Albumin-conjugated drug is irresistible by single gene mutation of endocytic system: Verification by genome-wide CRISPR-Cas9 loss-of-function screens

Article abstract of DOI:10.1016/j.jconrel.2020.04.035

Albumin-conjugated drugs attain KRAS mutant cancer targeting through KRAS-enhanced macropinocytosis and intensified lysosomal degradation due to reduced neonatal Fc receptor (FcRn) expression. The cytosolic delivery of active payloads relies on endocytosis and subsequent intracellular processing of albumin delivery vehicles, wherein complex regulatory mechanisms and molecular machineries are closely involved. Despite the obvious merit of KRAS targeting, could such an endocytic process involving extra molecular regulators also bring about extra vulnerabilities to albumin-conjugated drugs, particularly, unexpected drug resistance? To assess such risks, here we performed an unbiased drug resistance mechanism comparison in pancreatic cancer, between free triptolide (TP, a potent cytotoxin) and albumin-conjugated TP, using genome-wide CRISPR-Cas9 loss-of-function screens. GTF2H5, a subunit of GTF2H transcription factor complex, was the only hit identified regardless of forms of TP treatment. With drug efficacy tests on GTF2H5 knockout clones, we further concluded that GTF2H5 deficiency conferred drug resistance primarily due to the pharmacological mechanism of action (MoA) of TP. In addition, molecules previously considered to be able to affect endocytosis and intracellular processing were not enriched during the screening with albumin-conjugated TP. With the aid of genome-wide CRISPR-Cas9 loss-of-function screens, we conclude that the pharmacological resistance of the active payload, rather than any potential loss-of-function mutations in endocytic molecular machineries, is the solely crucial drug resistance mechanism of albumin-conjugated drugs.

Full text of DOI:10.1016/j.jconrel.2020.04.035

Journal Pre-proof
Albumin-conjugated drug is irresistible by single gene mutation
of endocytic system: Verification by genome-wide CRISPR-Cas9
loss-of-function screens
Fang Yuan, Mengnan Sun, Huiqin Liu, Feng Qian
PII:
S0168-3659(20)30249-2
DOI:
https://doi.org/10.1016/j.jconrel.2020.04.035
COREL 10288
Reference:
To appear in:
Journal of Controlled Release
Received date:
Revised date:
Accepted date:
25 February 2020
31 March 2020
22 April 2020
Please cite this article as: F. Yuan, M. Sun, H. Liu, et al., Albumin-conjugated drug
is irresistible by single gene mutation of endocytic system: Verification by genome-
wide CRISPR-Cas9 loss-of-function screens, Journal of Controlled Release (2020),
https://doi.org/10.1016/j.jconrel.2020.04.035
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.

Journal Pre-proof
Albumin-conjugated drug is irresistible by single gene mutation of endocytic
system: verification by genome-wide CRISPR-Cas9 loss-of-function screens
Fang Yuan1, Mengnan Sun1, Huiqin Liu1, Feng Qian*1
1
School of Pharmaceutical Sciences, Beijing Advanced Innovation Center for Structural Biology,
and Key Laboratory of Bioorganic Phosphorus Chemistry & Chemical Biology (Ministry of
Education), Tsinghua University, Beijing 100084, China.
* Correspondence to:
Feng Qian (Email: qianfeng@tsinghua.edu.cn)
Manuscript for Journal of Controlled Release

Journal Pre-proof
Abstract
Albumin-conjugated drugs attain KRAS mutant cancer targeting through KRAS-enhanced
macropinocytosis and intensified lysosomal degradation due to reduced neonatal Fc receptor
(FcRn) expression. The cytosolic delivery of active payloads relies on endocytosis and subsequent
intracellular processing of albumin delivery vehicles, wherein complex regulatory mechanisms
and molecular machineries are closely involved. Despite the obvious merit of KRAS targeting,
could such an endocytic process involving extra molecular regulators also bring about extra
vulnerabilities to albumin-conjugated drugs, particularly, unexpected drug resistance? To assess
such risks, here we performed an unbiased drug resistance mechanism comparison in pancreatic
cancer, between free triptolide (TP, a potent cytotoxin) and albumin-conjugated TP, using
genome-wide CRISPR-Cas9 loss-of-function screens. GTF2H5, a subunit of GTF2H transcription
factor complex, was the only hit identified regardless of forms of TP treatment. With drug efficacy
tests on GTF2H5 knockout clones, we further concluded that GTF2H5 deficiency conferred drug
resistance primarily due to the pharmacological mechanism of action (MoA) of TP. In addition,
molecules previously considered to be able to affect endocytosis and intracellular processing were
not enriched during the screening with albumin-conjugated TP. With the aid of genome-wide
CRISPR-Cas9 loss-of-function screens, we conclude that the pharmacological resistance of the
active payload, rather than any potential loss-of-function mutations in endocytic molecular
machineries, is the solely crucial drug resistance mechanism of albumin-conjugated drugs.
Keywords
Drug resistance, albumin-conjugated drug, CRISPR-Cas9 screen, triptolide, pancreatic cancer
Abbreviations
clustered regularly interspaced short palindromic repeats, CRISPR; CRISPR-associated protein 9,
Cas9; knockout, KO; wild type, WT; kirsten rat sarcoma, KRAS; pancreatic ductal
adenocarcinoma, PDAC; non-small-cell lung cancer, NSCLC; neonatal Fc receptor, FcRn;
triptolide, TP; mechanism of action, MoA; RNA interference, RNAi; genome-wide CRISPR-Cas9
knockout, GeCKO; single guide RNA, sgRNA; model-based analysis of genome-wide
CRISPR-Cas9 knockout, MAGeCK; 4-dimethylaminopyridine, DMAP; 1-hydroxybenzotriazole,

Journal Pre-proof
HOBT;
1-(3-Dimethylaminopropyl)-3-ethylcarbodiimide,
EDC;
5-(N-ethyl-N-
isopropyl)-amiloride, EIPA; false discovery rate, FDR; fold change, FC; half maximal inhibitory
concentration, IC50; fluorescence-activated cell sorting, FACS; matrix-assisted laser
desorption-ionization time of flight mass spectrometry, MALDI-TOF MS; general transcription
factor 2H, GTF2H; general transcription factor 2H subunit 5, GTF2H5

