Embed Size (px)
Citation preview
Article
Convergent Identification
and Interrogation ofTumor-Intrinsic Factors that Modulate CancerImmunity In VivoGraphical Abstract
Highlights
d Genome-scale in vivo CRISPR screens identified tumor-
intrinsic factors of oncogenesis
d Prkar1a is a strong genetic driver of oncogenesis
d Prkar1a loss altered the transcriptome and proteome in
immune responses
d Prkar1a mutation in cancer cells remodels the tumor
microenvironment
Codina et al., 2019, Cell Systems 8, 136–151February 27, 2019 ª 2019 Elsevier Inc.https://doi.org/10.1016/j.cels.2019.01.004
Authors
Adan Codina, Paul A. Renauer,
GuangchuanWang, ..., Jesse Rinehart,
Rong Fan, Sidi Chen
In Brief
Codina et al. performed genome-scale
in vivo CRISPR screens that robustly
identified multiple regulators of tumor-
intrinsic factors that alter the ability of
cells to grow as tumors across different
levels of immunocompetence.
Characterization ofPrkar1a, a convergent
hit from these screens, showed that its
knockdown leads to drastic alterations in
the phosphoproteome and
transcriptome, corresponding to changes
in the tumor microenvironment.
Cell Systems
Article
Convergent Identification and Interrogationof Tumor-Intrinsic Factors that ModulateCancer Immunity In VivoAdan Codina,1,2,3,4,16 Paul A. Renauer,1,2,3,4,16 Guangchuan Wang,1,2,3,16 Ryan D. Chow,1,2,3,5,16 Jonathan J. Park,1,2,3,5,17
Hanghui Ye,1,2,3,17 Kerou Zhang,3,6,17 Matthew B. Dong,1,2,3,5,17 Brandon Gassaway,1,3,13 Lupeng Ye,1,2,3
Youssef Errami,1,2,3 Li Shen,1,2,3 Alan Chang,1,2,3 Dhanpat Jain,7,8 Roy S. Herbst,8,9,10,11 Marcus Bosenberg,7,11,12
Jesse Rinehart,1,3,13 Rong Fan,3,6 and Sidi Chen1,2,3,4,5,11,14,15,18,*1System Biology Institute, Integrated Science & Technology Center, 850 West Campus Drive, West Haven, CT 06516, USA2Department of Genetics, Yale University School of Medicine, 333 Cedar Street, New Haven, CT 06510, USA3Center for Cancer Systems Biology, Integrated Science & Technology Center, 850 West Campus Drive, West Haven, CT 06516, USA4MCGD Program, Yale University, 333 Cedar Street, New Haven, CT 06510, USA5Yale M.D.-Ph.D. Program, 367 Cedar Street, New Haven, CT 06510, USA6Department of Biomedical Engineering, Yale University, 55 Prospect Street, New Haven, CT 06511, USA7Department of Pathology, Yale University School of Medicine, New Haven, CT 06510, USA8Department of Medicine, Yale University School of Medicine, New Haven, CT 06510, USA9Department of Pharmacology, Yale University School of Medicine, New Haven, CT 06510, USA10Smilow Cancer Hospital, 35 Park St, New Haven, CT 06510, USA11Yale Comprehensive Cancer Center, 20 York Street, Ste North Pavilion 4, New Haven, CT 06510, USA12Department of Dermatology, Yale University School of Medicine, New Haven, CT 06510, USA13Department of Cellular and Molecular Physiology, Yale University, 333 Cedar St., New Haven, CT 06520, USA14Immunobiology Program, The Anlyan Center, 300 Cedar Street, New Haven, CT 06520, USA15Yale Stem Cell Center, Yale University School of Medicine, New Haven, CT 06510, USA16These authors contributed equally17These authors contributed equally18Lead Contact
*Correspondence: [email protected]://doi.org/10.1016/j.cels.2019.01.004
SUMMARY
The genetic makeup of cancer cells directs oncogen-esis and influences the tumor microenvironment. Inthis study, we massively profiled genes that function-ally drive tumorigenesis using genome-scale in vivoCRISPR screens in hosts with different levels of immu-nocompetence. As a convergent hit from thesescreens, Prkar1a mutant cells are able to robustlyoutgrow as tumors in fully immunocompetent hosts.Functional interrogation showed that Prkar1a lossgreatly altered the transcriptome and proteomeinvolved in inflammatory and immune responses aswell as extracellular protein production. Single-celltranscriptomic profiling and flow cytometry analysismapped the tumor microenvironment of Prkar1amutant tumors and revealed the transcriptomic alter-ations in host myeloid cells. Taken together, our datasuggest that tumor-intrinsic mutations in Prkar1a leadto drastic alterations in the genetic program of cancercells, thereby remodeling the tumormicroenvironment.
INTRODUCTION
As a key hallmark of cancer (Hanahan andWeinberg, 2011), can-
cer cells have the ability to escape immune surveillance (Kim
136 Cell Systems 8, 136–151, February 27, 2019 ª 2019 Elsevier Inc.
et al., 2007) using mechanisms such as immuno-editing (Dunn
et al., 2002). Immune escape leads to resistance to various forms
of cancer immunotherapy, especially checkpoint blockade
immunotherapy (CBI), by primary, adaptive, or acquired resis-
tance. A greater understanding of this process is therefore of
central importance for improving clinical outcomes (Sharma
et al., 2017). Complex genetic, cellular, and immunological fac-
tors influence oncogenesis under a fully functional immune sys-
tem (Chen and Mellman, 2013). The interactions between the
immune system and cancer cells are continuous, dynamic, and
evolving. These interactions span from the initial establishment
of a cancer cell to the development of metastatic disease, which
is dependent on immune evasion, alteration of molecular fea-
tures, cellular composition, and the tumor-immune microenvi-
ronment (Chen and Mellman, 2017). Our understanding of
this phenomenon has just begun to be elucidated by pioneering
studies, which have shown both tumor-extrinsic and tumor-
intrinsic mechanisms. Tumor-extrinsic mechanisms include
non-tumor components within the tumor microenvironment
(TME), including regulatory T cells (Tregs), myeloid-derived sup-
pressor cells (MDSCs), M2 macrophages, and other inhibitory
immune checkpoints. All of these are able to contribute to the in-
hibition of anti-tumor immune responses and are thus also of
fundamental importance to CBI efficacy (Sharma et al., 2017).
Tumor-intrinsic mechanisms of immune evasion include a lack
of antigenic proteins, low mutational burden, impaired antigen
presentation by the silencing or deletion of key molecules such
as Transporter Associated with Antigen Processing (TAP),
Beta-2-Microglobulin (B2M), and Human Leukocyte Antigen
(HLA) (Sharma et al., 2017). Additionally, constitutive expression
of immunosuppressive ligands such as Programmed Cell Death
Ligand 1 (PD-L1), either via transcriptional activation, epigenetic
regulation, or truncation of the 30 UTR has been shown to be im-
mune suppressive (Kataoka et al., 2016). Immune evasion can
also arise from oncogenic pathways that directly influence the
malignant nature of cancer cells, resulting in the alteration of
the tumor-immune microenvironment. Examples include onco-
genic signaling through the MAPK pathway resulting in the inhi-
bition of T cell recruitment and function (Liu et al., 2013), loss of
PTEN that enhances PI3K or ATK signaling (Lastwika et al., 2016;
Parsa et al., 2007), EGFR mutations that activate PD-1 pathway
(Akbay et al., 2013), overexpression of MYC that enforces the
expression of CD47 or PD-L1 (Casey et al., 2016), expression
of CDK5 to dampen the ability of T cells to reject tumors (Dorand
et al., 2016), stabilization of b-catenin resulting in constitutive
WNT signaling that induces T cell exclusion (Spranger et al.,
2015), modulation of interferon gamma (IFNg) receptor signaling
molecule Janus kinase 2 (JAK2) (Zaretsky et al., 2016), direct
changes in the expression of INFg receptors 1 and 2 (IFNGR1
and IFNGR2) and interferon regulatory factor 1 (IRF1) (Gao
et al., 2016), as well as epigenetic modifications of immune
genes (Heninger et al., 2015).
Several screening systems in human cells and mouse models
have previously demonstrated success in identifying essential
genes in cancer immunotherapy (Manguso et al., 2017; Pan
et al., 2018; Patel et al., 2017). We reasoned that performing
in vivo tumorigenesis screens in hosts with different levels of
immunocompetence would provide an approach to discover
genetic factors in cancer cells that mediate tumor progression
and immune response. Rag1�/� mice lack mature T and
B cells because of the deficiency of Rag1.Nu/Nu are Foxn3 defi-
cient and athymic. C57BL/6 mice are commonly used wild-type
mice that are fully immunocompetent. Thus, these mice present
with different levels of immunocompetence. In this study,
through robust genetic screening, we have convergently identi-
fied Prkar1a as a potent suppressor of oncogenesis in multiple
tiers of immunocompetence. We functionally interrogated its
role in regulating gene expression and cellular properties for
modulating the immune response and remodeling the TME.
RESULTS
Convergent In Vivo Genome Scale Knockout ScreensIdentify Prkar1a as a Suppressor of Oncogenesis acrossMultiple Hosts with Different Immune CompetenceWe used a genome-scale in vivo CRISPR screening approach
to identify oncogenic drivers inmultiple hostswith different levels
of immune competence. Specifically, we utilized a p53�/�;Myc-
primed mouse hepatocyte cell line (denoted IM, or IMC9 after
transduction with Cas9-GFP), which is immortalized but not
transformed. We screened this line in multiple genotypes,
including Rag1�/� mice, Nu/Nu mice, and fully immunocompe-
tent C57BL/6 mice as a syngeneic host (Figure 1A). The hepato-
cytes were capable of forming tumors in Rag1�/� mice (10/10,
100%) but not Nu/Nu mice up to an injection dose of 10 million
cells (0/10, 0%), suggesting that the immune system in Nu/Nu
mice can keep tumor growth in check. In sharp contrast,
when we mutagenized the cells with a genome-scale CRISPR
knockout library (mBrie) (Doench et al., 2016) (STAR Methods),
all (21/21) injections led to rapid tumor formation (Kolmogorov-
Smirnov [KS] test, p = 7e�4) (Figures 1A, 1B, and S1A;
Table S1). Neither the parental line (0/4) nor themBrie library mu-
tagenized pool (0/20) was able to generate tumor in C57BL/6
mice observed for at least 90 days. These data together sug-
gested that the loss-of-function mutations in a subset of genes
enabled the cells to robustly grow tumors within the Nu/Nu host.
To determine which mutants comprised the tumors that grew
inNu/Numice, we deep-sequenced the sgRNA library represen-
tations within the tumors, plasmid, and transduced cells prior to
transplantation (STAR Methods). Library representation ana-
lyses revealed high coverage in three infection control replicates
(97.8%on average) and recurrent patterns of sgRNA representa-
tion in tumors (correlation test among tumors, p < 2.2e�16) (Da-
taset S1; Figures S1B and S1C). Analyses of gene-targeting and
non-targeting sgRNA distributions revealed high consistency
between replicates and the overall shifts among different sample
types (Figures S1D and S1E), providing a high-quality genome-
scale in vivo screen dataset for analysis of enriched sgRNAs in
the outgrown tumors. When comparing the mean library repre-
sentation of in vivo tumor cells to that in pre-injection cells, we
detected a population of sgRNAs that were highly enriched in
the tumors,many of which targeted the same genewith indepen-
dent sgRNAs, whereas non-targeting control (NTC) sgRNAs
rarely conferred tumor outgrowth (Figures 1C, S2A, and S2B).
By ranking the genes by the number of independent sgRNAs
across all 21 tumors, we found that 2,335 genes have at least
one significant sgRNAs passing a false discovery rate (FDR) cut-
off of 0.2% (Figure 1D). Out of those, 120 genes had more than
two significant sgRNAs, and 12 genes had all four of their target-
ing sgRNAs significantly enriched, serving as independent evi-
dence of the knockout effect (Figure 1D). We then analyzed
sgRNA enrichment using criteria for sgRNA abundance and sig-
nificant incidence rate across independent tumors (Figure 1E)
and independent sgRNAs. This approach identified genetic
mutants that represent the strength of selection, prevalence of
phenotype, and robustness of constructs. Our analyses revealed
a three-way convergence of robust hits, identifying 11 genes that
were significantly enriched over all statistical criteria, including
Apc, Csnk1a1, Nf2, Prkar1a, Pten, Rbsn, Serpinb7, Tsc1, Tsc2,
Vps45, and Zbtb40 (Figure 1E). Additional analyses using RIGER
(STARMethods), an established CRISPR screen analytical pipe-
line that utilizes the second best performing sgRNAs, revealed a
highly similar set of genes, with 16 genes at themaximum level of
statistical significance (Tables S2 and S3; Figures 1F and 2A).
