on direct protein inhibitory effects and permeability.
Similar to VHL-based PROTACs, no BCL-2 degradation was observed with these CRBN-based
PROTACs.
13

L
Figure 4. Western-blot analysis of BCL-X and BCL-2 protein levels in MOLT-4 T-ALL cells
after treating with CRBN-based BCL-X degraders at 50 nM for 16 h. Degradation activity
L
reported as % of total protein remaining after compound treatment relative to vehicle after
normalization with β-actin as quantified by Image J software.
Table 4. BCL-X
L
Degraders Designed with Various Linkers Tethering through an N-
Methylamino Group to Pomalidomide
O
O O
S
S
H
N
O
N
N
O
H
O
N
N
N
NH
H
N
O S
O
CF
3
Linker
O
Cl
a
a
Compd
Linker
IC50 (nM)
Compd
Linker
IC50 (nM)
14

13a
13b
14a
14b
15a
15b
16a
770
769
177
179
29.2
35.0
32.5
16b
16c
16d
16e
17a
38.7
60.9
147
>2000
35.0
10.1
19.3
1
7b
(
XZ739)
O
O
17c
O
O
a
IC50 values are the means of at least three independent experiments; reduction of MOLT-4 cell viability after 48 h treatment.
Subsequently, the N-methylamine precursor for analogs in Table 2 was employed for CRBN-
based PROTACs and the cell viability data are shown in Table 4. Compounds 13a and 13b with
an amide linkage to connect the 1,2,3-triazole containing linker and the precursor displayed
moderate cytotoxicity against MOLT-4 cells. The extension of the alkane-chain portion of the
linker resulted in an improvement in cellular activity (13a vs. 14a, 13b vs. 14b). The PEG-chain
portion was more tolerated for modification, supported by the similar activity for 13a and 13b
(or 14a and 14b). Switching the amide linkage in compounds 14a-b to C-N linkage, which
resulted in 15a-b, significantly increased cytotoxicity in MOLT-4 cells. Protein degradation data
indicated that 15a and 15b are better BCL-X degraders than their amide analogs (Figure 4).
L
15

Similar to the compounds in Table 2, we also synthesized both alkane-chain and PEG-chain
containing PROTACs with a C-N linkage. Compared to the corresponding VHL-based
PROTACs with the same linker (Table 2, Figure 3), the CRBN-based PROTACs displayed much
higher cytotoxicity and BCL-X degradation activity. Compound 17b (XZ739), which contains a
L
PEG linker with linker length of 11 atoms, was the most potent BCL-X degrader against
L
MOLT-4 cells and was ~22 times more potent than ABT-263.
Table 5. Cellular Activity of Selected Analogs in Human Platelets and a Panel of Human Cancer
Cell Lines
a
IC50 (nM) 48 h treatment
b
c
Compd
DC50 (nM)
IC50 ratio
MOLT-4
227
RS4;11
49.0
213
H146
43.8
278
Platelets
242
ABT-263
DT2216
ND
53
1.1
77.1
>10,000
>10,000
>10,000
1,560
>130
>122
>123
90
2
6
1
1
1
1
b
82.0
76
203
93
c
81.3
189
265
71
2c
5a
6a
17.3
38.5
62.2
129
24.6
61.7
70.2
25.3
4.5
6.3
10.6
2.5
29.2
6,250
214
101
120
32.5
3,296
7b (XZ739)
10.1
41.8
1,217
a
b
c
IC50 values are the means of at least three independent experiments; in MOLT-4 cells, 16 h treatment; IC50 ratio between
human platelets and MOLT-4 cells.
L
We selected several potent BCL-X degraders for further investigation. The DC50 values (the
concentrations at which 50% protein has been degraded) were determined in MOLT-4 cells after
treatment with each individual degrader for 16 h. As shown in Table 5 and Figure S2, the
CRBN-based PROTACs 12c, 15a, 16a, and 17b (XZ739) were more potent in inducing BCL-XL
degradation and reducing the viability of MOLT-4 cells than the VHL-based PROTACs DT2216,
16