Journal Pre-proof
1. Introduction
Oncogenic KRAS is one of the most frequent mutations in human cancer. With a mutation
incidence as high as 90-95%, KARS activating mutations in pancreatic ductal adenocarcinoma
(PDAC) are considered to transform cells with strong oncogene addictions at the beginning of
disease progression, and efficacious inhibitions of such driver molecule may be of great value in
clinical treatment1. Pharmacological inhibitors for KRAS oncoprotein are generally challenging,
mostly due to its shallow binding pocket and high functional affinity which is historically
considered to be undruggable2. One recent breakthrough is direct targeting of KRASG12C, a mutant
allele of oncogenic KRAS, which has a cysteine residue and thus provides inhibitor covalently
binding opportunity. KRASG12C inhibitors have been demonstrated with expected efficacy in
several pre-clinical models of NSCLC and colon cancer3. However, for PDAC, KRASG12C only
represents one of many mutation versions of oncogenic KRAS (mutation rate <2% in PDAC), and
for the remaining alleles, especially KRASG12D, KRASG12V, and KRASG12R with a larger patient
population and worse prognosis, direct targeting remains challenging4.
Previously, we demonstrated that albumin conjugation was a viable oncogenic KRAS
targeted drug delivery strategy5. In non-transformed endothelial cells, the albumin-conjugated
drugs undergo clathrin-mediated endocytosis, where FcRn-mediated recycling exocytosis
minimizes the intracellular release of active payloads thus the cytotoxicity. In comparison,
enhanced macropinocytosis and lysosomal degradation of albumin-conjugated drugs occur in
KRAS mutant cancer cells, and produce significantly increased intracellular drug release.
Benefiting from these differential intracellular processing mechanisms dictated by KRAS
genotype, albumin-conjugated drugs attain targeted drug actions in KRAS mutant PDAC and
substantially enlarged the therapeutic window of payloads by ~10 times.
Distinct from free drugs, the cytosolic delivery of albumin-conjugated drugs relies heavily on
the endocytosis and intracellular trafficking within the targeted cells, wherein complex regulatory
mechanisms and molecular machineries are involved, as suggested by studies such as
genome-wide RNA interference (RNAi) screens6. It has been acknowledged that perturbations of
endocytic regulators could divert the delivery route of a drug that is conjugated to or encapsulated
in its delivery vehicle, yet it remains unknown that to what degree such perturbations could affect
the overall pharmacological efficacy of such drug delivery systems. Could these engineered drug

Journal Pre-proof
delivery systems, such as albumin-conjugated drugs, encounter extra drug resistance
vulnerabilities due to the aforementioned complexity? This requires a conclusive answer to avoid
any future application risks.
TP is a diterpenoid triepoxide extracted from Tripterygium wilfordii Hook F., and is one of
the most potent cytotoxins towards PDAC with much higher activity compared to common
chemotherapeutic agents including gemcitabine, 5-fluorouracil, nab-paclitaxel, irinotecan and
platinum7,8. TP attracted considerable interest as an anti-cancer drug candidate, while its
translation may be hampered by severe toxicities on multiple organs9. We believe that TP could be
able to offer unique clinical benefits for PDAC, if its narrow therapeutic window could be
widened by drug delivery strategies, such as albumin conjugation. Here, we exploited the
unbiased genome-wide CRISPR-Cas9 loss-of-function screening, a powerful and unprecedented
genetic tool with prior-proved feasibility and robustness, to identify any possible genes that could
critically affect the efficacy of free TP and albumin-conjugated TP10. The established
genome-wide single guide RNA (sgRNA) library includes 123,411 sgRNAs targeting 19,050
genes, and 4-6 guides per target gene on average were designed to further minimize off-target
potentials and gain high signal-to-noise interpretations11.
2. Methods and materials
2.1. Materials
Reagents were purchased from commercial sources: Triptolide (Chengdu Purifa Technology
Development Co., Ltd, China), bovine serum albumin (Amresco, USA), succinic anhydride (J&K
Scientific, China), triethylamine (Macklin, China), 4-dimethylaminopyridine (Macklin, China),
1-hydroxybenzotriazole (Macklin, China), 1-(3-dimethylaminopropyl)-3-ethylcarbodiimide
(Macklin, China).
2.2. Cell line
Pancreatic cancer cell line MIAPaCa-2 was originally obtained from American Type Culture
Collection and maintained in lab regularly. Cell culture of MIA PaCa-2 and its derived knockout
clones was based on DMEM (Gibco) medium supplemented with 10% fetal bovine serum (FBS)
(Gemini, # 900108) and 1% penicillin and streptomycin (Gibco).
2.3. Synthesis and characterization of albumin-conjugated TP