Further analysis using MAGeCK, a model-based approach with
consideration of the empirical null distribution of NTCs (STAR
Methods), also identified a similar set of genes, with 9 genes
having a top-level statistical significance of positive selection
signature (Tables S4 and S5; Figures 1G and 2A). Therefore,
this screen identified high-confidence candidate genes that
were strongly enriched in tumors that escaped the immune sys-
tem of the Nu/Nu host.
To further enhance the rigor and stringency of identification, we
repeated the screen with an independent CRISPR knockout li-
brary (mGeCKO) using the same optimized screening strategy
(Dataset S2). This screen identified five top genes that passed
Cell Systems 8, 136–151, February 27, 2019 137
A
B C D
E F G
Figure 1. Genome-Scale In Vivo CRISPR Screens in Different Mouse Hosts Identified Genes Regulating Oncogenesis
(A) Schematics of the experiment. Left, p53�/�;Mycmutated non-malignant hepatocyte cell lines are capable of growing as tumors in immunodeficient Rag1�/�
mice with deficiency in both innate and adaptive immune system; however, they are rejected byNu/Numice. Right, genome-scale CRISPR screen and validation
for mutants that can grow as tumors. Genome-scale CRISPR libraries (mBrie or mGeCKO) mutagenized cell pools have robust tumor growth in Nu/Nu host.
Validation of individual gene mutant cells with top hits in mouse hosts with different levels of immune competence.
(B) Tumor growth curves of mBrie library transduced hepatocytes (n = 21), as compared to various injections using the parental cell line (n = 3 each) over a 30-day
period (Kolmogorov-Smirnov [KS] test, mBrie versus all controls, p = 7e�4). Error bars are SEM.
(C) A color-coded scatterplot showing hit identification for genes whose loss of function lead to consistent tumor growth. The screen was performed using a
second-generation genome-scale CRISPR knockout library (mBrie) transforming hepatocytes that can grow in fully immunodeficient Rag1�/� mice but not in
Nu/Numice that have competent innate immunity although lacking mature T cells. Library-transduced cells were capable of escaping the innate immune system
and grew as tumors at 100% penetrance. Average sgRNA library representation across all Nu/Nu tumors (n = 21) was plotted against average sgRNA library
representation across all cell replicates (n = 3). The 11 top hits that pass all three statistical criteria were plotted as color-coded dots, with each dot representing
one sgRNAwhere the same color represents independent sgRNAs for the same gene. Non-targeting control sgRNAs (NTCs) were shown as dark gray dots. Other
gene-targeting sgRNAs (GTSs) were shown as light gray dots.
(D) A barplot of the number of independent scoring sgRNAs in themBrie screen. False discovery rate (FDR) of 0.2%was used as a cutoff for scoring in each tumor.
Four sgRNAs per gene were contained in the mBrie library. Highly consistent hits withR3 scoring sgRNAs (14 genes with 3/4 scoring sgRNAs, 12 genes with 4/4)
were shown.
(E) A Venn diagram showing the overlap of hits identified using three selection criteria, revealing 11 genes significant regardless of criteria. The first criterion is by
abundance (i.e., a gene with one or more sgRNA(s) comprising ofR2% total reads in a tumor); the second criterion is by phenotypic penetrance (i.e., a gene with
the frequency of one or more sgRNA(s) passing FDR 0.2% cutoff across 25% of all independent tumor samples); the third criteria is by independent construct
(i.e., a gene with R2 sgRNA(s) passing FDR 0.2% cutoff).
(F) RIGER analysis of the genome-scale screen showing top hits.
(G) MAGeCK analysis of the genome-scale screen showing top hits.
See also Figures S1 and S2.
138 Cell Systems 8, 136–151, February 27, 2019
Figure 2. Convergent Screen Analysis and Validation Revealed Prkar1a Mutant Cancer Cell Robustly Generate Tumors in a Fully Immuno-
competent Host
(A) A Venn diagram showing the convergence of two genome-scale in vivo CRISPR screens, with hits identified using the custom 3-Way Hit Caller pipeline. (Left)
The screen was repeated with a genome-scale CRISPR library (mGeCKO), and 5 genes were identified as top hits passing all statistical criteria. ThemBrie screen
andmGeCKO screen shared 3 hits:Prkar1a,Apc, and Tsc2. (Middle) Convergence of two genome-scale in vivoCRISPR screens, with hits identified using RIGER.
(Right) Convergence of mBrie in vivo CRISPR screens, showing consistent hits using three different hit calling algorithms (custom 3-Way HitCaller, MAGeCK,
and RIGER).
(B) Quantification of tumor sizes of the top screen hit mutants. (Top) tumor volume measurement of mutants in C57BL/6J host 2 weeks post injection. Prkar1a
mutant cells grew rapidly in this host, whereas NTC, Apc, Tsc2, or Csnk1a1 mutants did not (n = 10 each); (Bottom) tumor volume measurement of mutants in
Nu/Nu host 2.5 weeks post injection. Two-sided unpaired Mann-Whitney test was used to assess statistical significance for Prkar1a, Apc, Tsc2, or Csnk1a1
mutants, relative to NTC (ns, not significant, i.e., p > 0.05; *, p < 0.05; **, p < 0.01; ***, p < 0.001).
(C) The rate of tumor incidence of the top screen hit mutants. (Top) The tumor incidence rate of mutants in C57BL/6J host. Two-sided Fisher’s exact test was used
to assess statistical significance for Prkar1a, Apc, Tsc2, or Csnk1a1 mutants, relative to NTC (ns, not significant, i.e., p > 0.05; *, p < 0.05; **, p < 0.01;
***, p < 0.001).
(D) Representative H&E images of tumors in (C). Scale bar, 250 mm.
See also Figures S2 and S3.
Cell Systems 8, 136–151, February 27, 2019 139
A
B C D
E
F G
(legend on next page)
140 Cell Systems 8, 136–151, February 27, 2019
the stringent statistical criteria using our custom 3-way hit caller
pipeline (FDR < 0.2%, R2 sgRNAs, R2% read abundance, and
R25% of all tumors), of which three (Apc, Prkar1a, and Tsc2)
overlap with the most stringent hits from the mBrie screen
(Figure 2A). Similarly, four genes were identified as overlapping,
stringent hits from both screens using RIGER and MAGeCK
(Figure 2A). Multiple independent sgRNAs targeting these genes
were consistently found at high levels across tumor samples
relative to the infection-control cell samples, upon verification
of single genes (Figures S2C–S2F). Apc, Prkar1a, and Tsc2 are
the three genes that robustly emerged independent of statistical
criteria, computational algorithm, and screening library (Figures
2A and S2C–S2F).
To validate Prkar1a and other top candidates, we performed
tumor challenges that used single sgRNAs to target these genes
in immortalized hepatocyte cells, followed by transplantation
into Nu/Nu mice or fully immunocompetent C57BL/6 mice.
Compared to NTC controls, all four top mutants identified from
the screens were validated as potent tumor generators when
transplanted into Nu/Nu mice (Figures 2B–2D and S3A–S3E).
In sharp contrast, only the Prkar1a mutant is capable of gener-
ating tumors in immunocompetent C57BL/6 mice (Figures 2B–
2D), whereas mice challenged by cells with strong drivers
mutated, including Apc, Tsc2, and Csnk1a1, remained tumor-
free for the whole duration of the experiment (60 days) (Figures
2B–2D). These data indicated that Prkar1a knockout endowed
the immortalized hepatocytes with the ability to develop tumors
in fully immunocompetent hosts.
Phosphoproteomic Characterization of Prkar1aMutants Identified Its Downstream Signaling TargetsBecause of the strong phenotype of Prkar1a knockout, we set
out to interrogate the molecular regulation of Prkar1a and its
downstream pathways. Prkar1a encodes the alpha regulatory
subunit of the cAMP-dependent type I protein kinase (Protein
kinase A, PKA). As a result of the potential direct regulatory roles
of Prkar1a in phosphorylation signaling, we investigated the
changes in kinase activity and the phosphoproteome upon
Figure 3. Phosphoproteomic Characterization of Prkar1a Mutant Cells
(A) Schematics of Prkar1a phosphoproteomic experiments. Two independent set
proteomics approach that included stable isotope labeling using amino acids
enrichment of phosphopeptides, cation exchange chromatography (CEX) fracti
quently quantified, identified, and assembled into groups of proteins, based on s
(B) Waterfall plots of the top upregulated and downregulated proteins as MS-ide
mutants (Prkar1a/NTC log2 normalized-protein ratio [nr]) are plotted. As expecte
(C) Color-coded scatterplot of differential phosphorylation between independen
mutant sample (Prkar1a/NTC log2 nr) are plotted for each phosphorylation site (PS
red dots, and top differential PS (DPS, >3-fold change) are labeled with protein s
(D) Western of STAT3 and phospho-STAT3 (pSTAT3) in Prkar1a mutant and NTC
(E) Waterfall plots of the top hyper-phosphorylation site (PS) and hypo-PS in Prk
mutants (Prkar1a/NTC log2 nr) are labeled with the gene symbol (bold), UniProt a
peptide sequence in a 31 residue window, centered on the PS residue.
(F) Waterfall plot of the Ingenuity Pathway Analysis of protein groups with PS
pathways are ranked by significance (�log10 p value), and the bar colors represe
(G) Identification of hyper- and hypo-phosphorylation motifs in Prkar1amutant ce
hypo-PS (bottom) motifs that are over-represented in Prkar1amutant cells. Each
score for the enrichment below the motif. (Right) Logos are shown for the top kin
(bottom) motifs in Prkar1amutant cells. Up to four of the top kinase-substrate mo
substrate motifs are shown with the significance of the motif match (upper right of
(upper left of each Logo).
Prkar1a loss via stable isotope labeling using amino acids in
cell culture (SILAC) (Figure 3A). Because of the amount of
materials needed for SILAC experiments, we performed the
proteomics experiments in culture outside a host context, with
the goal to map out the cancer cells’ intrinsic properties.
We labeled the Prkar1a mutant cells (sgPrkar1a) with Lys6/
Arg6 ‘‘heavy’’ stable isotope and control cells (NTC) with Lys0/
Arg0 ‘‘light’’ stable isotope, and proteins were extracted,
mixed, trypsin-digested, TiO2-enriched for phosphopeptides,
fractionated by cation exchange chromatography (CEX) and
analyzed by liquid chromatography-tandem mass spectrometry
(LC-MS/MS) using an LTQ Orbitrap XL (STAR Methods). Two
independent SILAC-pair replicates were used to compare
Prkar1a mutants to control cells at the peptide, protein, and
post-translational modification levels, at which we identified a
respective 19,979 unique peptides (peptide spectrum matches,
PSMs), 2,940 protein groups and 2,994 phosphorylation sites
(PSs) after strict quality control filtering (Datasets S3, S4, and
S5). From this dataset, we identified differentially regulated pro-
teins, as well as a set of differentially phosphorylated peptides
upon Prkar1a loss, based on phosphorylation fold change at a
peptide PS and by the number of differentially phosphorylated
peptides in a protein group (Figures 3B, 3C, and 3E). As a robust-
ness check, PRKAR1A protein level was confirmed to be one of
the most downregulated proteins in the sgPrkar1a group (Fig-
ure 3B), validating the proteomics measurements. Additional
validation of the data was assessed by the western blot quanti-
fication of STAT3, one of the most upregulated proteins in the
dataset and a key player in immune and inflammation signaling
(Figures 3B and 3D). The proteins with the most significant hy-
per-phosphorylation upon Prkar1a loss included IWS1, LIMCH1,
and GOLGB1 (Figure 3E). The phospho-signaling pathway
changes in Prkar1a mutants, based on differentially phosphory-
lated proteins, showed that the most significantly enriched
pathway change in Prkar1a mutants was integrin-linked kinase
(ILK) signaling, Rho GTPase signaling, PKA signaling, mTOR
signaling, JAK/STAT signaling, and others, all of which modulate
oncogenesis and extracellular matrix (ECM) remodeling
Identified Its Downstream Signaling Targets
s of Prkar1amutant and NTC cells were compared using a multiplexed shotgun
in cell culture (SILAC), followed by protein extraction, trypsin digestion, TiO2
onation, and LC-MS/MS analysis by LTQ Orbitrap XL. Peptides were subse-
hared peptide sequences.