2
b, and 6c. Among these PROTACs, XZ739 was the most potent BCL-X degrader with a DC
L 50
value of 2.5 nM. Likely due to the reduced membrane permeability relative to their parent
compound ABT-263, coupled with the lack of BCL-2 degradation, none of these ABT-263
derived BCL-X degraders were significantly more potent than ABT-263 in reducing the
L
viability of RS4;11, a B-cell ALL cell line that mainly depends on BCL-2 for survival [13]. The
CRBN-based BCL-X
cells, a small cell lung cancer cell line that depends on both BCL-2 and BCL-X
has low VHL expression (Figure S1). Subsequently, the BCL-X degraders in Table 5 were
L
degraders were also more potent than the VHL-based degraders in H146
L
for survival and
L
tested for their toxicity against human platelets and most displayed >100-fold selectivity for
MOLT-4 cells over platelets. In contrast, ABT-263 showed no selectivity between these platelets
and MOLT-4 cells.
L L
Figure 5. XZ739-induced BCL-X degradation. (A) Western blots showing the BCL-X protein
levels in MOLT-4 cells treated with the indicated concentrations of XZ739 for 16 h. (B) BCL-XL
protein levels in MOLT-4 cells treated with 100 nM of XZ739 at the indicated time points. (C)
BCL-X protein levels in MOLT-4 cells after incubation with 100 nM of XZ739 for 16 h
L
followed by drug washout, resuspension and incubation of the cells for additional time as
indicated in a drug-free medium. (D-E) Pretreatment with 10 µM pomalidomide (POM) or 1 µM
L
MG-132 for 2 h blocked the degradation of BCL-X induced by XZ739. (F) Western blot
17

analysis of BCL-X in MOLT-4 cells treated with XZ739-NC at indicated concentrations for 16
L
h. Data are representative of two independent experiments.
Western blot analysis revealed that XZ739 dose-dependently induced BCL-X
L
degradation in
MOLT-4 cells (Figure 5A). In addition, the BCL-X degradation induced by XZ739 in MOLT-4
L
was rapid, starting within 2 h; and 8 h after drug treatment, more than 96% of the protein was
degraded with 100 nM of XZ739 (Figure 5B). The effects of XZ739 on BCL-X protein levels in
L
MOLT-4 were long-lasting and also reversible, as indicated in the ‘washout’ experiment (Figure
5
C). Further, XZ739-induced BCL-X degradation can be abrogated by proteasome inhibitor
L
MG-132 (Figure 5D) or excess competitive CRBN ligand pomalidomide (Figure 5E), indicating
that the degradation depends on both proteasomes and the CRBN E3 ligase. To further confirm
L
that the CRBN E3 ligase is involved in XZ739-induced BCL-X degradation. We synthesized
XZ739-NC, a negative control compound of XZ739, in which a methyl group is installed on the
amino group in the pomalidomide moiety of XZ739 to block CRBN binding. Not surprisingly,
XZ739-NC did not induce BCL-X degradation in MOLT-4 cells (Figure 5F).
L
Figure 6. XZ739 shows improved selectivity because of its low activity in inducing BCL-XL
degradation in human platelets, and the remaining cytotoxicity to human platelets mainly
L L
depends on BCL-X inhibition. (A) Western blot analysis of BCL-X levels after treatment of
18