Journal Pre-proof
The synthesis of triptolide-succinate (TPS) and albumin-conjugated TP was based on a
previously reported synthetic method12. Briefly, TP (200 mg, 0.54 mM) and succinic anhydride
(300 mg, 3 mM) were dissolved into 5 mL chloroform, then 0.5 mL triethylamine (TEA) and
4-dimethylaminopyridine (DMAP, 122 mg, 1 mM) were added. The reaction solution was stirred
at room temperature overnight under nitrogen protection. The progress was detected by thin layer
chromatography with chloroform/methanol (20:1, v/v) to ensure esterification of all TP. After the
reaction, chloroform was evaporated, and the remaining solids were washed with water and
subsequently dried under vacuum. To purify TPS, a silica gel system eluting with
chloroform/methanol (20:1, v/v) was used. 205.4 mg (80.0%) faint yellow product (TPS) was
finally purified and the structure was characterized by 1H NMR (400 MHz) and MALDI-TOF MS.
BSA (100 mg, 5mg/mL) was dissolved in 0.1 M borate buffer. TPS (0.2 mM) together with
1-(3-dimethylaminopropyl)-3-ethylcarbodiimide (EDC, 100 mg) and 1-hydroxybenzotriazole
(HOBT, 50 mg) were additionally dissolved in 2 mL acetonitrile and added dropwise to the
protein solution. Subsequently the solution was stirred for 24 h in refrigerator. Insoluble products,
uncoupled TPS and others were removed by centrifuging and passing through size exclusion
chromatography twice (GE Healthcare, Superdex 200 5/150 GL # 28906561). The protein eluent
was collected in distilled water for molecular weight identification using a MALDI-TOF MS, and
the TP:albumin molar ratio in conjugates was calculated based on the molecular weight difference
between the conjugate and albumin.
For tetramethyl rhodamine (TMR) labeling, 2 mg TMR was dissolved into 100 μL DMSO
and added dropwise into the conjugate solution (100mg conjugates in PBS, pH 7.4). The mixture
was stirred at room temperature for 3-4 h. To remove uncoupled TMR, size exclusion
chromatography (GE Healthcare, Superdex 200 5/150 GL # 28906561) was engaged for
purification.
2.4. Drug release study of the albumin-conjugated TP
Release kinetics of TP from the albumin conjugates was determined by measuring the content
of free TP in a drug release study, using DMEM as the release medium to simulate the
extracellular environment. Briefly, 10 mg of albumin-conjugated TP was dissolved in 1 mL
DMEM medium at 37 ºC, and 100 μL of sample was collected at different time points and added
into 500 μL methanol/acetonitrile (3:1). To completely extract the released TP and precipitate

Journal Pre-proof
albumin, sample mixtures were vortexed and then centrifuged at 12,000 × g for 10 min at 4 ºC.
The TP concentration of the supernatant was determined using a HPLC system (Prominence-i9
LC-2030, Shimadzu) with a C18 column and a mobile phase of water/acetonitrile (25:75) at flow
rate of 1mL/min, and detection was performed at 218 nm using a PDAdetector.
2.5. Endocytosis study of albumin-conjugated TP
Macropinocytosis inhibitor, 5-(N-ethyl-N-isopropyl)-amiloride (EIPA), was used to confirm
the cell entry mechanism of albumin-conjugated TP. Cells were pre-treated with 50 μM EIPA for
30 min and then 1 mg/mL TMR-labeled conjugates were added and incubated for 2-4 h. At the
end of incubation time, cells were rinsed with PBS and collected for cytometer detection (PE
channel, BD LSRFortessa, BD Bioscience) measuring TMR fluorescent intensity.
2.6. Plasmids and library screening
The principal of CRISPR-Cas9 mediated knockout is that endonuclease Cas9 can be guided
to a specific genomic target by sgRNAs and generate double strand break and loss-of-function
mutagenesis. In this library screening, Genome-wide CRISPR-Cas9 Knockout (GeCKO) sgRNA
pooled library (Addgene # 1000000049), FUCas9-Cherry, were used to express sgRNAs and Cas9
protein as suggested9. Plasmids were firstly packaged into lentiviral particles with RSV, VSV and
pMDL packaging plasmids using FuGENE6 transfection reagent and 293 T cell. For each screen,
cherry fluorescence was confirmed before the infection to ensure expression of the Cas9
component. The sgRNA coverage was set ~100, indicating that more than 100 cells received any
single sgRNA. Cells were infected with library lentivirus at a low multiplicity of infection (MOI,
normally ~0.1-0.3 to ensure for each cell only one guide enters) with spinfection (2000 rpm, 1 h)
and incubated overnight. The medium was aspirated 17-24 h post infection and replaced with
puromycin containing medium selecting for sgRNA transfected cells (selection pressure: 0.75
μg/mL puromycin, 5 days). After selection, sgRNAtransfected cells were collected and underwent
drug selection at lethal doses for either albumin-conjugated TP (albumin-TP) or free TP. Under the
lethal dose selection pressure, cells without resistant mutagenesis could not survive, and only cells
whose loss-of-function mutations protected them from drug induced cell killing would be enriched
during the screening. The final enriched drug resistant cells were further collected, and their
genome DNAs were extracted and subjected to deep sequencing for sgRNAs inside. From the
comparison of sgRNA distribution between the experimental and control groups, the enriched

Journal Pre-proof
sgRNAs were isolated and corresponding genes that conferred drug resistance phenotype were
identified. The overall schematic illustration of the screening strategy is shown in Figure 1.
sgRNA library
A library of KO cells
Transduction
MOI ~0.3
Cas9+MIA PaCa-2
Whole-genome
19,050 genes
Selection pressure: the lowest
drug concentration sufficient to kill
all the control cells
Drug treatment
Albumin-TP
Drug treatment
No treatment
free TP
Albumin-TP resistant cells
Sequencing
TP resistantcells
Control cells
Sequencing
Sequencing
Enriched
sgRNAs
Enriched
sgRNAs
sgRNAs
MAGeCK analysis: sgRNAs distribution comparison
between experimental groups and control group
Gene whose deficiency confers drug
resistance for each drug
Figure 1, Screening design and workflow. Transducing Cas9 expressing MIA PaCa-2 cells with
the unbiased genome-wide CRISPR-Cas9 sgRNA library, and a library of knockout (KO) cells
was consequently generated. The KO cells were then divided into different groups, and one of
them stayed as control group while the others underwent drug selection with either
albumin-conjugated TP (albumin-TP) or free TP. After the drug selection, sgRNAs from drug
resistant cells were amplified for each group and subjected to deep sequencing. Through
MAGeCK analysis method, the enriched sgRNAs, sgRNAs distribution changes between
experimental group and control group, and eventually, the gene whose deficiency confers drug
resistance, would be identified.