ntified protein groups in Prkar1a mutants. The protein level changes in Prkar1a
d, the PRKAR1A protein is among the most downregulated.
t phosphoproteomic experiments. Phosphorylation changes in each Prkar1a
). Hypo-PS and hyper-PS (>1.5-fold change) are shown as respective blue and
ymbols.
cell samples.
ar1a mutants (Left and right panels, respectively). The PS changes in Prkar1a
ccession number, and phosphorylated residue (in parentheses), as well as the
changes (DPS; >1.5-fold change) in Prkar1a mutants. The top ten canonical
nt the activation Z score, which is presented to the right of each bar.
lls. (Left) Sequence log-odds diagrams (Logos) are shown for hyper-PS (top) or
motif is presented above the Logo with an asterisk following the PS and a motif
ase-interacting motifs that match to the identified hyper-PS (top) or hypo-PS
tifs (Bonferroni-adjusted p < 1e�3) are presented for each PS motif. All kinase-
each Logo) and the kinase name with phosphorylated residues in parentheses
Cell Systems 8, 136–151, February 27, 2019 141
A
−2
−1
0
1
2
sgPrkar1a
z-score
sgNTCsg1 sg1 sg1 sg1sg1 sg1 sg2 sg2 sg2sg3 sg3 sg3 sg3sg3sg3 -6 -4 -2 0 2 4 6
sgPrkar1a/sgNTC Expression(log2 fold change)
0
40
80
120260
300
pvalue(-log
10)
●●●●●●●●●●●●
Extracellular regionExtracellular spaceProteinaceous extracellular matrixExtracellular exosomecAMP catabolic processResponse to drug3',5'-cyclic-AMP phosphodiesterase activityExtracellular matrix
Molecular mediator of immune responseRibonucleotide catabolic processFatty acid metabolic process
GSEA: Upregulated with sgPrkar1a
responseMolecular mediator of
immune responseRibonucleotide
catabolic process
-log10 FDR q-value
Fatty acidmetabolic process
0.0 0.5 1.0 1.5 2.0 2.5
DAVID: Upregulated with sgPrkar1a
-log10 q value0 2 4 6 8 10 12
3',5'-cyclic-AMPphosphodiesterase activity
Response to drug
cAMP catabolic processExtracellular exosome
Proteinaceous extracellularmatrix
Extracellular space
Extracellular region
0.3
Enrichmentscore
0.20.1
0.40.50.60.7
0.0
sgPrkar1ahigh
sgNTChigh
B
C F G
H
D
E
sgPrkar1asgNTCsg1 sg1 sg1 sg1sg1 sg1 sg2 sg2 sg3sg3 sg3 sg3 sg3sg2sg3
FmodCol6a3Col5a2Col8a2PrelpCol15a1Mamdc2CrtapCcdc80Ccdc80CtgfCtgfCyr61Adamts20Mmp8HpseSpon2VcanEcm1Ccbe1Ecm1Col24a1Lama4Ecm1Sfrp1Adamts9Wnt11VcanThbs4EpycHmcn1
-2 -1 0 1 2
Proteinaceous extracellular matrixAcute inflammatory response
-2 -1 0 1 2
Il20rbPhb2OsmrAsh1lC2Adam8C2Ash1lCfhCfhCfpRhbdd3Phb2Il20rbOsmrCfpRhbdd3PhbRhbdd3CfhIl20rbPhb2PhbC1qbpPhb2Il20rbCd55Tnfrsf11aCd55Il6Ccr7PpargPtgs2Tnfrsf11aPla2g4aPhb2C1qbpPhbOsmrCd55Cd55PhbOsmrPhb2C8gRhbdd3Cd46Ash1lRhbdd3SpnCd55CfhSusd4C2Il6stPros1Il6stGstp1Gstp1Masp1Il6stC3Tnfsf11
Serping1C4bPhbOsmrSerping1Rhbdd3Rhbdd3Adora1C1qbpAdam8CfhOsmrIl6stSpnC8aIl6st
sgPrkar1asgNTCsg1 sg2 sg2 sg2sg1 sg1 sg3 sg3 sg1sg3 sg3 sg3 sg1sg1sg3
z Score
sgPrkar1a
sgNTC
0
200
400
600
800
1000
ELISA: Secreted IL-6
pg/mlIL-6
sgPrkar1a
sgNTC
* below detection
*
p <0.0001
(legend on next page)
142 Cell Systems 8, 136–151, February 27, 2019
(Figure 3F). Comparative analysis of all PSs as a set showed a
shift in the kinase activity of Prkar1amutants, emphasized by al-
terations in the targetmotifs of differentially phosphorylated pep-
tides, for which the hypo-phosphorylated peptides are enriched
in the S*Pmotif, whereas the hyper-phosphorylated peptides are
enriched in RRxS* and RxxS* motifs (Figure 3G). These data
together revealed the proteome and phospho-proteome level al-
terations in Prkar1a mutant cells.
Whole-Transcriptome Profiling Identified DeregulatedGene Expression Programs in Prkar1a MutantsTo further dissect the gene product changes downstream of
phosphorylation, we utilized mRNA sequencing (mRNA-seq) to
profile the transcriptome of Prkar1a mutant cells together with
matched NTCs in culture, using three independent sgRNAs for
sgPrkar1a and two independent NTCs, with three biological
replicates for each sample group (STARMethods). We collected
total RNA from IMC9 cells at 7 days post-infection in biological
triplicates and performed strand-specific, poly-A enriched
mRNA-seq (Dataset S6). Using a differential expression analysis,
we identified a set of 371 transcripts that are differentially ex-
pressed between Prkar1a mutant and wild-type cells, with 119
transcripts being downregulated and 252 upregulated upon
Prkar1a loss (Dataset S7; Figures 4A and 4B). These differentially
expressed genes belonged to a distinct set of pathways with the
most significantly upregulated genes centered on immune, in-
flammatory, and secreted protein signatures (Figure 4B). Gene
Ontology analysis by the Database for Annotation, Visualization
and Integrated Discovery (DAVID) revealed that extracellular
proteins were among the top categories of Prkar1a-upregulated
genes, along with the anticipated on-target cAMP catabolic
process signature mediated by the feedback regulation of
PKA signaling via the Phosphodiesterases (Pde) gene family
(Tables S6 and S7; Figure S4A). The immune, inflammatory,
and cAMP-catabolism signatures were recapitulated with gene
set enrichment analysis (GSEA) (Tables S8 and S9; Figures 4D
and 4E). A closer look of specific transcripts revealed that
Figure 4. Whole Transcriptome Profiling of Prkar1aMutants Identified
Extracellular Protein Production
(A) An overall heatmap of all differentially expressed genes between Prkar1a mu
pendent NTCs were subjected to whole transcriptome analysis by mRNA-seq (n
gulated or downregulated in Prkar1a mutants) is shown by Z scores.
(B) A color-coded volcano plot of all differentially expressed transcripts betwee
corrected p value) was plotted against the log2 fold-change of normalized gene
most highly enriched gene sets, described in the legend.
(C) Waterfall plots of most significantly enriched gene ontologies for genes that
by the Database of Annotation, Visualization and Integrated Discovery (DAVID), a
sets including extracellular or secreted proteins emerged on top as most enric
phosphodiesterase activity, indicated the on-target activity of Prkar1a-regulated
(D) A waterfall plot of most significantly enriched gene sets of all Prkar1a upregula
levels considered during enrichment calculation (methods), and the top gene set
inflammatory response and molecular mediator of immune response emerged o
catabolic process indicated the on-target activity of Prkar1a-regulated PKA path
(E) A GSEA individual pathway analysis plot showing the enrichment of acute infl
(F) A heatmap of all differentially expressed transcripts between Prkar1a mutant
Relative gene expression levels were shown as Z scores.
(G) A heatmap of all differentially expressed transcripts between Prkar1amutant a
Relative gene expression levels were shown as Z scores.
(H) A barplot of IL-6 ELISA between Prkar1a mutant and control cells.
See also Figure S4.
Prkar1a significantly upregulated genes encode inflammatory
and secretory factors, such as cell surface proteins, cytokines,
ECM proteins, and molecular mediators of immune response
(Figures 4F, 4G, and S4B). Using ELISA, we further confirmed
that Prkar1a mutant cells turned on the protein production
switch of interleukin (IL)-6, a pro-inflammatory cytokine with
important roles in promoting tumor progression (Coussens
and Werb, 2002) (Prkar1a versus NTC, >800-fold change;
p < 0.0001; Figure 4H). Moreover, STAT3, the transcription factor
associated with IL-6 expression was also found to be upregu-
lated at the mRNA level, showing consistency between our
proteomic and transcriptomic data sets (p = 6.1e�25). Echoing
the phenotypes, these data suggest that cancer-cell-intrinsic
Prkar1a-loss leads to upregulated production of secreted cyto-
kines and extracellular proteins implicated in the inflammatory
response, which may influence the host TME.
Single-Cell Transcriptome Profiling Mapped the TumorMicroenvironment ofPrkar1aMutant Tumors in Hosts ofVarying ImmunocompetenceTo directly interrogate the host TME, we performed single-cell
RNA sequencing (scRNA-seq) (STAR Methods), an approach
that can elucidate both the gene expression signature and
cell population structure in a high-throughput manner. We re-
challenged Rag1�/�, Nu/Nu, and C57BL/6 mice with Prkar1a
and NTC cells (Figure 5A). As expected, Prkar1a cells grew tu-
mors in all three mice genotypes, whereas NTC cells only grew
in Rag1�/� (Figures 5A and 5B). We then sought to profile the
tumor-infiltrating immune repertoire and simultaneously profile
the cancer cells in the same tumor. However, since the vast
majority of cells in the tumor single-cell suspensions are
cancer cells, we performed fluorescence-activated cell sorting
(FACS) to purify both CD45+ immune cells and GFP+ cancer
cells. Using these enriched immune cell populations, we
subsequently spiked-in a smaller fraction of cancer cells taken
from the same tumor. The single cells were barcoded and had
cDNA libraries constructed in a single-cell manner, based on
an Upregulated Gene Expression Program of Immune Response and
tant and control cells. Three independent sgRNAs for Prkar1a and two inde-
= 3 biological replicates for each sgRNA). Differential gene expression (upre-
n Prkar1a mutant and control cells. The statistical significance (�log10 FDR-
expression levels. Each dot represents one gene, which is color-coded by the
are upregulated in Prkar1a mutants. The enrichment analysis was performed
nd the top gene sets are shown, as ranked by FDR-corrected q values. Gene
hed; Gene sets including cAMP catabolic process and response, as well as
PKA pathway.
ted genes. Pathway analysis was performed using GSEA with gene expression
s are shown, as ranked by FDR-corrected q values. Gene sets including acute
n top as the most significantly enriched ones; the gene set of ribonucleotide
way; gene sets are not mutually exclusive.
ammatory response genes in the Prkar1a mutant cells as compared to NTC.
and control cells for the genes in the acute inflammatory response category.
nd control cells for the genes in the proteinaceous extracellular matrix category.
Cell Systems 8, 136–151, February 27, 2019 143
A
C
B
Figure 5. Single-Cell Profiling of Prkar1a Mutant Tumors in Hosts of Different Immune Competence
(A) Schematic for single-cell RNA-seq (scRNA-seq) experiments. IMC9 tumor cells were transfected with sgPrkar1a or NTC to generate Prkar1a knockout or
control cells, respectively. Transfected IMC9 cells were then subcutaneously injected into immunocompromised (Rag1�/�), partially immunocompromised
(Nu/Nu), and immunocompetent (C57BL/6) mice as shown, and the resulting tumors were extracted, and then sorted to CD45+ tumor-infiltrating immune cells
and tumor cells that were mixed in a 1:10 ratio for scRNA-seq. One representative mouse in each group was sequenced.
(B) H&E staining of tumors subjected to tumor-immune scRNA-seq. Scale bar, 250 mm.
(C) Cellular populations within the TME, visualized by t-distributed stochastic neighbor embedding (t-SNE) plot of clustered, dimensionality-reduced single-cell
transcriptional profiles. Each dot represents a single cell that is arranged relative to other cells based on transcriptional profiles, such that similar cells are grouped
together.
See also Figures S5 and S6.
previously established methods (Gierahn et al., 2017; Macosko
et al., 2015) (STAR Methods) (Figure 5A) (Datasets S8, S9, S10,
S11, S12, S13, and S14; Tables S10, S11, S12, S13, and S14).