human platelets with indicated concentration of XZ739 for 16 h. (B) IC50 fold changes in
MOLT-4 cells and human platelets after blocking with 10 µM pomalidomide (POM). IC values
5
0
in MOLT-4 cells and platelets were normalized to 1.0.
In human platelets, no significant changes in BCL-X
L
protein levels were observed after 16 h
treatment with up to 1.0 µM of XZ739 (Figure 6A). The cytotoxicity of high concentrations of
XZ739 to platelets most likely derived from BCL-X inhibition rather than degradation as pre-
L
incubation of platelets with pomalidomide did not affect the cytotoxicity of XZ739 to platelets
while a ~27-fold shift in IC50 values was observed in MOLT-4 cells with pomalidomide
pretreatment (Figure 6B).
Figure 7. Characterization of XZ739-mediated apoptosis in MOLT-4 cells. (A) Western blot
analysis of cleaved-PARP and cleaved-caspase-3 after 16 h treatment with indicated
concentrations of XZ739. (B) Flow cytometry analysis of apoptosis using Annexin-V and PI
staining. MOLT-4 cells were treated with XZ739 or ABT-263 at 10 and 100 nM for 48 h. XZ739
(100 nM) significantly increased the percentage of apoptotic cells, and QVD (10 µM) pre-
19

treatment for 2 h inhibited the apoptosis induced by XZ739. Data are representative of two
independent experiments.
We next assessed the impact of XZ739 on apoptosis. Western blot analysis showed that XZ739
dose-dependently increased poly (ADP-ribose) polymerase (PARP) and caspase-3 cleavage in
MOLT-4 cells (Figure 7A), suggesting the apoptotic cell-death mechanism. Further, to confirm
that XZ739 induces cell death through caspase-mediated apoptosis, we did flow cytometry
analysis of apoptosis using Annexin-V and propidium iodide (PI) staining (Figure 7B). We found
that, 10 nM of XZ739 treatment for 48 h significantly increased the percentage of Annexin-V-
positive cells in MOLT-4 cells compared to the vehicle group. In comparison, ABT-263 showed
similar effect at much higher concentration (100 nM). In addition, pretreatment with 10 µM of
pan-caspase inhibitor Q-VD-OPh (QVD) for 2 h partially inhibited the XZ739-induced
apoptosis, which confirms that XZ739 induces cell death through a caspase-dependent
mechanism.
CRBN-targeting immunomodulatory drugs (IMiDs) such as thalidomide, pomalidomide and
lenalidomide can induce degradation of a number of neo-substrate proteins through a C2H2
degron. PROTACs that are based on these IMiDs may have the potential to exert some off-target
effects by degrading some of these C2H2 zinc finger proteins [48]. To test this possibility, we
evaluated the effects of XZ739 on the expression of the Ikaros proteins that are well known
targets of those CRBN E3 ligands, and found that IKZF1 and IKZF3 protein levels were not
affected in MOLT-4 cells treated with 10 nM XZ739 but downregulated when the cells were
treated with 100 nM XZ739 (Figure S6).
20

3
. Chemistry
2 4 3
Scheme 1. Reagents and conditions: (a) i. MeNH , MgSO , THF, rt; ii. NaBH CN, THF, rt; (b)
Troc-Cl, TEA, DCM, rt; (c) TFA, DCM, rt; (d) 24, TEA, acetonitrile, reflux; (e) 26, EDCI,
DMAP, DCM, rt; (f) Zn, HOAc, THF, rt.
Precursor 18 that contains a piperazinyl moiety for linker attachment and aldehyde 20 were
prepared as described in our recent publication [34]. The synthesis of methylamino group
containing precursor 19 is outlined in Scheme 1. Briefly, aldehyde 20 was converted to amine 21
through a reductive amination reaction. The amino group in 21 was protected with 2,2,2-
trichloroethoxycarbonyl (Troc) group by treating with trichloroethyl chloroformate (Troc-Cl) to
give 22. Removal of the Boc protection group in 22 afforded 23, which was subjected to S Ar
N
substitution with 24 to yield N-acylsulfonamide 25. Coupling of 25 with acid 26 in the presence
of EDCI and DMAP afforded 27. Removal of the Troc group in 27 with Zn/HOAc gave 19.
21