Journal Pre-proof
2.7. Deep sequencing, MAGeCK analysis, and hit identification criteria
Genome DNA (gDNA) from enriched drug resistant cells was collected with GeneJET
genomic DNA extraction kit (Thermo K0712) and corresponding amplicon library was generated
with one-step PCR (35 cycles of reaction). Model-based Analysis of Genome-wide CRISPR-Cas9
knockout (MAGeCK) screening analysis method, a well-established algorithm, was adapted13.
Amplicon samples were sent for deep sequencing (Hiseq-PE150 from Novogene Genetics, China).
The obtained sequencing data were further analyzed with MAGeCK software (version 0.5.7) and
visualized figures were generated directly. For hit identification criteria, the identified hit should
demonstrate consistency among multiple sgRNAs designed for it, as well as a much lower p value
and false discovery rate (FDR) value compared to the rest14.
2.8. Generation of GTF2H5 knockout clones
Fgh1t-UTG vector encoding inducible sgRNA and aforementioned FUCas9-mcherry were
applied to generate loss-of-function clones15. The guide sequence was selected from GeCKO
library (GTF2H5|2|+(4 rv) UID: HGLibB_21533 sequence: GTACAGCAGAAACTGCTTCA),
and 1 µg/mL doxycycline was latter used to induce sgRNA expression. After the transduction,
cells were selected by fluorescence-activated cell sorting (FACS, GFP positive, mCherry positive,
BD Influx high-speed cell sorter, BD Bioscience) and sorted into single clones. Due to the lack of
suitable antibodies, deep sequence was utilized for confirmation of induced mutations around
target region. Primers used for PCR amplicon are: forward ACAAGCTGTCTTACAATCATGTG;
reverse CAGGGCATTGGACTCATC.
2.9. Selection pressure, cell viability, and clonogenic assay
Before screening, dose-response curves of drugs were performed to determine an optimized
selection pressure, i.e., the minimal dose that can kill all the control cells. For both free TP and
albumin-conjugated TP, a titration of concerned drug was tested for drug effects on cell viability
and clonogenic ability, from 1 nM to 100 μM for both drugs.
For cell viability assay, 1000-3000 cells were seeded into 96 well plates, and once attached,
cells were then treated with respective drugs for 36-48 h. The culture medium was aspirated and
replaced with fresh one without drug for further cell viability observation. CellTiter Glo (Promega)
was used to assess cell viability according to the product protocol, and corresponding drug

Journal Pre-proof
response curves and IC50 values were calculated by Graph Pad Prism software (version 6.0). For
clonogenic assay, 100 cells were originally seeded in 6-well plate, and then underwent drug
treatment for 36-48 h. After the treatment, medium was aspirated and replaced with fresh one.
Cells were then cultured for 7-14 days, and the colonies formed were stained with crystal violet
and quantified by Image J automatic particle counting.
2.10. Statistical Analysis
The statistical significance of CRISPR screening candidate genes was calculated by
MAGeCK algorithm (version 0.5.7)13. Log2 fold change > 1 and false discovery rate < 10% were
set as thresholds for significantly enriched genes. In the line graphs, data are presented as mean
and standard deviation (S.D.).
3. Results
3.1. Synthesis and characterization of albumin-TPconjugates
The synthesis schematic of albumin-conjugated TP (albumin-TP) is illustrated in Figure 2. TP
was firstly coupled with succinic anhydride to generate triptolide succinate (TPS) intermediate
(Figure 2A), and the structure of TPS was confirmed by both MALDI-TOF MS (Figure 2C) and
1H NMR (Supplementary Figure 1). TPS was then conjugated to BSA through the formation of
amide bonds with amine groups of lysine residues of albumin (Figure 2B), and the average
molecular weight of the purified conjugates demonstrated a clear shift from albumin at 66.35 KD
to albumin-TP at 66.96 KD (Figure 2D), based onwhich the average molar ratio of TP:albumin in
conjugates was determined to be ~1.4:1.
The albumin-TP conjugates displayed a hydrodynamic diameter below 10 nm (Figure 2E).
We further examined the drug release profile of albumin-TP in DMEM cell culture media, and
observed that less than 4% of TP was released while more than 96% albumin-TP remained
conjugated up to 48 h, indicating negligible extracellular free drug during the time course of
screening study, which is undesired and could complicate or negate the screening result of intact
albumin-TP conjugates (Figure 2F).
As reported and previously confirmed5,16, KRAS enhanced macropinocytosis is a major
endocytosis mechanism for albumin, which could be validated by selective inhibition of
macropinosome formation using EIPA. To confirm such mechanism remains for albumin-TP

Journal Pre-proof
conjugates, internalization of TMR-labeled albumin-TP conjugates by KRAS mutant cells was
quantified after EIPA blockage and a ~50% attenuation was observed, confirming strong
macropinocytosis dependence (Figure 2G). Once internalized through macropinocytosis,
albumin-drug conjugates undergo lysosomal degradation and the therapeutic payloads are
subsequently released to exert their pharmacological effects5. As expected, both albumin-TP and
free TP are highly cytotoxic towards MIA PaCa-2 cells (Figure 2H), while the relatively reduced
potency of albumin-TP (IC50 ~100 nM) is likely caused by the kinetics of intracellular TP release
throughout the dynamic endocytic transport process of albumin-TP.

Journal Pre-proof
A
B
[TP-COONa]+
C
D
50000
40000
30000
20000
10000
0
2500
483.16
2000
1500
1000
500
0
66350.6
66967.4
Albumin
Albumin-TP
50000 55000 60000 65000 70000 75000 80000
0
100
200
300
400
500
600
Mass(s/z)
Mass(s/z)
E
F
30
20
10
0
100
50
Albumin-TP
40
0
0
20
0.1
1
10
100
1000
Time (h)
Size (nm)
G
H
+ EIPA
100
TP
Albumin-TP
50
0
Control
-10 -9 -8 -7 -6 -5 -4 -3
Log[dose], M
Figure 2, Preparation and characterization of albumin-TP conjugates. (A-B) Synthesis of
triptolide succinate and further conjugation with albumin through the formation of amide bonds.
(C-D) MS spectra of TPS (C), albumin and albumin-TP (D). (E-F) Size distribution (E) and drug
release kinetics (F) of albumin-TP in DMEM cell culture medium without FBS addition, pH 7.4.
(G) Decreased internalization of albumin-TP into MIA PaCa-2 cells by EIPA blockage, upper
panel: intracellular fluorescent intensity of TMR-labeled conjugates in the presence of EIPA,