The single-cell transcriptional data were dimensionality-reduced
by principal-component analysis (PCA), visualized by t-distrib-
uted stochastic neighbor embedding (t-SNE) and clustered by
shared nearest neighbor (SNN) methods. The single-cell clus-
tering identified distinct different subpopulations of the
sequenced TME cellular repertoire for the Prkar1a and NTC tu-
mors in Rag1�/� mice as well as the Prkar1a tumors in Nu/Nu
and C57BL/6 mice (Figure 5C).
By merging the transcriptome data from the single-cell popu-
lations of sgPrkar1a- and sgNTC-generated tumors in Rag1�/�
mice, we identified 3 clusters (Figure 6A), based on the optimal
cell separation in the t-SNE embedding, canonical immune
144 Cell Systems 8, 136–151, February 27, 2019
gene expression patterns, and optimal rank selection using
non-negative matrix factorization (NMF) (Brunet et al., 2004;
Gaujoux and Seoighe, 2010). We distinguished cluster 1 and 2
as immune cells, largely comprising myeloid populations, based
on the significantly upregulated expression of Cd14, Fcgr3
(Cd16), and Fcgr2b (Cd32), as well as increased expression of
Itgam (Cd11b) (MAST analysis; adjusted p < 1e�9) (Figures 6C
and S6), whereas cluster 3 cells are primarily cancer cells, based
on expression of tumor-specific marker genes, Mgp, Fkbp11,
Cdkn2a, andMrpl33, identified by comparing themost highly ex-
pressed genes in the IMC9 cells from bulk RNA-seq analysis to a
reference panel of immune cells comprising Immunological
GenomeProject data (ImmGen version 1) (Heng et al., 2008) (Fig-
ures 6C, S5B, and S6). In addition, we classified the clusters into
stromal and innate immune cell types by comparing the mean
Figure 6. Differential Single-Cell Expression Analysis of the Immune Microenvironment in the Tumors Induced by Prkar1a Mutant and
Control Cancer Cells in Rag1–/– Mice
(A) t-distributed stochastic neighbor embedding (t-SNE) visualization of aligned single-cell transcriptional data generated from Prkar1a�/� or NTC tumors in
Rag1�/� mice. Cells of the t-SNE plot are labeled by the tumor dataset (top) and by unsupervised clusters that were identified (bottom).
(legend continued on next page)
Cell Systems 8, 136–151, February 27, 2019 145
expression of cluster-specific differentially expressed genes
(MAST analysis of each cluster relative to all other cells; adjusted
p < 1e�3, 95th-percentile of jlog-scale fold-change expressionj)to an ImmGen reference panel, comprising 85 cell types (Figures
6B and S5D). This transcriptional comparison shows moderate
correlation (Kendall correlation coefficient t > 0.3) between
clusters 1–2 andmyeloid cell types, including macrophage, den-
dritic cell (DC), andmonocyte cells, with cluster 1 also correlating
with granulocytic neutrophils (Figure 6B). In contrast, the clus-
ter 3 population moderately correlated (t > 0.3) with fibroblast
cells and weakly (t > 0.2) with endothelial and epithelial cells
(Figure 6B).
Further categorization suggested that cluster 1 is best
described as a population of tumor-associated macrophages
(TAMs), with significantly increased expression (adjusted p <
1e�8) of key TAM markers, including Cd68, Cd163, and Csf1r
(Cd115) (Bronte et al., 2016). In addition, our data suggest that
both immune cell clusters contain MDSCs with notable expres-
sion of multiple MDSC markers (Bronte et al., 2016), including
significantly increased expression of Arg1, Cybb (Nox2), and
Il4ra markers, relative to tumor cells of cluster 3 (Figures 6C
and S6). Once the clusters were characterized, we assessed
the transcriptional differences related to Prkar1a loss, as it af-
fects each cell cluster. These differential expression analyses
show that Prkar1a loss is significantly associated with the
increased expression of the inflammatory marker Saa3 in all
three clusters (adjusted p < 1e�5), indicating that the inflamma-
tory phenotype observed by RNA-seq is conferred to the TME.
Decreased expression of MHCII genes in both immune cell
clusters was also observed, including Cd74, H2-Aa, H2-Ab1,
and H2-Eb1 (adjusted p < 1e�9; Figure 6D). In addition, the
chemokine gene Cxcl13 is significantly increased in cluster 1
cells (adjusted p < 4.23e�22; Figure 6D). Together, these data
mapped the TME of Prkar1a mutant tumors in Rag1�/� mice
and revealed the cellular repertoire as well as transcriptional
changes in immune genes.
Next, we investigated whether the same immune infiltrates are
found within immunocompetent C57BL/6 tumors generated by
the loss of Prkar1a. Wild-type tumors were not assessed in the
immunocompetent background, as NTC tumors did not form in
C57BL/6; therefore, only Prkar1a mutant tumors can be profiled
in this host to serve as a refinement but not comparison. We
repeated the single-cell transcriptomics analysis of eGFP+
IMC9 and CD45+ immune cells of C57BL/6 tumors as a 1:10
mixed cell ratio, with the 10X Genomics platform for scRNA-
(B) Identification of cell cluster populations based on expression correlation to kno
differentially expressed genes (MAST analysis; FDR-adjusted p < 0.001) with
expression of these marker genes in each cluster was compared by Kendall c
innate immune and stromal cells). The scaled correlation coefficients for the clus
unsupervised hierarchical clustering, performed separately within 6 broad cell ty
(C) Classification of cell clusters by cell-type-specific characteristic gene express
identified as myeloid cell populations by the gene expression of canonical myelo
(Cd11b). These clusters also express known markers of tumor-associated mac
represents the IMC9 cancer cells based on the expression of the previously defin
(D) Volcano plots showing differentially expressed genes in the TME between the
Each panel shows a cluster-specific differential expression analysis, for which the
fold-changes are shown in the x axes. Differentially expressed genes were defined
knockout tumor model. Overexpressed and underexpressed genes are depicted
See also Figure S7.
146 Cell Systems 8, 136–151, February 27, 2019
seq library prep prior to sequencing. After pre-processing, we
applied the Potential of Heat-diffusion for Affinity-based Trajec-
tory Embedding (PHATE) technique (Figure S5C Left) (Moon
et al., 2017). The ensuing population structure of the data
showed a notable split into two discrete groups of cells, which
represented immune and tumor subsets, based on the expres-
sion of Ptprc (Cd45) and Mgp genes, respectively (Figure S5C
Right). As the expression of these genes was not detected in
all cells, we partitioned the population structure into two clusters
using unsupervised SNN clustering. Next, we investigated the
immune cell populations in finer detail by subsetting the Ptprc+
cluster and visualizing the subpopulation structure by PHATE
dimensionality reduction, followed by unsupervised clustering
(Figure S5D). We next sought to categorize the five immune sub-
populations by comparing the expression of cluster-specific
genes (STAR Methods) in each cluster to 107 known immune
cells of a reference panel comprising neutrophils, macrophages,
monocytes, DCs, natural killer cells, and T cells (Figure S5E). The
correlation heatmap shows a macrophage signature in all im-
mune subclusters, with the strongest macrophage correlations
to subclusters 1, 2, and 4 (t > 0.5). These three clusters also
have moderately high correlations to DCs and monocytes
(t > 0.4), whereas subcluster 3 correlates moderately high with
neutrophil and macrophage cell types (t > 0.4). In support of
this cell-type categorization, we show significantly increased
expression of TAM-specific genes Adgre1 (F4/80), Csf1r
(Cd115), and H2-Aa (MHC-II) in subcluster 4 (adjusted p <
1e�5; Figure S7). We also found significantly increased expres-
sion of various myeloid markers in the adjacent subclusters of
1 and 3, which might represent populations of MDSCs. We
also show significantly upregulated expression of MDSC regula-
tory genes (Ddit3 and S100a8) and M-MDSC enzymes (Arg1 and
Nos2) in subcluster 3 cells (Figure S7). Furthermore, the data
show increased trends in PMN-MDSC-related genes, Cybb
(Nox2) and Cd244 (Figure S7). We see the expression of these
immune cell type genes follow gradient-like expression patterns
across cells in the PHATE plot (Figure S7 Bottom). As the PHATE
dimensionality reduction technique has been shown to preserve
both local and global relationships between cells (Moon et al.,
2017), these gradient-like transcriptional trends may reflect a
spectrum of the phenotype in these different myeloid cell types,
which will be interesting to explore in future single-cell analyses.
These data further corroborated the refined TME of Prkar1a
mutant tumors in the fully immunocompetent C57BL/6 wild-
type host.
wn immune cell populations. Cluster-specific marker genes were identified as
the top 95% fold-change between clusters, and the log-normalized mean
orrelation analysis to ImmGen V1 transcriptional data (85 cell populations of
ter-specific markers are shown for each ImmGen cell population, arranged by
pe categories of the ImmGen data.
ion patterns, as shown by expression-labeled t-SNE plots. Clusters 1 and 2 are
id cell surface markers such as Cd14, Fcgr3 (Cd16), Fcgr2b (Cd32), and Itgam
rophages (TAMs) and myeloid-derived suppressor cells (MDSCs). Cluster 3
ed IMC9-specific gene markers including Mgp, Fkbp11, Cdkn2a, and Mrpl33.
myeloid-like immune cells in tumors induced by Prkar1a�/� and control cells.
FDR-adjusted p values are shown in the y axes and the log-scale expression
as those with an adjusted p value < 0.01 and a fold-change > 1.8 in the Prkar1a
by red and blue points, respectively.
CB
A
63.6%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%66666666666666666363336333333333333 666.66666666666666666666666666663666333333333333 63 63 633 63 63 666666666666666666666666666666666663333333 633 63 666666666666666666666666666666666666666663663636333633333 63 6666333.63.663 666666666666666666666666666666666636636333333333333333333.63 63 63 6666666666666666666666666363336333333.63 6.63.6663 63 663 63 66666666666666666633333 6663.633 66666666666666666666666666663633333333 63 6333 63 63 6333 66666666666666666666666666666633333333.666633 63.63 666666666666666666666666666666666666666666666666666666636336333 633 63 633 63666666666666666666666666666666636666333333333.63.6.63333333.36666666666666666666363333.33.3.33..3.3666666666666666636333333.3.3.33.666666666666663333.33.3366666666666633333333
0 104
105
0
-103
103
104
105
0 103
104
105
106
0
102
103
104
105
106
M-MDSCs
PMN-MDSCs
sgNTC TME
sgNTC
sgPrkar1a
DCs
TAMs
M-MDSCs
PMN-MDSCs
0
10
20
30
40
Immune cell populations
%ofCD45+Cells
**
**
0 104
105
PE/Cy7 CD45
APC/Cy7CD11b
FITC Gr-1
APCLy-6C
0
-103
103
104
105
0 103
104
105
106
0
102
103
104
105
106
M-MDSCs
PMN-MDSCs
sgPrkar1a TME
0 104
105
0
103
104
105
APC/Cy7 CD45.2
SSC-A
APC Ly6C
FITC
CD11b
PerCP/Cy5.5 MHC-II
SSC-A
PE F4/80
PE/Cy7CD24
59.4%
0 104
105
106
0
103
104
105
0 103
104
105
0
103
104
105
0 104
105
106
0
-103
103
104
105
106
DC
TAM
0 104
105
0
103
104
105
0 104
105
106
0
103
104
105
0 103
104
105
0
103
104
105
0 104
105
106
0
-103
103
104
105
106
DC
TAM
sgNTC TME
sgPrkar1a TME
APC/Cy7 CD45.2
SSC-A
APC Ly6C
FITC
CD11b
PerCP/Cy5.5 MHC-II
SSC-A
PE F4/80
PE/Cy7CD24
PE/Cy7 CD45
APC/Cy7CD11b
FITC Gr-1
APCLy-6C
Figure 7. Flow Cytometry Analysis of the
Immune Microenvironment in the Tumors
Induced by Prkar1aMutant and Control Can-
cer Cells
(A and B) Flow cytometry plots present the differ-
ences in the quantities of specific immune cell
subsets from tumors that were generated with or
without Prkar1a loss. (A) The gating schemes for the
detection of dendritic cells (DCs) and tumor-
associated macrophages (TAMs) are presented
using composite flow cytometry data from five
sgPrkar1a-generated tumors or six sgNTC-gener-
ated tumors in Rag1�/� mice. (B) Plots show the
flow cytometry gating scheme for the quantifi-
cation of monocytic and polymorphonuclear
myeloid-derived suppressor cells (M-MDSCs and
PMN-MDSCs, respectively) in tumors that were
generated with or without Prkar1a loss. The com-
posite flow cytometry data are shown for eight
sgPrkar1a-generated tumors or six sgNTC-gener-
ated tumors in Rag1�/� mice.