OH
OH
O
O
a, b
HO
N
HO
(
)n
N
N
H2N
+
(
)n OR
H
O
O
N
S
O
R = Me or Et
O
N
S
O
H
O
H
2
2
2
9a, n = 0 29d, n = 3
9b, n = 1 29e, n = 4
9c, n = 2 29f, n = 5
30a, n = 0 30d, n = 3
30b, n = 1 30e, n = 4
30c, n = 2 30f, n = 5
N
N
28
O
O O
S
S
N
H
OH
a
O
N
N
N
H
18
N
O
O
S
N
N
CF3
(
)n
N
H
O
O
N
S
O
H
1
1
1
a, n = 0
b, n = 1
c, n = 2
1d, n = 3
1e, n = 4
1f, n = 5
N
Cl
HO
O
OH
HO
O
OH
30g R =
30h R =
O
O
O
O
O
29g
O
O
N
R
N
H
a
HO
HO
O
HO
O
O
O
O
OH
O
N
H
S
28
O
O
O
O
29h
29i
N
HO
OH
30i R =
O
O
O
O
O
O
O O
S
S
N
OH
H
a
N
N
N
H
N
H
N
18
N
O
O
S
N
CF3
L
O
N
S
O
O
O
H
N
3a
3b
3c
Cl
L =
O
O
O
O
O
O
O O
OH
S
S
N
O
O
H
a
N
N
N
N
N
19
H
H
N
O S
O
O
N
S
30e
CF3
O
H
N
5
Cl
Scheme 2. Reagents and conditions: (a) HATU, TEA, DCM, rt; (b) LiOH monohydrate,
MeOH, H O, rt.
2
The syntheses of VHL-based PROTACs are illustrated in Schemes 2 and 3. The VHL ligand
motif 28 was prepared by following reported methods [46]. Coupling of 28 with acid 29a-f
followed by hydrolysis with LiOH afforded 30a-f. Direct coupling between dicarboxylic acid
22

2
9g-i and 28 also worked smoothly in good yields (62-89%). PROTACs 1a-f, 3a-c, and 5 were
obtained via coupling of precursor 18/19 with 30a-i (Scheme 2).
2 3
Scheme 3. Reagents and conditions: (a) i. K CO , NaI, DMSO, 80 °C; ii. LiOH monohydrate,
MeOH, THF, H O, rt; (b) 28, HATU, TEA, DCM, rt.
2
The synthesis of the series with C-N linkage started from bromides 31a-f or tosylates 34a-d
Scheme 3). Nucleophilic substitution with amine 18/19 in the presence of K CO and NaI at
(
2
3
8
0 °C, followed by ester hydrolysis, afforded acid intermediates 32a-d, 33a-f, 35a-b, and 36a-c,
respectively, which were coupled with VHL ligand 28 to generate the corresponding final
products 2a-d, 4a-b, 6a-f, and 7a-c.
23

Scheme 4. Reagents and conditions. (a) DIPEA, DMF, 90 °C; (b) 18, HATU, DIPEA, DCM, rt;
(c) TFA, DCM, rt.
The syntheses of CRBN-based PROTAC BCL-X degraders are illustrated in Schemes 4-7.
L
As shown in Scheme 4, amino acids 37a-e were coupled with 38 in the presence of DIPEA to
afford intermediates 39a-e. Compound 8a-e were synthesized by amide coupling of 39a-e with
1
8. PEG linker containing PROTACs 10a-c were synthesized by initial coupling of amine 40a-c
with 38, followed by removal of the tert-butyl on them to give intermediates 41a-c and amide
coupling with 18.
Scheme 5. Reagents and conditions: (a) HATU, DIPEA, DCM, rt; (b) CuSO ·5H O, sodium
4
2
t
L-ascorbate, BuOH, THF, H O, 50 °C; (c) K CO , NaI, DMSO, 80 °C.
2
2
3
24