Journal Pre-proof
lower panel: intracellular fluorescent intensity of TMR-labeled conjugates without EIPA. (H)
Toxicity of free TP and albumin-TP towards MIAPaCa-2 cell viability, measured by CellTiter Glo.
Error bars represent S.D.
3.2. Enrichments of GTF2H5 loss-of-function cells in both albumin-TP and free TP screen
panels
The loss-of-function screens were performed using IC95 concentration of albumin-TP and
free TP. Based on the MAGeCK analysis of sgRNA distribution, we observed that GTF2H5
loss-of-function cells were robustly enriched in albumin-TP screen panel with a p value of
2.63E-07, a false discovery rate of 0.00495 and a calculated enrichment fold change value of 326
(Figure 3A and 3B). GTF2H5 ranked 1st by MAGeCK robust ranking aggregation (RRA)
algorithm, an algorithm that can prioritize genes enriched in the drug resistant clones compared to
control clones (Figure 3C).
Genes ranking top 10 by the algorithm were analyzed. GTF2H4 and GTF2H2C, ranking 3rd
and 4th, are molecules also from GTF2H family, working within the same complex with GTF2H5
and may therefore be true hits, though obviously not as strong as GTF2H5. ARMC5, ranking 2nd,
was also enriched in one replicate of the free TP screening, but not exclusively in albumin-TP
screening, thus was not considered to be associated with the albumin conjugation delivery system.
For other top 10 gene candidates (Figure 3D), enrichment was obtained when merely 1 out of 6
total guides was applied (Figure 3E-3M). Along with skeptical and far-outnumbered FDR values
(Table 1), an accidental outcome instead of a true hit appears to be highly likely.
GTF2H5 loss-of-function mutation was enriched not only in albumin-TP screen panel but
also in free TP screen panel. Under free TP selection, GTF2H5, stood out robustly with a p value
of 3.62E-07 (Figure 4A), a false discovery rate of 4.95E-03, a calculated enrichment fold change
value of 124 (Figure 4B). All its 6 independent guides were enriched after TP screening, and
GTF2H5 ranked 1st by the positive selection RRA score, justifying it as a true hit (Figure 4C).

Journal Pre-proof
Figure 3, Identification of GTF2H5 deficiency conferring albumin-TP resistance. (A)
Identification of top candidate gene GTF2H5 by MAGeCK p value analysis after albumin-TP
selection. GTF2H5 is labeled as a large red dot. (B) Scatterplot showing enrichment of GTF2H5
loss-of-function mutation after albumin-TP selection. Fold change for each gene represents its
frequency from albumin-TP resistant cells versus its frequency from control cells. GTF2H5 is
labeled as a large red dot. (C) Gene ranking by robust ranking aggregation (RRA) score from
MAGeCK analysis under albumin-TP treatment showing GTF2H5 outperformed others. Ranking
top 10 genes are shown in the plot. (D) Enrichment consistency among 6 guides designed for
GTF2H5 after albumin-TP selection. 5 out of 6 guides showed enrichment under the selection.
(E-M), Inconsistent guides distribution for other RRA score top 10 genes: (E) ARMC5, (F)

Journal Pre-proof
GTF2H4, (G) GTF2H2C, (H) SGCZ, (I) CCL14, (J) RAB5A, (K) has-mir-6508, (L) has-mir-4778,
(M) MICU2.
Rank
Gene
GTF2H5
ARMC5
Number of enriched sgRNAs
FDR
1
2
5
2
2
2
1
1
1
1
1
1
4.95E-03
0.056931
0.146865
0.216584
0.654455
0.659241
0.751397
0.751397
0.751397
0.751397
3
GTF2H4
GTF2H2C
SGCZ
4
5
6
CCL14
7
RAB5A
8
hsa-mir-6508
hsa-mir-4778
MICU2
9
10
Table 1, Top 10 genes ranked by RRA score in albumin-TP screen panel. Number of enriched
guides and false discovery rate (FDR) value of respective genes are denoted.

Journal Pre-proof
Figure 4, Identification of GTF2H5 deficiency conferring TP resistance. (A) Identification of
top candidate gene GTF2H5 by MAGeCK p value analysis after free TP selection. GTF2H5 is
labeled as a large red dot. (B) Scatterplot showing enrichment of GTF2H5 loss-of-function
mutation after free TP selection. Fold change for each gene represents its frequency from free TP
resistant cells versus its frequency from control cells. GTF2H5 is labeled as a large red dot. (C)
Consistent enrichment among multiple guides designed for GTF2H5 after free TP selection. all 6
guides showed enrichment.
3.3. Genes governing endocytosis or intracellular trafficking were not enriched in the
albumin-TP screen panel
Atotal of 2 genes emerged in the albumin-TP screen panel, with the threshold of FDR < 10%.
However, neither of them was related with endocytosis or intracellular vesicle processing.

Journal Pre-proof
Previously known macropinocytosis regulators such as CDC42, RAC1, PI3K and PAK117, and
molecules involved in lysosome functions including TFE3, IPO8, MITF, and LAMP218, were not
enriched after albumin-TP selection, implying loss-of-function of any of them could not rescue
cells from albumin-TP mediated cell killing (Table 2).
Number of
Gene
Log2 FC
FDR
p value
enriched sgRNAs
CDC42
RAC1
PI3K
0
0
0
0
0
0
0
0
-4.93
-4.59
0.995217
0.981163
0.9942
0.84418
0.56318
0.75334
0.39302
0.17464
0.39302
0.79629
0.61232
-4.5645
-4.33
PAK1
TFE3
IPO8
0.962701
0.936214
0.962701
0.9942
-5.23
-4.2042
-5.83
MITF
LAMP2
-5.45
0.987627
Table 2, Status of genes relevant to macropinocytosis and lysosome function in albumin-TP
screen panel. Number of enriched guides, log2 fold change (log2 FC), false discovery rate (FDR),
and p value of respective genes are denoted.
3.4. Validation of GTF2H5 deficiency conferring TP resistance
TP has been reported to be a covalent inhibitor of XPB, a subunit of GTF2H transcription
factor complex, and is able to inhibit its RNA polymerase II-mediated transcription and therefore
exert anti-proliferation effect19. Given GTF2H5 as a subunit of GTF2H, the molecular target of TP,
we hypothesized that GTF2H5 deficiency resulted in the pharmacological resistance of TP itself
but not resistance of albumin delivery strategy, and led to the drug resistance to
albumin-conjugated TP. To validate the assumption, we re-generated GTF2H5 knockout MIA
PaCa-2 clones using the CRISPR-Cas9 system. Exploited sgRNA sequence was adapted from
GeCKO library, with a puromycin selection marker. We generated and sorted the edited cells.
Clone 4 and clone 8, 2 derived clones with validated indel mutations around the CRISPR-Cas9
editing site, were used for further drug efficacy tests (Supplementary Figure 2). We found that the