(C) Quantitative analysis of immune cell infiltration
differences between Prkar1a�/� and NTC tumors.
Immune cells were quantified from the flow
cytometry data in (A) and (B) and normalized to the
proportion of CD45/CD45.2+ cells for each tumor.
The immune cell populations were compared
between Prkar1a�/� and NTC tumors by unpaired
t test with Holm-Sidak correction for multiple-
testing. Significant population changes are de-
picted with asterisks (**, adjusted p < 1e�5).
Flow Cytometry Confirms the Alterations of TumorMicroenvironment in Prkar1a Mutant TumorsTo directly quantify Prkar1a knockout-mediated differences in
tumor-infiltrating immune cell populations, we performed flow cy-
tometry on eight tumors with Prkar1a loss and six tumors without.
First, we quantified DCs and TAMs (Figures 7A and 7C); however,
there are no significant changes between Prkar1a knockout and
wild-type tumors, yet there was a noteworthy decrease in TAMs
in Prkar1a knockout tumors (Figures 7A and 7C). Subsequently,
we inspected MDSC populations to show a significant increase
of PMN-MDSCs and a significant decrease of M-MDSCs in
Prkar1a knockout tumors (adjusted p = 4.99e�7 and 1.88e�6,
respectively; Figures 7B and 7C). Taken together, these data
show that the loss of Prkar1a leads to a polarized recruitment of
PMN-MDSCs, as well as a decrease in TAMs and the M-MDSCs
from which they are derived, representing a substantial alteration
of the host TME.
DISCUSSION
Malignant transformation, oncogenesis, and evading immune
destruction are all key hallmarks of cancer with fundamental roles
in thecancer immunecyclewithdirect consequenceson immuno-
Cell S
therapy outcomes (Chen and Mellman,
2013; Hanahan and Weinberg, 2011;
Sharma et al., 2017). Using multiple
whole-genomeknockout screens,we iden-
tified several candidate factors whose loss
of function consistently led to tumor
outgrowth in mouse hosts with different levels of immunocompe-
tence. Knocking out several hits enabled the cells to grow out in
Nu/Nu mice as tumors, while only Prkar1a knockout cells are
able to grow in fully immunocompetent C57BL/6mice. This is sur-
prising given that under the same genetic context, knockouts of
well-known tumor suppressors such as Tsc2 and Apc failed to
grow in the C57BL/6 host. Prkar1a encodes a regulatory subunit
of the PKA holoenzyme complex. Prkar1a is the only subunit of
the four whose knockout leads to pronounced disease-related
phenotypes. Prkar1a mutations were associated with disease in
the Carney complex (CNC), a disease syndrome characterized
by hyper-pigmented skin and sporadic neoplasias (Akin et al.,
2017; Alband and Korbonits, 2014; Stratakis, 2016). In this
context, hereditaryorgermlineLOFmutations inPrkar1aweresuf-
ficient to produce the disease in a dominant manner. In addition,
Prkar1a mutations are associated with several tumor types such
as cardiac, endocrine, thyroid, and lung adenocarcinoma (Kondo
et al., 2017; Wang et al., 2016) (Saloustros et al., 2017). Prkar1a
overexpression is also associatedwith tumor formation inmultiple
contexts, indicating that altering Prkar1a function leads to differ-
ential phenotypesacrosscellular and genetic backgrounds. Addi-
tionally, recent clinical findings show that patientswith a rare form
of liver cancer, known as fibrolamellar HCC, harbor the classical
ystems 8, 136–151, February 27, 2019 147
DNAJB1-PRKACA fusion, which was recapitulated in a mouse
model of cancer (Kastenhuber et al., 2017; Rondell et al., 2017).
This fusion protein results in hyperactive PKA signaling, similar
to Prkar1a knockouts. Supporting the similarity, a subset of
CNC patients also develop fibrolamellar HCC, indicating signifi-
cant overlap between the phenotypes (Rondell et al., 2017). Given
the increased likelihood with which patients with CNC develop
various types of tumors,Prkar1awas speculated to be a suppres-
sor for early oncogenesis, where the immune system plays a
fundamental role in clearing mutated cells with aberrant growth
to preserve tissue homeostasis (Brown et al., 2017).
While its role as a tumor suppressor is now supported, the
molecular changes associated with Prkar1a�/�-mediated
tumorigenesis are less clear. Reports have shown that Prkar1a
knockout increases PKA activity, as well as resistance to
apoptosis via mTOR activation (Robinson-White et al., 2006;
Pringle et al., 2014). Additionally, it has also been shown that
Prkar1a regulates the ERK, Snail and E-cadherin pathways
(Wang et al., 2016). We investigated the mechanistic mecha-
nisms of Prkar1a in cancer cells, showing that its genetic pertur-
bation leads to drastic alterations in the proteome, phosphopro-
teome, and the transcriptome. These changes center on acute
inflammatory responses, mediators of immune response, and
production of secreted ECM proteins and cytokines. For
example, IL-6 production is dramatically upregulated, potentially
enhancing Prkar1a�/� cells’ ability to drive inflammation in the
absence of other cell types, as well as to drive early malignant
transformation in both an autocrine and paracrine manner.
Notably, an miRNA overexpressed in various cancers, miRNA-
1246, targets PKA pathway genes leading to an inflammatory
phenotype (Bott et al., 2017), which further supports our findings.
Single-cell transcriptome profiling and comparative analysis
of Prkar1a mutant and wild-type tumor cells in Rag1�/� mice re-
vealed the TME alterations in this host. The phenotypic effect of
Prkar1a knockout is too strong to have comparable tumors in the
hosts with more potent immune systems (Nu/Nu and C57BL/6,
where Prkar1a wild-type cells cannot grow as tumors). The
downregulation of MHC-II in the immune cells of Prkar1a tumors
on Rag1�/� mice may suggest gene expression changes in the
TAM populations or recruitment of myeloid-derived suppressive
cells. Alternatively, the downregulation of antigen presentation
in resident macrophages is in line with the reduced MHCII
expression. Furthermore, characterization of the immune cell
infiltrates by flow cytometry in Rag1�/� mice showed a signifi-
cant increase of PMN-MDSCs relative to controls. PMN-MDSCs
are antigen-specific immunosuppressive cells, known to down-
regulate cytotoxic T cell activity resulting in tumor-specific
T cell tolerance (Kumar et al., 2016). While additional studies
will need to be performed to understand the significance of
this finding, it raises the possibility that the PMN-MDSCs could
play an outsized role in regulating the growth of Prkar1a�/�
tumors in hosts with intact adaptive immune system. Future
molecular characterization of Prkar1a may further unveil the
detailed molecular, cellular, and immunological mechanisms
underlying TME changes driven by this gene or other tumor-
intrinsic factors. In summary, by convergent genome-scale
screens and multiple sets of functional interrogation, we have
identified Prkar1a as a gene regulating a tumor-intrinsic genetic
program for remodeling the TME.
148 Cell Systems 8, 136–151, February 27, 2019
STAR+METHODS
Detailed methods are provided in the online version of this paper
and include the following:
d KEY RESOURCES TABLE
d CONTACT FOR REAGENT AND RESOURCE SHARING
d EXPERIMENTAL MODEL AND SUBJECT DETAILS
B Institutional Approval
d METHOD DETAILS
B Cell Culture
B Mice
B Virus Production of CRISPR Library
B Cellular CRISPR Library Production
B Cell Library Transplantation
B Mouse Organ Dissection, Imaging, and Histology
B Mouse Tissue Collection for Molecular Biology
B Genomic DNA Extraction
B Demultiplexing and Read Preprocessing
B Mapping of sgRNA Spacers and Quantitation of
sgRNAs
B Normalization and Summary-Level Analysis of sgRNA
Abundances
B Determining Significantly Enriched sgRNAs (Custom 3-
Way Hit Caller)
B Individual Gene Validation in Nu/Nu and
C57BL/6J Mice
B SILAC Proteomics Sample Preparation
B Proteomics Analysis
B Phosphorylation Site Motif Analysis
B Western Blot
B RNA-seq Differential Expression Analysis
B Pathway Enrichment Analysis
B ELISA
B TIL Isolation for scRNA-seq
B Pre-processing the Single-Cell RNA-seq Data
B Filtering and Visualization of scRNA-seq Data
B Cluster Analysis of scRNA-seq Data
B Flow Cytometry
B Blinding Statement
d QUANTIFICATION AND STATISTICAL ANALYSIS
d DATA AND SOFTWARE AVAILABILITY
B Reagent Availability
B Code Availability
B Data Availability and Accessions
SUPPLEMENTAL INFORMATION
Supplemental Information includes 7 figures, 14 tables, and14 datasets andcan
be found with this article online at https://doi.org/10.1016/j.cels.2019.01.004.
ACKNOWLEDGMENTS
We thank S. Zhang, M. Martinez, Y. Cheng, L. Guo, B. Fitzgerald, G. Gabel, A.
Ristic, A. Araki, G.L. Wang, C. Castaldi, D. Rimm, and A. Brooks for helping in
parts of the study design, experiments, or analyses. We thank all members in
Chen laboratory, as well as various colleagues in the Department of Genetics,
Systems Biology Institute, Cancer Systems Biology Center, MCGD Program,
Immunobiology Program, BBS Program, Cancer Center, and StemCell Center
at Yale for assistance and/or discussion. We thank the Center for Genome
Analysis, Center for Molecular Discovery, Pathology Tissue Services,
Histology Services, High Performance Computing Center, West Campus
Analytical Chemistry Core and West Campus Imaging Core, and Keck
Biotechnology Resource Laboratory at Yale for technical support.
S.C. is supported by Yale SBI/Genetics Startup Fund, NIH/NCI
(1DP2CA238295-01, 1R01CA231112-01, 1U54CA209992-8697, 5P50CA196530-
A10805, and 4P50CA121974-A08306), DamonRunyonDale Frey Award (DFS-
13-15), Melanoma Research Alliance (412806 and 16-003524), St-Baldrick’s
Foundation (426685), Breast Cancer Alliance, Cancer Research Institute
(CLIP), AACR (499395 and 17-20-01-CHEN), The Mary Kay Foundation
(017-81), The V Foundation (V2017-022), Ludwig Family Foundation, DoD
(W81XWH-17-1-0235), Sontag Foundation, and Chenevert Family Foundation.
R.S.H. is supported by Yale SPORE in lung cancer (5P50CA196530). R.F. is
supported by NIH/NCI (1U54CA209992). A. Codina and P.A.R. are supported
by Yale Ph.D. training grant from NIH (T32GM007499). R.D.C., J.J.P., and
M.B.D. are supported by the Yale MSTP training grant from NIH
(T32GM007205). G.W. is supported by Cancer Research Institute Irvington
and RJ Anderson Postdoctoral Fellowships. H.Y. and K.Z. are supported by
USTC student scholarships.
AUTHOR CONTRIBUTIONS
A. Codina designed the experiments with S.C. and initiated the study. A. Co-
dina., G.W., and P.A.R. performed most experiments. P.A.R. and R.D.C. per-
formed most data analyses. J.J.P. and A. Chang assisted data analysis. H.Y.,
K.Z., M.D., L.Y., Y.E., L.S. and B.G. assisted experiments. D.J., R.S.H.,
M.B.D., J.R., and R.F. provided conceptual advice. S.C. conceived the study,
secured funding, and supervised the work. A. Codina., P.A.R., R.D.C., and
S.C. prepared the manuscript.
DECLARATION OF INTERESTS
The authors declare no competing interests.
Received: May 24, 2018
Revised: November 3, 2018
Accepted: January 22, 2019
Published: February 20, 2019
REFERENCES
Akbay, E.A., Koyama, S., Carretero, J., Altabef, A., Tchaicha, J.H.,
Christensen, C.L., Mikse, O.R., Cherniack, A.D., Beauchamp, E.M., Pugh,
T.J., et al. (2013). Activation of the PD-1 pathway contributes to immune
escape in EGFR-driven lung tumors. Cancer Discov. 3, 1355–1363.
Akin, S., Noyan, S., Dagdelen, S., Pasaoglu, I., Kaynaroglu, V., Askun, M.M.,
Bilen, C.Y., Kiratli, H., Baydar, D.E., Onder, S., et al. (2017). Unusual presenta-
tions of Carney complex in patient with a novel PRKAR1A mutation. Neuro
Endocrinol. Lett. 38, 248–254.
Alband, N., and Korbonits, M. (2014). Familial pituitary tumors. Handb. Clin.
Neurol. 124, 339–360.