As shown in Scheme 5, compounds 13a-b and 14a-b were produced through coupling of
azido acid 42a-b with precursor 19, followed by a click reaction with alkynes 44a/44b [37]. The
synthesis of 15a-b started from a nucleophilic substitution of tosylate 45 with 19, followed by
click reaction.
Scheme 6. Reagents and conditions. (a) DIPEA, DMF, 90 °C; (b) TFA, DCM, rt; (c) i.
carbonyldiimidazole, TEA, DCM, rt; ii. 18, DIPEA, rt; (d) i. 1,1'-thiocarbonyldiimidazole, TEA,
DCM, rt; ii. 18, DIPEA, rt.
The synthesis of compounds 11a and 11b is shown in Scheme 6. The key intermediate 48 was
synthesized by coupling of compounds 47 and 38, followed by removal of the Boc group by
treating with TFA/DCM. Compound 48 was treated with carbonyldiimidazole or 1,1'-
thiocarbonyldiimidazole and then with precursor 18 to afford compound 11a and 11b,
respectively.
25

O
O
O O
S
S
NH
O
N
H
O
N
N
N
N
H
H
N
N
O
O
S
N
O
1
8
CF3
(
)n
c
9a, n = 1
OHC
9b, n = 2
9c, n = 3
HO
(
)n
N
O
(
)n
N
O
3
8
H
b
H
HO
(
)n NH2
N
O
N
O
a
O
O
Cl
NH
O
NH
O
4
9a, n = 1
9b, n = 2
9c, n = 3
9d, n = 4
9e, n = 6
50a, n = 1
50b, n = 2
50c, n = 3
50d, n = 4
50e, n = 6
51a, n = 1
51b, n = 2
51c, n = 3
51d, n = 4
51e, n = 6
O
O O
4
4
4
4
S
S
N
1
9
H
c
N
N
H
CF3
N
(
)n
N
H
O
N
O
S
O
N
O
O
1
6a, n = 1
6b, n = 2
6c, n = 3
6d, n = 4
1
1
1
NH
O
Cl
16e, n = 6
HO
O
MsO
O
O
O O
(
O
)m
NH2
( O
)m
NH
O
O
S
S
1
8
N
5
5
5
5
2a, m = 0
2b, m = 1
2c, m = 2
2d, m = 3
NH
H
54a, m = 0
54b, m = 1
54c, m = 2
N
O
e
N
N
N
H
N
O
S
N
O
m
O
CF3
( O
)
NH
O
O
O
d
NH
3
8
a
1
1
1
2a, m = 0
2b, m = 1
2c, m = 2
N
O
HO
O
Cl
O
(
O
)m
NH
O
O
NH
N
O
R
O
5
5
5
5
3a, m = 0
3b, m = 1
3c, m = 2
3d, m = 3
N
O
b
O
O
O O
S
S
O
N
H
N
O
H
OHC ( O
)m
NH
O
N
N
O
O
O
1
9
N
N
(
O
)m
NH
H
N
O S
N
O
c
CF3
O
5
5
5
5a, m = 1
5b, m = 2
5c, m = 3
O
17a, m = 1, R = H
17b, m = 2, R = H
1
7c, m = 3, R = H
Cl
XZ739-NC, m = 2, R = Me
1
9
e
F
HO
O
)2
MsO
O
)2
O
O
( O
NH
O
O
( O
NH
O
5
2c
O
N
d
N
N
N
O
a
N
O
N
O
O
O
O
5
6
57
58
2
Scheme 7. Reagents and conditions. (a) DIPEA, DMF, 90 °C; (b) i. COCl , DMSO, DCM, -78
°
C; ii. 50 or 53, DCM, -78 °C; iii. TEA, DCM, -78 °C then rt; (c) NaBH CN, TEA, DCM, rt; (d)
3
MsCl, TEA, DCM, rt; (e) DIPEA, NaI, 1,4-dioxane, 90 °C.
The syntheses of CRBN-based PROTACs with a C-N linkage are illustrated in Scheme 7.
Since the pomalidomide moiety is unstable under the treatment of inorganic bases, they were
synthesized either by nucleophilic substitution of mesylates with precursor 18/19 in the presence
of DIPEA or by reductive amination of aldehydes with 18/19. Briefly, coupling of 38 with 49a-
e/52a-d afforded 50a-e and 53a-d, respectively. Swern oxidation of 50a-e produced aldehyde
26