Journal Pre-proof
derived GTF2H5 deficient clones showed strong resistance to TP cytotoxicity, under both
clonogenic assay (Figure 5A, Supplementary Table 1) and cell viability assay (Figure 5B).
Similarly, the albumin-conjugated TP was unable to bypass GTF2H5 deficiency, and displayed
further efficacy reduction due to their reliance on TP drug release to achieve therapeutic effect
(Figure 5C).
KO clone 8
WT cells
KO clone 4
A
B
C
150
100
50
100
50
Vector
KO clone 4
KO clone 8
KO pool cells
TP
Albumin-TP
0
-10
0
-9
-8
-7
-10 -9
-8
-7
-6
-5
-4
Log[dose], M
Log[dose], M
Figure 5, Validation of GTF2H5, whose deficiency impairs drug effect of free TP and
therefore albumin-TP. (A-B) Deficiency of GTF2H5 resulted in reduced TP cytotoxicity with
lower ability to inhibit cell clonogenesis (A) and cell viability (B). (C) Simultaneously
insensitization of cells towards albumin-TP under the GTF2H5 deficiency. Wild type (WT).
Knock out (KO). Error bars represent S.D.
4. Discussion
Unlike free drugs, drugs delivered by advanced drug delivery systems, such as protein
conjugations, nanoparticle encapsulations, etc., rely heavily on the endocytic and intracellular
processing within the targeted cells, and activate payloads become available to their sites of action
only upon the completion of such steps and released from their carriers20. A recent study showed
that endocytosis engaging active transport mechanisms might also be responsible for intra-tumor
transport of nanoparticles, challenging the traditional passive transport theory21. An efficient
endocytic transport attains desired pharmacological effects; however, a complex process
controlled by various molecular machineries could sometimes be disrupted by genetic

Journal Pre-proof
perturbations. Panarella et al. reported an RNAi-based high-throughput screen with microscopy
strategy on fluorescence labeled polystyrene nanoparticles in HeLa cells, and found that out of a
library of 408 genes, myosin and a few Rab proteins dominated endocytosis and intracellular
traffic of their polystyrene nanoparticles22. Also, Sahay et al conducted a small molecule-based
library screen combining automated confocal microscopy, and identified pathways related to
NPC1, V-ATPase, mTOR, cathepsins, and ionic balance, affected the cellular entry and trafficking
of their siRNA loading lipid nanoparticles23.
Here we aimed to investigate whether any single gene deficiency could affect
macropinocytosis, intracellular trafficking or lysosomal degradation, to the extent of jeopardizing
the efficacy of albumin-conjugated drugs. Direct microscopic observation of fluorescence labeled
particles remains as the major state-of-the-art characterization strategy to understand interactions
between drug carriers and cells. Despite its undisputable contributions to this field, this method
does not provide a systemic and mechanistic evaluation at the molecular level. In the current study,
we presented a genome-wide CRIPSR-Cas9 loss-of-function screening method to reveal the direct
and unambiguous connections between gene and the pharmacological efficacy of drug delivery
systems, and demonstrated that it could serve as a complementary research tool to provide
otherwise unreachable molecular insights. Obviously, this methodology could be similarly applied
to understand the biological interactions between other drug delivery systems with different size
or chemistry, and any other types of cancerous or non-cancerous cells. Similar strategies used in
this research to optimize the screening conditions, i.e., the determination of drug sensitive cells,
the selection of screening pressure to provide a clear-cut phenotype signal, etc., could also be
applied in future other studies.
With a low infection MOI, most cells receive only one sgRNA, and mutagenesis happens
around one gene locus per cell. A library of knockout cells wherein each cell harbored one single
gene mutation was therefore generated. Considering the selection pressure was set as the drug
concentration sufficient to kill the majority of control cells, only genes of crucial impact rather
than partial or even marginal influence would be identified. During the screening with albumin-TP,
two replicates were conducted but only TP related resistance mechanism was identified, while
none of the endocytosis or intracellular trafficking related genes was enriched. However, we do
believe that there are genes whose knockout confer partial endocytic resistance, but not to the