Bailey, T.L., Boden, M., Buske, F.A., Frith, M., Grant, C.E., Clementi, L., Ren,
J., Li, W.W., and Noble, W.S. (2009). Meme SUITE: tools for motif discovery
and searching. Nucleic Acids Res. 37, W202–W208.
Bayona-Bafaluy, M.P., Acın-Perez, R., Mullikin, J.C., Park, J.S., Moreno-
Loshuertos, R., Hu, P., Perez-Martos, A., Fernandez-Silva, P., Bai, Y., and
Enrıquez, J.A. (2003). Revisiting the mouse mitochondrial DNA sequence.
Nucleic Acids Res. 31, 5349–5355.
Bott, A., Erdem, N., Lerrer, S., Hotz-Wagenblatt, A., Breunig, C., Abnaof, K.,
Worner, A., Wilhelm, H., M€unstermann, E., Ben-Baruch, A., and Wiemann, S.
(2017). miRNA-1246 induces pro-inflammatory responses in mesenchymal
stem/stromal cells by regulating PKA and PP2A. Oncotarget. 8, 43897–43914.
Bray, N.L., Pimentel, H., Melsted, P., and Pachter, L. (2016). Near-optimal
probabilistic RNA-seq quantification. Nat. Biotechnol. 34, 525–527.
Bronte, V., Brandau, S., Chen, S.H., Colombo, M.P., Frey, A.B., Greten, T.F.,
Mandruzzato, S., Murray, P.J., Ochoa, A., Ostrand-Rosenberg, S., et al.
(2016). Recommendations for myeloid-derived suppressor cell nomenclature
and characterization standards. Nat. Commun. 7, 12150.
Brown, S., Pineda, C.M., Xin, T., Boucher, J., Suozzi, K.C., Park, S., Matte-
Martone, C., Gonzalez, D.G., Rytlewski, J., Beronja, S., et al. (2017). Correction
of aberrant growth preserves tissue homeostasis. Nature 548, 334–337.
Brunet, J.P., Tamayo, P., Golub, T.R., and Mesirov, J.P. (2004). Metagenes
and molecular pattern discovery using matrix factorization. Proc. Natl. Acad.
Sci. USA 101, 4164–4169.
Butler, A., Hoffman, P., Smibert, P., Papalexi, E., and Satija, R. (2018).
Integrating single-cell transcriptomic data across different conditions, technol-
ogies, and species. Nat. Biotechnol. 36, 411–420.
Casey, S.C., Tong, L., Li, Y., Do, R., Walz, S., Fitzgerald, K.N., Gouw, A.M.,
Baylot, V., G€utgemann, I., Eilers, M., et al. (2016). MYC regulates the antitumor
immune response through CD47 and PD-L1. Science 352, 227–231.
Chen, D.S., and Mellman, I. (2013). Oncology meets immunology: the cancer-
immunity cycle. Immunity 39, 1–10.
Chen, D.S., and Mellman, I. (2017). Elements of cancer immunity and the
cancer-immune set point. Nature 541, 321–330.
Chen, S., Sanjana, N.E., Zheng, K., Shalem, O., Lee, K., Shi, X., Scott, D.A.,
Song, J., Pan, J.Q.,Weissleder, R., et al. (2015). Genome-wide CRISPR screen
in a mouse model of tumor growth and metastasis. Cell 160, 1246–1260.
Church, D.M., Goodstadt, L., Hillier, L.W., Zody, M.C., Goldstein, S., She, X.,
Bult, C.J., Agarwala, R., Cherry, J.L., DiCuccio, M., et al. (2009). Lineage-
specific biology revealed by a finished genome assembly of the mouse.
PLoS Biol. 7, e1000112.
Coussens, L.M., and Werb, Z. (2002). Inflammation and cancer. Nature 420,
860–867.
Dobin, A., Davis, C.A., Schlesinger, F., Drenkow, J., Zaleski, C., Jha, S., Batut,
P., Chaisson,M., andGingeras, T.R. (2013). STAR: ultrafast universal RNA-seq
aligner. Bioinformatics 29, 15–21.
Doench, J.G., Fusi, N., Sullender, M., Hegde, M., Vaimberg, E.W., Donovan,
K.F., Smith, I., Tothova, Z., Wilen, C., Orchard, R., et al. (2016). Optimized
sgRNA design to maximize activity and minimize off-target effects of
CRISPR-Cas9. Nat. Biotechnol. 34, 184–191.
Doench, J.G., Hartenian, E., Graham, D.B., Tothova, Z., Hegde, M., Smith, I.,
Sullender, M., Ebert, B.L., Xavier, R.J., and Root, D.E. (2014). Rational design
of highly active sgRNAs for CRISPR-Cas9-mediated gene inactivation. Nat
Biotechnol. 32, 1262–1267.
Dorand, R.D., Nthale, J., Myers, J.T., Barkauskas, D.S., Avril, S., Chirieleison,
S.M., Pareek, T.K., Abbott, D.W., Stearns, D.S., Letterio, J.J., et al. (2016).
Cdk5 disruption attenuates tumor PD-L1 expression and promotes antitumor
immunity. Science 353, 399–403.
Dunn, G.P., Bruce, A.T., Ikeda, H., Old, L.J., and Schreiber, R.D. (2002).
Cancer immunoediting: from immunosurveillance to tumor escape. Nat.
Immunol. 3, 991–998.
Finak, G., McDavid, A., Yajima, M., Deng, J., Gersuk, V., Shalek, A.K., Slichter,
C.K., Miller, H.W., McElrath, M.J., Prlic, M., et al. (2015). MAST: a flexible sta-
tistical framework for assessing transcriptional changes and characterizing
heterogeneity in single-cell RNA sequencing data. Genome Biol. 16, 278.
Gao, J., Shi, L.Z., Zhao, H., Chen, J., Xiong, L., He, Q., Chen, T., Roszik, J.,
Bernatchez, C., Woodman, S.E., et al. (2016). Loss of IFN-gamma pathway
genes in tumor cells as a mechanism of resistance to anti-CTLA-4 therapy.
Cell 167, 397–404.e9.
Gaujoux, R., and Seoighe, C. (2010). A flexible R package for nonnegative ma-
trix factorization. BMC Bioinformatics 11, 367.
Gierahn, T.M., Wadsworth, M.H., 2nd, Hughes, T.K., Bryson, B.D., Butler, A.,
Satija, R., Fortune, S., Love, J.C., and Shalek, A.K. (2017). Seq-Well: portable,
low-cost RNA sequencing of single cells at high throughput. Nat. Methods 14,
395–398.
Gupta, S., Stamatoyannopoulos, J.A., Bailey, T.L., and Noble, W.S. (2007).
Quantifying similarity between motifs. Genome Biol. 8, R24.
Hanahan, D., and Weinberg, R.A. (2011). Hallmarks of cancer: the next gener-
ation. Cell 144, 646–674.
Heng, T.S., and Painter, M.W.; Immunological Genome Project (2008). The
Immunological Genome Project: networks of gene expression in immune cells.
Nat. Immunol. 9, 1091–1094.
Cell Systems 8, 136–151, February 27, 2019 149
Heninger, E., Krueger, T.E., and Lang, J.M. (2015). Augmenting antitumor im-
mune responses with epigenetic modifying agents. Front. Immunol. 6, 29.
Hu, J., Rho, H.S., Newman, R.H., Zhang, J., Zhu, H., and Qian, J. (2014a).
PhosphoNetworks: a database for human phosphorylation networks.
Bioinformatics 30, 141–142.
Hu, J.F., Rho, H.S., Newman, R.H., Hwang, W.C., Neiswinger, J., Zhu, H.,
Zhang, J., and Qian, J. (2014b). Global analysis of phosphorylation networks
in humans. Biochim. Biophys. Acta 1844, 224–231.
Huang, da W., Sherman, B.T., and Lempicki, R.A. (2009a). Bioinformatics
enrichment tools: paths toward the comprehensive functional analysis of large
gene lists. Nucleic Acids Res. 37, 1–13.
Huang, da W., Sherman, B.T., and Lempicki, R.A. (2009b). Systematic and
integrative analysis of large gene lists using David bioinformatics resources.
Nat. Protoc. 4, 44–57.
Kastenhuber, E.R., Lalazar, G., Houlihan, S.L., Tschaharganeh, D.F., Baslan,
T., Chen, C.C., Requena, D., Tian, S., Bosbach, B., Wilkinson, J.E., et al.
(2017). DNAJB1-PRKACA fusion kinase interacts with beta-catenin and the
liver regenerative response to drive fibrolamellar hepatocellular carcinoma.
Proc. Natl. Acad. Sci. USA 114, 13076–13084.
Kataoka, K., Shiraishi, Y., Takeda, Y., Sakata, S., Matsumoto, M., Nagano, S.,
Maeda, T., Nagata, Y., Kitanaka, A., Mizuno, S., et al. (2016). Aberrant PD-L1
expression through 3’-UTR disruption inmultiple cancers. Nature 534, 402–406.
Kim, R., Emi, M., and Tanabe, K. (2007). Cancer immunoediting from immune
surveillance to immune escape. Immunology 121, 1–14.
Kolde, R. 2019. Package ‘pheatmap’. https://cran.r-project.org/web/packages/
pheatmap/pheatmap.pdf.
Kondo, K., Harada, M., Konomoto, T., Hatanaka, M., and Nunoi, H. (2017).
Novel PRKAR1A mutation in Carney complex with cardiac myxoma. Pediatr.
Int. 59, 840–841.
Kumar, M.S., Pester, R.E., Chen, C.Y., Lane, K., Chin, C., Lu, J., Kirsch, D.G.,
Golub, T.R., and Jacks, T. (2009). Dicer1 functions as a haploinsufficient tumor
suppressor. Genes Dev. 23, 2700–2704.
Kumar, V., Patel, S., Tcyganov, E., and Gabrilovich, D.I. (2016). The nature of
myeloid-derived suppressor cells in the tumor microenvironment. Trends
Immunol. 37, 208–220.
Langmead, B., Trapnell, C., Pop, M., and Salzberg, S.L. (2009). Ultrafast and
memory-efficient alignment of short DNA sequences to the human genome.
Genome Biol. 10, R25.
Lastwika, K.J., Wilson, W., 3rd, Li, Q.K., Norris, J., Xu, H., Ghazarian, S.R.,
Kitagawa, H., Kawabata, S., Taube, J.M., Yao, S., et al. (2016). Control of
PD-L1 expression by oncogenic activation of the AKT-mTOR pathway in
non-small cell lung cancer. Cancer Res. 76, 227–238.
Li, W., Xu, H., Xiao, T., Cong, L., Love, M.I., Zhang, F., Irizarry, R.A., Liu, J.S.,
Brown, M., and Liu, X.S. (2014). MAGeCK enables robust identification of
essential genes from genome-scale CRISPR/Cas9 knockout screens.
Genome Biol. 15, 554.
Liu, C., Peng, W., Xu, C., Lou, Y., Zhang, M., Wargo, J.A., Chen, J.Q., Li, H.S.,
Watowich, S.S., Yang, Y., et al. (2013). BRAF inhibition increases tumor infiltra-
tion by T cells and enhances the antitumor activity of adoptive immunotherapy
in mice. Clin. Cancer Res. 19, 393–403.
Lun, A.T., McCarthy, D.J., and Marioni, J.C. (2016). A step-by-step workflow
for low-level analysis of single-cell RNA-seq data with Bioconductor.
Version 2. F1000Res, 2122.
Luo, B., Cheung, H.W., Subramanian, A., Sharifnia, T., Okamoto, M., Yang, X.,
Hinkle, G., Boehm, J.S., Beroukhim, R., Weir, B.A., et al. (2008). Highly parallel
identification of essential genes in cancer cells. Proc Natl Acad Sci U S A. 105,
20380–20385.
Macosko, E.Z., Basu, A., Satija, R., Nemesh, J., Shekhar, K., Goldman, M.,
Tirosh, I., Bialas, A.R., Kamitaki, N., Martersteck, E.M., et al. (2015). Highly par-
allel genome-wide expression profiling of individual cells using nanoliter drop-
lets. Cell 161, 1202–1214.
Manguso, R.T., Pope, H.W., Zimmer, M.D., Brown, F.D., Yates, K.B., Miller,
B.C., Collins, N.B., Bi, K., LaFleur, M.W., Juneja, V.R., et al. (2017). In vivo
150 Cell Systems 8, 136–151, February 27, 2019
CRISPR screening identifies Ptpn2 as a cancer immunotherapy target.
Nature 547, 413–418.
Martin, M. (2011). Cutadapt removes adapter sequences From high-
throughput sequencing reads. EMBnet J. 17, 10–12.