5
1a-e. Other oxidation methods, including PCC, PDC, Dess-Martin periodinane, and IBX,
afforded unsatisfactory results. Subsequent reductive amination of the aldehydes with the
corresponding precursor 18 or 19 yielded compounds 9a-c and 16a-e, respectively. Compounds
1
7a-c were obtained by using the same strategy. Compounds 12a-c were synthesized by using a
different method. Mesylates 54a-c, which are more stable than the corresponding aldehyde
intermediates, were prepared from 53a-c. Nucleophilic substitution of 54a-c with 18 in the
presence of DIPEA and NaI afforded 12a-c. Similarly, XZ739-NC, the negative control
compound of XZ739, was synthesized starting from compound 56.
4
. Conclusions
In this study, we reported the design, synthesis, and evaluation of a series of PROTAC BCL-
X degraders. Through tethering BCL-2/BCL-X dual inhibitor ABT-263 and a VHL ligand or a
L
L
CRBN ligand, and varying the linker unit, we obtained a number of BCL-X
showed improved cytotoxicity against BCL-X dependent MOLT-4 cells when compared with
ABT-263. Protein degradation assay confirmed that these BCL-X PROTACs were able to
L
degraders that
L
L
degrade BCL-X but not BCL-2 despite that the warhead ABT-263 binds to both proteins. In
L
addition, human platelets were more tolerant to the treatment of these degraders whereas the
parent BCL-X inhibitor ABT-263 showed no selectivity for MOLT-4 cells over platelets. Both
L
CRBN and VHL-based PROTACs were able to achieve >100-fold selectivity in vitro for MOLT-
4
cells over human platelets. XZ739, a CRBN-based BCL-X PROTAC, was the most potent
L
analog against various cancer cell lines. The mouse plasma stability assay and preliminary
pharmacokinetic study revealed that XZ739 is bioavailable (Figure S4). Overall, we have
27

developed a novel strategy to reduce drug on-target toxicity by conversion of a conventional
inhibitor into a protein degrader.
Despite the high molecular weight of PROTACs that are derived from PPI inhibitors, a
number of PPI-targeting degraders have been successfully developed, including MCL-1 [38],
BCL-2/MCL-1 [39], MDM2 [49], BCL-6 [50], and STAT3 [51,52]. Due to their catalytic mode
of action, common ‘rules’ such as ‘rule-of-five’ may not be suitable to predict the drug-like
L
properties of PROTACs. The size and molecular weight of our BCL-X specific degraders are
among the largest PROTACs, which further highlighted the broad utility of this therapeutic
modality.
5
. Experimental
5
.1. Chemistry
THF, DCM, toluene, and acetonitrile were obtained via a solvent purification system by
filtering through two columns packed with activated alumina and 4 Å molecular sieve,
respectively. All other chemicals obtained from commercial sources were used without further
purification. Flash chromatography was performed using silica gel (230–400 mesh) as the
stationary phase. Reaction progress was monitored by thin-layer chromatography (silica-coated
glass plates) and visualized by UV light, and/or by LC-MS. NMR spectra were recorded in
1
CDCl CD OD, or acetone-d6 at 400 or 600 MHz for H NMR. Chemical shifts
δ are given in
3
,
3
using tetramethylsilane as an internal standard. Multiplicities of NMR signals are designated as
singlet (s), broad singlet (br s), doublet (d), doublet of doublets (dd), triplet (t), quartet (q), and
multiplet (m). All final compounds for biological testing were of
≥95.0% purity as analyzed by
LC-MS, performed on an Advion AVANT LC system with the expression CMS using two
28