Journal Pre-proof
critical point to create sufficient drug resistance and rescue the cells from IC95 selection pressure.
Therefore, the albumin-conjugated drug is irresistible by any single gene mutation of the
endocytic system, and such interpretation is consistent with previous knowledge that a regulatory
network rather than one single gene underpinning cellular and subcellular processes like
macropinocytosis and lysosomal activities6,24,25. Owing to the regulatory network, compensation
mechanisms, using alternative endocytosis routes or alternative macropinocytosis mediators, could
still support albumin cellular and subcellular transport and metabolism under tough conditions. It
would certainly be interesting to carry out future studies on tumor cells suffering genetic chaos
with the existence of multiple gene knockouts, to reveal whether and to what degree a
combination of different gene perturbations could insensitize cells towards albumin-conjugated
drugs or other drug delivery systems.
It must be noted that, there could be other causes to our screen finding. One could be, there
do exist genes whose knockout could substantially negate the endocytosis or intracellular
metabolism capabilities of albumin-conjugated drugs, thus confer substantial drug resistance;
however, knockout of any of these genes causes reduction of cell viability or clonal expansion
ability and subsequently the screen selection progress would eliminate cells like this and
apparently no enrichment would be observed. For example, KRAS, an activator for
macropinocytosis and also an essential gene, did appear to following this pattern and did not
display enrichment after the screens. Several other genes related with macropinocytosis may not
be enriched after the screens either: Yao et al reported that knockdown of SDC1, which encode the
macropinocytosis mediator protein syndecan 1, would lead to decreased PDAC genesis and
maintenance both in vitro and in vivo26; Jayashankar et al observed that a mutant variant of actin
capping protein regulator CARMIL1, which aided a specific impairment of macropinocytosis in
breast cancer cells, could also inhibit retard tumor growth27. As being demonstrated again recently,
macropinocytosis inhibition could be indeed lethal for pancreatic cancer cells 28, therefore, any
un-identified genetic targets essential for macropinocytosis could be missed by this screen method.
The fact that none of the macropinocytosis related genes were enriched in this genome-wide
CRIPSR-Cas9 loss-of-function screen might have further suggested the critical role of
macropinocytosis during the proliferation of PDAC.
On the other hand, the revealed hit, GTF2H5, encodes a protein named p8 or TTDA. p8 itself

Journal Pre-proof
has been previously reported to have a critical role in GTF2H complex repair activity, where it
interacts with XPB, the major molecular target of TP19, and stimulates XPB ATPase activity29. The
additional significance of p8 includes its role as a stabilizer for GTF2H complex so as to maintain
its intracellular concentration30. The loss-of-function of p8 via some unknown mechanisms could
have hindered the binding between TP and XPB, and induced resistance against TP. The exact
molecular mechanism could become a tempting pharmacological research topic to be explored
further.
5. Conclusion
Using a genome-wide CRISPR-Cas9 loss-of-function screening method, we hereby
comprehensively assessed the risk of drug resistance against albumin-conjugated drugs under
genetic perturbation of any single gene. Albumin-conjugated TP and the free TP, were screened
for their resistance vulnerabilities. GTF2H5 deficiency, a mechanism that belongs to the resistance
mechanism of TP itself, was mutually identified in both screen panels, while no endocytosis or
intracellular processing related genes stood out in the panel of albumin-conjugated TP. Our
findings indicate that, comparing to any single gene deficiency along the endocytosis and
intracellular trafficking pathway, pharmacological resistance of the payload remains the solely
crucial drug resistance vulnerability for albumin-conjugated drugs, and the whole-genome
CRISPR-Cas9 loss-of-function screening offers an unique tool to pinpoint crucial genes governing
the efficacy of drug delivery systems.
Acknowledgements
We would like to thank Prof. David CS Huang for the GeCKO and Cas9 plasmids and
helpful discussions, Beinan Wang and Haoyu Yang for experimental advice.
Author contribution
F.Q. conceived the idea of exploiting CRISPR-Cas9 screening method to investigate the
intracellular deposition of drug delivery systems. F.Q. and F.Y designed the research and F.Y.
performed the experiments with library screening, MAGeCK analysis and preliminary validation.

Journal Pre-proof
M.S. prepared drug conjugates. F.Y. and F.Q. organized the manuscript, and H.L. assisted with the
discussion.
Grant Support
This research is supported by Beijing Advanced Innovation Center for Structural Biology,
China National Nature Science Foundation (project number 81773649), and Novagenesis
Theraputix Ltd.
Supplementary data
Attached below.

Journal Pre-proof
Reference
1.
2.
3.
4.
5.
Waters AM, Der CJ. KRAS: The Critical Driver and Therapeutic Target for Pancreatic Cancer.
Cold Spring Harb Perspect Med. 2018;8(9):a031435. doi:10.1101/cshperspect.a031435
McCormick F. Targeting KRAS Directly. Annu Rev Cancer Biol. 2018;2(1):81-90.
doi:10.1146/annurev-cancerbio-050216-122010
Canon J, Rex K, Saiki AY, et al. The clinical KRAS(G12C) inhibitor AMG 510 drives
anti-tumour immunity. Nature. 2019;575(7781):217-223. doi:10.1038/s41586-019-1694-1
Bryant KL, Mancias JD, Kimmelman AC, Der CJ. KRAS: Feeding pancreatic cancer
proliferation. Trends Biochem Sci. 2014;39(2):91-100. doi:10.1016/j.tibs.2013.12.004
Liu H, Sun M, Liu Z, et al. KRAS-enhanced macropinocytosis and reduced FcRn-mediated
recycling sensitize pancreatic cancer to albumin-conjugated drugs. J Control Release.
2019;296(January):40-53. doi:10.1016/j.jconrel.2019.01.014
6.
Collinet C, Stöter M, Bradshaw CR, et al. Systems survey of endocytosis by multiparametric
image analysis. Nature.2010;464(7286):243-249. doi:10.1038/nature08779
7.
8.
Antileukemic N. 0 = 118.09. 1970;502(9):7194-7195.
Beglyarova N, Banina E, Zhou Y, et al. Screening of conditionally reprogrammed
patient-derived carcinoma cells identifies ERCC3-MYC interactions as a target in pancreatic
cancer. Clin CancerRes. 2016;22(24):6153-6163. doi:10.1158/1078-0432.CCR-16-0149
Xi C, Peng S, Wu Z, Zhou Q, Zhou J. Toxicity of triptolide and the molecular mechanisms
involved. Biomed Pharmacother. 2017;90:531-541. doi:10.1016/j.biopha.2017.04.003
Sharma S, Petsalaki E. Application of CRISPR-Cas9 based genome-wide screening approaches
to study cellular signalling mechanisms. Int J Mol Sci. 2018;19(4). doi:10.3390/ijms19040933
Sanjana NE, Shalem O, Zhang F. Improved vectors and genome-wide libraries for CRISPR
screening. Nat Methods. 2014;11(8):783-784. doi:10.1038/nmeth.3047
9.
10.
11.
12.
Zheng Q, Gong T, Sun X, Zhang ZR. Synthesis, characterization and in vitro evaluation of
triptolide-lysozyme conjugate for renal targeting delivery of triptolide. Arch Pharm Res.
2006;29(12):1164-1170. doi:10.1007/BF02969309
13.
14.
15.
16.
Li W, Xu H, Xiao T, et al. MAGeCK enables robust identification of essential genes from
genome-scale CRISPR/Cas9 knockout screens. Genome Biol. 2014;15(12):554.
doi:10.1186/preaccept-1316450832143458
Joung J, Konermann S, Gootenberg JS, et al. Genome-scale CRISPR-Cas9 knockout and
transcriptional
activation
screening.
Nat
Protoc.
2017;12(4):828-863.
doi:10.1038/nprot.2017.016
Aubrey BJ, Kelly GL, Kueh AJ, et al. An Inducible Lentiviral Guide RNA Platform Enables
the Identification of Tumor-Essential Genes and Tumor-Promoting Mutations InVivo. Cell Rep.
2015;10(8):1422-1432. doi:10.1016/j.celrep.2015.02.002
Commisso C, Davidson SM, Soydaner-Azeloglu RG, et al. Macropinocytosis of protein is an
amino acid supply route in Ras-transformed cells. Nature. 2013;497(7451):633-637.
doi:10.1038/nature12138
17.
18.
Recouvreux MV, Commisso C. Macropinocytosis: A metabolic adaptation to nutrient stress in
cancer. Front Endocrinol (Lausanne).2017;8(SEP):1-7. doi:10.3389/fendo.2017.00261
Perera RM, Stoykova S, Nicolay BN, et al. Transcriptional control of the autophagy-lysosome
system in pancreatic cancer HHS Public Access. Nature. 2015;524(7565):361-365.
doi:10.1038/nature14587