Mi, H., Muruganujan, A., Casagrande, J.T., and Thomas, P.D. (2013). Large-
scale gene function analysis with the PANTHER classification system. Nat
Protoc. 8, 1551–1566.
Moon, K.R., van Dijk, D., Wang, Z., Chen, W., Hirn, M.J., Coifman, R.R.,
Ivanova, N.B., Wolf, G., and Krishnaswamy, S. (2017). PHATE: A dimension-
ality reduction method for visualizing trajectory structures in high-dimensional
biological data. bioRxiv. https://doi.org/10.1101/120378v3.
Newman, R.H., Hu, J.F., Rho, H.S., Xie, Z., Woodard, C., Neiswinger, J.,
Cooper, C., Shirley, M., Clark, H.M., Hu, S.H., et al. (2013). Construction of hu-
man activity-based phosphorylation networks. Mol. Syst. Biol. 9, 655.
Pan, D., Kobayashi, A., Jiang, P., Ferrari de Andrade, L., Tay, R.E., Luoma,
A.M., Tsoucas, D., Qiu, X., Lim, K., Rao, P., et al. (2018). A major chromatin
regulator determines resistance of tumor cells to T cell-mediated killing.
Science 359, 770–775.
Parsa, A.T., Waldron, J.S., Panner, A., Crane, C.A., Parney, I.F., Barry, J.J.,
Cachola, K.E., Murray, J.C., Tihan, T., Jensen,M.C., et al. (2007). Loss of tumor
suppressor PTEN function increases B7-H1 expression and immunoresist-
ance in glioma. Nat. Med. 13, 84–88.
Patel, S.J., Sanjana, N.E., Kishton, R.J., Eidizadeh, A., Vodnala, S.K., Cam, M.,
Gartner, J.J., Jia, L., Steinberg, S.M., Yamamoto, T.N., et al. (2017). Identification
of essential genes for cancer immunotherapy. Nature 548, 537–542.
Pimentel, H., Bray, N.L., Puente, S., Melsted, P., and Pachter, L. (2017).
Differential analysis of RNA-seq incorporating quantification uncertainty.
Nat. Methods 14, 687–690.
Pringle, D.R., Vasko, V.V., Yu, L., Manchanda, P.K., Lee, A.A., Zhang, X.,
Kirschner, J.M., Parlow, A.F., Saji, M., Jarjoura, D., et al. (2014). Follicular thy-
roid cancers demonstrate dual activation of PKA and mTOR as modeled by
thyroid-specific deletion of Prkar1a and Pten in mice. J Clin Endocrinol
Metab. 99, E804–E812.
Robinson-White, A.J., Leitner, W.W., Aleem, E., Kaldis, P., Bossis, I., and
Stratakis, C.A. (2006). PRKAR1A inactivation leads to increased proliferation
and decreased apoptosis in human B lymphocytes. Cancer Res. 66,
10603–10612.
Robinson, M.D., McCarthy, D.J., and Smyth, G.K. (2010). edgeR: a
Bioconductor package for differential expression analysis of digital gene
expression data. Bioinformatics 26, 139–140.
Rondell, P.G., Lackner, K., Terracciano, L., Gonzalez-Cantu, Y., Maleszewski,
J.J., Greipp, P.T., Simon, S.M., and Torbenson, M.S. (2017). Fibrolamellar car-
cinoma in the Carney complex: PRKAR1A loss instead of the classic DNAJB1-
PRKACA fusion. Hepatology 68, 1441–1447.
Saloustros, E., Salpea, P., Starost, M., Liu, S., Faucz, F.R., London, E., Szarek,
E., Song, W.J., Hussain, M., and Stratakis, C.A. (2017). Prkar1a gene knockout
in the pancreas leads to neuroendocrine tumorigenesis. Endocr. Relat. Cancer
24, 31–40.
Sanjana, N.E., Shalem, O., and Zhang, F. (2014). Improved vectors and
genome-wide libraries for CRISPR screening. Nat. Methods 11, 783–784.
Satija, R., Farrell, J.A., Gennert, D., Schier, A.F., and Regev, A. (2015). Spatial
reconstruction of single-cell gene expression data. Nat. Biotechnol. 33,
495–502.
Schwartz, D., and Gygi, S.P. (2005). An iterative statistical approach to the
identification of protein phosphorylation motifs from large-scale data sets.
Nat. Biotechnol. 23, 1391–1398.
Shalem, O., Sanjana, N.E., Hartenian, E., Shi, X., Scott, D.A., Mikkelson, T.S.,
Heckl, D., Ebert, B.L., Root, D.E., Doench, J.G., et al. (2014). Genome-scale
CRISPR-Cas9 knockout screening in human cells. Science 343, 84–87.
Sharma, P., Hu-Lieskovan, S., Wargo, J.A., and Ribas, A. (2017). Primary, adap-
tive, and acquired resistance to cancer immunotherapy. Cell 168, 707–723.
Spranger, S., Bao, R., and Gajewski, T.F. (2015). Melanoma-intrinsic beta-
catenin signalling prevents anti-tumour immunity. Nature 523, 231–235.
Stratakis, C.A. (2016). Carney complex: A familial lentiginosis predisposing to a
variety of tumors. Rev. Endocr. Metab. Disord. 17, 367–371.
Subramanian, A., Tamayo, P., Mootha, V.K., Mukherjee, S., Ebert, B.L.,
Gillette, M.A., Paulovich, A., Pomeroy, S.L., Golub, T.R., Lander, E.S., et al.
(2005). Gene set enrichment analysis: a knowledge-based approach for inter-
preting genome-wide expression profiles. Proc. Natl. Acad. Sci. USA 102,
15545–15550.
Tyanova, S., Temu, T., and Cox, J. (2016). The MaxQuant computational plat-
form for mass spectrometry-based shotgun proteomics. Nat. Protoc. 11,
2301–2319.
van der Maaten, L.J.P. (2014). Accelerating t-SNE using Tree-Based
Algorithms. J. Mach. Learn. Res. 15, 3221–3245.
van der Maaten, L.J.P., and Hinton, G.E. (2008). Visualizing High-Dimensional
Data Using t-SNE. J. Mach. Learn. Res. 9, 2579–2605.
Waltman, L., and van Eck, N.J. (2013). A smart local moving algorithm for
large-scale modularity-based community detection. Eur. Phys. J. B 86.
Wang, S., Cheng, Y., Zheng, Y., He, Z., Chen, W., Zhou, W., Duan, C., and
Zhang, C. (2016). PRKAR1A is a functional tumor suppressor inhibiting
ERK/Snail/E-cadherin pathway in lung adenocarcinoma. Sci. Rep.
6, 39630.
Zaretsky, J.M., Garcia-Diaz, A., Shin, D.S., Escuin-Ordinas, H., Hugo, W., Hu-
Lieskovan, S., Torrejon, D.Y., Abril-Rodriguez, G., Sandoval, S., Barthly, L.,
et al. (2016). Mutations associated with acquired resistance to PD-1 blockade
in melanoma. N. Engl. J. Med. 375, 819–829.
Zender, L., Spector, M.S., Xue, W., Flemming, P., Cordon-Cardo, C., Silke, J.,
Fan, S.T., Luk, J.M., Wigler, M., Hannon, G.J., et al. (2006). Identification and
validation of oncogenes in liver cancer using an integrative oncogenomic
approach. Cell 125, 1253–1267.
Cell Systems 8, 136–151, February 27, 2019 151
STAR+METHODS
KEY RESOURCES TABLE
REAGENT or RESOURCE SOURCE IDENTIFIER
Antibodies
Phospho-Stat3 (Tyr705) (D3A7) XP Cell Signaling Technologies RRID:AB_2491009
Stat3a (D1A5) XP Rabbit mAb #8768 Cell Signaling Technologies RRID:AB_2722529
GAPDH antibody Santa Cruz Biotech RRID:AB_627679
Ultra-LEAF Purified anti-mouse CD3 3Antibody
(clone: 145-2C11)
Biolegend RRID:AB_11149115
APC anti-mouse CD3 3Antibody (clone: 145-2C11) Biolegend RRID:AB_312677
FITC anti-mouse Ly-6G/Ly-6C (Gr-1) Antibody
(clone: RB6-8C5)
Biolegend RRID:AB_313370
FITC anti-mouse/human CD11b Antibody (clone: M1/70) Biolegend RRID:AB_312789
PE/Dazzle594 anti-mouse CD11c Antibody (clone: N418) Biolegend RRID:AB_2563655
PerCP/Cy5.5 anti-mouse I-A/I-E Antibody (clone: m5/114.15.2) Biolegend RRID:AB_2191071
APC anti-mouse Ly-6C Antibody (clone: HK1.4) Biolegend RRID:AB_1732076
PE anti-mouse F4/80 Antibody (clone: BM8) Biolegend RRID:AB_893486
APC/Cy7 anti-mouse CD45.2 Antibody (clone: 104) Biolegend RRID:AB_830789
APC/Cy7 anti-mouse/human CD11b Antibody (clone: M1/70) Biolegend RRID:AB_830641
PE/Cy7 anti-mouse CD24 Antibody (clone: M1/69) Biolegend RRID:AB_756048
PE/Cy7 anti-mouse CD45 Antibody (clone: 30-F11) Biolegend RRID:AB_312978
Bacterial and Virus Strains
One Shot Stbl3 Chemical Competent E. Coli ThermoFisher Catalog Number: C737303
Endura ElectroCompetent Cells Lucigen Catalog Number: 60242-2
Critical Commercial Assays
Qubit dsDNA HS Assay Kit ThermoFisher Catalog Number: Q32854
Chromium Single Cell 10x Genomics N/A
Gibson Assembly Master Mix NEB Catalog Number: E2611
Collagenase, Type IV, powder ThermoFisher Catalog Number: 17104019
QuickExtract DNA Extraction Solution Epicenter Catalog Number: QE09050
Proteinase K Sigma-Aldrich Catalog Number: 3115879001
RNase A Qiagen Catalog Number: 19101
Phusion Flash High-Fidelity PCR Master Mix ThermoFisher Catalog Number: F548L
DreamTaq Green PCR Master Mix (2X) ThermoFisher Catalog Number: K1082
E-Gel Low Range Quantitative DNA Ladder ThermoFisher Catalog Number: 12373031
T7E1 NEB Catalog Number: M0302L
TRIzol Plus RNA Purification Kit ThermoFisher Catalog Number: 12183555
QIAquick Gel Extraction Kit Qiagen Catalog Number: Q28706
Deposited Data
UniProt C57BL6J UP000000589 proteome
(last modified: May 16, 2017)
Bayona-Bafaluy et al.,
(2003); Church et al., (2009)
http://www.uniprot.org/proteomes/
UP000000589
PhosphoNetworks kinase phosphorylation motif data Hu et al., (2014a), (2014b);
Newman et al., (2013)
http://www.phosphonetworks.org/
download/motifMatrix.csv
Immunological Genome Project expression data (version 1) Heng et al., (2008) GEO: GSE15907
Mouse genome (mm10) Macosko et al., (2015) GEO: GSE63472
Prkar1a mutant mRNA-seq This study GEO: PRJNA506858
Prkar1a mutant proteomics This study PRIDE: PXD011837
Prkar1a mutant single-cell RNA-seq This study GEO: PRJNA509929
(Continued on next page)
e1 Cell Systems 8, 136–151.e1–e7, February 27, 2019
Continued
REAGENT or RESOURCE SOURCE IDENTIFIER
Experimental Models: Cell Lines
HEK293FT ThermoFisher Catalog Number: R70007
Myc;p53-/- hepatocyte Zender et al. (2006) Clone IM
Myc;p53-/- hepatocyte Zender et al. (2006) Clone 172
Experimental Models: Organisms/Strains
C57BL/6J Jackson Laboratory RRID:IMSR_JAX:000664
Nu/Nu Jackson Laboratory RRID:IMSR_JAX:002019
Rag1-/- Jackson Laboratory RRID:IMSR_JAX:002216
Recombinant DNA
mGeCKOa Sanjana et al., (2014) https://www.addgene.org/pooled-library/
zhang-mouse-gecko-v2/
mBrie Doench et al. (2014) https://www.addgene.org/pooled-library/
broadgpp-mouse-knockout-brie/
Software and Algorithms
FlowJo software 9.9.6 FlowJo, LLC
Cutadapt 1.12 Martin, (2011) https://github.com/marcelm/cutadapt/
Bowtie 1.1.2 Langmead et al., (2009) https://sourceforge.net/projects/bowtie-bio/
files/bowtie/1.1.2
GSEA Subramanian et al., (2005) http://software.broadinstitute.org/gsea/
DAVID Huang et al., (2009b) https://david.ncifcrf.gov/
Scran R package Lun et al., (2016) https://github.com/MarioniLab/scran/
Rtsne R package van der Maaten, (2014);
van der Maaten
and Hinton, (2008)
https://github.com/jkrijthe/Rtsne/
Edge R package Robinson et al., (2010) https://github.com/StoreyLab/edge/
Panther classification system Mi et al., (2013) http://pantherdb.org/
RIGER Luo et al. (2008) https://software.broadinstitute.org/GENE-E/
extensions/RIGER.jar
MAGeCK Li et al. (2014) http://mageck.sourceforge.net
Dropseq_tools-1.12 Macosko et al., (2015) http://mccarrolllab.com/dropseq/
Seurat R package Satija et al., (2015) http://satijalab.org/seurat/
STAR 2.5.3a Dobin et al., (2013) https://github.com/alexdobin/STAR/
R: The R Project for Statistical Computing The R Foundation https://www.r-project.org/
MaxQuant version 1.6.0.1 Tyanova et al., (2016) https://www.biochem.mpg.de/5111795/
maxquant
MEME software suite version 4.12.0 Bailey et al., (2009) http://meme-suite.org/
Tomtom motif comparison Gupta et al., (2007) http://meme-suite.org/
Kallisto Bray et al., 2016 https://pachterlab.github.io/kallisto/
Sleuth Pimentel et al., (2017) http://pachterlab.github.io/sleuth/
Picard 2.9.0 Broad Institute, 2018 http://broadinstitute.github.io/picard/
Modification Motifs (MoMo) Cheng, and Mellman,
(2017) (BioRx)
http://meme-suite.org/
Ingenuity Pathway Analysis Qiagen https://analysis.ingenuity.com
phateR R package Moon et al., (2017) (BioRx) https://github.com/KrishnaswamyLab/phateR
MAST R package Finak et al., 2015 https://github.com/RGLab/MAST/
pheatmap R package Kolde, (2019) https://github.com/raivokolde/pheatmap
CONTACT FOR REAGENT AND RESOURCE SHARING
All resources related to this work are available upon reasonable request to the lead contact, Sidi Chen ([email protected]).