Journal Pre-proof
19.
20.
Titov D V, Gilman B, He Q-L, et al. XPB, a subunit of TFIIH, is a target of the natural product
triptolide. Nat Chem Biol. 2011;7(3):182-188. doi:10.1038/nchembio.522
Duncan R, Richardson SCW. Endocytosis and intracellular trafficking as gateways for
nanomedicine delivery: Opportunities and challenges. Mol Pharm. 2012;9(9):2380-2402.
doi:10.1021/mp300293n
21.
22.
Sindhwani S, Syed AM, Ngai J, et al. The entry of nanoparticles into solid tumours. Nat Mater.
doi:10.1038/s41563-019-0566-2
Panarella A, Bexiga MG, Galea G, et al. A systematic High-Content Screening microscopy
approach reveals key roles for Rab33b, OATL1 and Myo6 in nanoparticle trafficking in HeLa
cells. Sci Rep. 2016;6(January):1-11. doi:10.1038/srep28865
23.
24.
25.
26.
27.
28.
Sahay G, Querbes W, Alabi C, et al. Efficiency of siRNA delivery by lipid nanoparticles is
limited by endocytic recycling. Nat Biotechnol.2013;31(7):653-658. doi:10.1038/nbt.2614
Zhang Y, Commisso C. Macropinocytosis in Cancer: A Complex Signaling Network. Trends
in Cancer. 2019;5(6):332-334. doi:10.1016/j.trecan.2019.04.002
Pelkmans L, Fava E, Grabner H, et al. Genome-wide analysis of human kinases in clathrin- and
caveolae/raft- mediated endocytosis. Nature.2005;436(7047):78-86. doi:10.1038/nature03571
Yao W, Rose JL, Wang W, et al. Syndecan 1 is a critical mediator of macropinocytosis in
pancreatic cancer. Nature.2019. doi:10.1038/s41586-019-1062-1
Jayashankar V, Edinger AL. Macropinocytosis confers resistance to therapies targeting cancer
anabolism. Nat Commun. 2020. doi:10.1038/s41467-020-14928-3
Thu PM, Zheng ZG, Zhou YP, et al. Phellodendrine chloride suppresses proliferation of KRAS
mutated pancreatic cancer cells through inhibition of nutrients uptake via macropinocytosis.
Eur J Pharmacol. 2019. doi:10.1016/j.ejphar.2019.01.060
29.
30.
Egly JM, Coin F. A history of TFIIH: Two decades of molecular biology on a pivotal
transcription/repair
factor.
DNA
Repair
(Amst).
2011;10(7):714-721.
doi:10.1016/j.dnarep.2011.04.021
Greber BJ, Nguyen THD, Fang J, Afonine P V., Adams PD, Nogales E. The cryo-electron
microscopy structure of human transcription factor IIH. Nature. 2017;549(7672):414-417.
doi:10.1038/nature23903

Journal Pre-proof

Journal Pre-proof
Graphic Abstract
Cas9
sgRNA
Genome-wide library screening to
identify genes whose loss confer
albumin-conjugated drug resistance
Albumin-TP
Macropinocytosis
Lysosomal
degradation
GTF2H5
identified
deficienc
as the on
y
ly
vulnera
bility
Cell d
eath
XPB/G
TH2H
Payload therapeutic MoA
Albumin
TP
Albumin fragments
Highlights


In KRAS mutant pancreatic cancer cells, albumin-TP conjugates rely on macropinocytosis
and lysosomal degradation to achieve cellular entry and intracellular release of TP, thus to
produce pharmacological effects and cause cell death.
With the pooled genome-wide CRISPR-Cas9 loss-of-function screens, GTF2H5 deficiency, a
resistance mechanism of the payload cytotoxic agent TP, is identified as the only drug
resistance vulnerability.

Mutations of endocytic genes are not enriched in albumin-TP drug resistant cells.

Journal Pre-proof
Author contribution
F.Q. conceived the idea of exploiting CRISPR-Cas9 screening method to investigate the
intracellular deposition of drug delivery systems. F.Q. and F.Y designed the research and F.Y.
performed the experiments with library screening, MAGeCK analysis and preliminary validation.
M.S. prepared drug conjugates. F.Y. and F.Q. organized the manuscript, and H.L. assisted with the
discussion.