Cell Systems 8, 136–151.e1–e7, February 27, 2019 e2
EXPERIMENTAL MODEL AND SUBJECT DETAILS
Institutional ApprovalAll experimental work involving recombinant DNA was performed under the guidelines of Yale University Environmental Health and
Safety (EHS) committee with an approved protocol (Chen-rDNA-15-45). All animal work was performed under the guidelines of Yale
University Institutional Animal Care and Use Committee (IACUC) with approved protocols (Chen-2015-20068; Chen-2018-20068)
and was consistent with the Guide for Care and Use of Laboratory Animals, National Research Council, 1996 (Institutional Animal
Welfare Assurance No. A-3125-01).
METHOD DETAILS
Cell CultureImmortalized hepatocyte cell lines with the background of p53-/-;cMyc (IM and 172) were from (Zender et al., 2006). Lung cancer cells
were from (Kumar et al., 2009). Cells were transduced with lenti-Cas9 or lenti-Cas9-GFP to generate Cas9-positive cells. All cell lines
were grown under standard conditions using DMEM containing 5% FBS & 1% Penstrep in a CO2 incubator unless otherwise noted.
Cell culture for SILAC experiment was described below.
MiceSeveral mouse strains with various levels of immunocompetence were used as transplant hosts, including C57BL/6, Nu/Nu and
Rag1-/-. Mice, both sexes, between the ages of 6-12 weeks of age were used for the study. All animals were housed in standard
individually ventilated, pathogen-free conditions, with 12h:12h or 13h:11h light cycle, room temperature (21-23�C) and 40-60% rela-
tive humidity. When a cohort of animals was receiving multiple treatments, animals were randomized by 1) randomly assign
animals to different groups using littermates, 2) random mixing of females prior to treatment, maximizing the evenness or represen-
tation of mice from different cages in each group, and/or 3) random assignment of mice to each group, in order to minimize the
effect of gender, litter, small difference in age, cage, housing position, where applicable. All mouse experiments follow ARRIVE
guidelines.
Virus Production of CRISPR LibraryFirst-generation and second-generation mouse genome-scale CRISPR sgRNA libraries, mGeCKO and mBrie, respectively, from
(Sanjana et al., 2014) and (Doench et al., 2016). Libraries were amplified and re-sequenced by barcoded Illumina sequencing with
customized primers to ensure proper representation. For virus production, envelope plasmid pMD2.G, packaging plasmid psPAX2,
and library plasmid were added at ratios of 1:1:2.5, and then polyethyleneimine (PEI) was added and mixed well by vortexing. The
solution was stand at room temperature for 10-20min, and then themixture was dropwisely added into 80-90% confluent HEK293FT
cells and mixed well by gently agitating the plates. Six hours post-transfection, fresh DMEM supplemented with 10% FBS and 1%
Pen/Strep was added to replace the transfection media. Virus-containing supernatant was collected at 48 h and 72 h post-
transfection, and was centrifuged at 1500 g for 10 min to remove the cell debris; aliquoted and stored at -80�C. Virus was titrated
by infecting IM cells at a number of different concentrations, followed by the addition of 2 mg/mL puromycin at 24 h post-infection
to select the transduced cells. The virus titers were determined by calculating the ratios of surviving cells 48 or 72 h post infection
and the cell count at infection.
Cellular CRISPR Library ProductionCells were transduced at 500x or higher coverage and at MOI between 0.2 to 0.4 to ensure sufficient coverage of the library and low
percentage of multiple infectants, using a method similar to (Chen et al., 2015; Shalem et al., 2014).
Cell Library TransplantationUn-transduced and library-transduced cells were injected subcutaneously into the right and left flanks ofNu/Numice at 2e7 cells per
flank. The mGeCKO- or mBrie- transduced IMC9 cells were treated in a similar manner, with the exception that 2.5e7 cells were
injected into the right flank only. Tumors were measured every 2 to 3 days by caliper, and their sizes were estimated as spheres.
Statistical significance was assessed by Welch’s t-test, given the unequal sample numbers and variances for each treatment
condition.
Mouse Organ Dissection, Imaging, and HistologyMicewere sacrificed by carbon dioxide asphyxiation or deep anesthesia with isoflurane followed by cervical dislocation. Mouse livers
and other organs were manually dissected and examined under a fluorescent stereoscope (Zeiss). Organs were then fixed in 4%
formaldehyde or 10% formalin for 48 to 96 hours, embedded in paraffin, sectioned at 6 mm and stained with hematoxylin and eosin
(H&E) for pathology. For tumor size quantification, H&E slides were scanned using an Aperio digital slidescanner (Leica). Tumor sizes
in area were measured using ImageScope (Leica). Immunohistochemistry (IHC) was done with well-established markers using
standard protocols (see Key Resources Table for reagent information). IHC staining quantification was done manually using
ImageScope (Leica), where applicable.
e3 Cell Systems 8, 136–151.e1–e7, February 27, 2019
Mouse Tissue Collection for Molecular BiologyMouse organs were dissected and collected manually under a dissection scope where applicable. For molecular biology, tissues
were flash frozen with liquid nitrogen, ground in 24 Well Polyethylene Vials with metal beads in a GenoGrinder machine (OPS diag-
nostics). Homogenized tissues were used for DNA/RNA/protein extractions using standard molecular biology protocols.
Genomic DNA ExtractionA total of 3e7 cells per replicate, or 200-800 mg of frozen ground tissue each sample, was used for gDNA extraction for library
readout. Samples were re-suspended in 6 mL of NK Lysis Buffer (50 mM Tris, 50 mM EDTA, 1% SDS, pH 8.0) supplemented with
30 mL of 20 mg/mL Proteinase K (Qiagen) in 15 mL conical tubes, and incubated at 55�C bath for 2 h up to overnight. After all the
tissues have been lysed, 30 mL of 10 mg/mL RNAse A (Qiagen) was added, mixed well and incubated at 37�C for 30 min. Samples
were chilled on ice and then 2 mL of pre-chilled 7.5 M ammonium acetate (Sigma) was added to precipitate proteins. The samples
were inverted and vortexed for 15-30 s and then centrifuged atR 4,000 g for 10 min. The supernatant was carefully decanted into a
new 15mL conical tube, followed by the addition of 6 mL 100% isopropanol (at a ratio of� 0.7), inverted 30-50 times and centrifuged
atR 4,000 g for 10 minutes. Genomic DNA should be visible as a small white pellet. After discarding the supernatant, 6 mL of freshly
prepared 70% ethanol was added, mixed well, and then centrifuged at R 4,000 g for 10 min. The supernatant was discarded by
pouring; and remaining residues was removed using a pipette. After air-drying for 10-30 min, DNA was re-suspended by adding
200-500 mL of Nuclease-Free H2O. The genomic DNA concentration was measured using a Nanodrop (Thermo Scientific), and
normalized to 1000 ng/mL for the following readout PCR.
Demultiplexing and Read PreprocessingRaw single-end fastq read files were filtered and demultiplexed using Cutadapt (Martin, 2011). To remove extra sequences down-
stream (i.e. 3’ end) of the sgRNA spacer sequences, we used the following settings: cutadapt –discard-untrimmed -aGTTTTAGAGCT
AGAAATGGC. As the forward PCR primers used to readout sgRNA representation were designed to have a variety of barcodes to
facilitate multiplexed sequencing, we then demultiplexed these filtered reads with the following settings: cutadapt -g file:fbc.fasta
–no-trim, where fbc.fasta contained the 12 possible barcode sequences within the forward primers. Finally, to remove extra se-
quences upstream (i.e. 5’ end) of the sgRNA spacers, we used the following settings: cutadapt –discard-untrimmed –gGTGGAAAGG
ACGAAACACCG. Through this procedure, the raw fastq read files could be pared down to the 20 bp sgRNA spacer sequences.
Mapping of sgRNA Spacers and Quantitation of sgRNAsHaving extracted the 20 bp sgRNA spacer sequences from each demulitplexed sample, we then mapped the sgRNA spacers to the
library from which they originated (either mGeCKO or mBrie). To do so, we first generated a bowtie index of either sgRNA library
using the bowtie-build command in Bowtie 1.1.2 (Langmead et al., 2009). Using these bowtie indexes, we mapped the filtered fastq
read files using the following settings: bowtie -v 1 –suppress 4,5,6,7 –chunkmbs 2000 –best. Using the resultant mapping output, we
quantitated the number of reads that had mapped to each sgRNA within the library. To generate sgRNA representation bar plots, we
set a detection threshold of 1 read and counted the number of unique sgRNAs present in each sample.
Normalization and Summary-Level Analysis of sgRNA AbundancesWe normalized the number of reads in each sample by converting raw sgRNA counts to reads per million (rpm). The rpm values
were then subject to log2 transformation for certain analyses. To generate correlation heatmaps, we used the NMF R package
(Gaujoux and Seoighe, 2010) and calculated the Pearson correlations between individual samples using log2 rpm counts. To calcu-
late the cumulative distribution function for each sample group, we first averaged the normalized sgRNA counts across all
samples within a given group. We then used the ecdfplot function in the latticeExtra R package to generate empirical cumulative
distribution plots.
Determining Significantly Enriched sgRNAs (Custom 3-Way Hit Caller)Three criteria were used to identify enriched sgRNAs / genes: 1) if an sgRNA comprisedR 2% of the total reads in at least one tumor
sample; 2) if an sgRNAwas deemed statistically significantly enriched inR 25%of all tumor samples (formBrie) orR 50%of all tumor
samples (for mGeCKO) using a false-discovery rate (FDR) threshold of 2%based on the abundances of all nontargeting controls; or 3)
ifR 2 independent sgRNAs targeting the same gene were each found to be statistically significant at FDR < 2% in at least one tumor
sample. For the first and second criteria, individual sgRNA hits were collapsed to genes to facilitate comparisons with the hits from
the third criteria.
Individual Gene Validation in Nu/Nu and C57BL/6J MiceTwo to three sgRNAswere chosen for each validation target, corresponding to themost highly enriched sgRNAs foundwithin tumors.
Additionally, 3 NTCs sgRNAswere chosen randomly from themGeCKO library to serve as negative controls. All sgRNA spacers were
cloned in to lentiviral backbone, and virus was produced in HEK293FT cells as described above at a lower scale. IM.C9 cells were
infected with validation constructs and selected for via 5 mg/mL puromycin treatment to produce KO cell lines. For each cell line 4e5
cells were injected into both flanks of a Nu/Nu or C57BL/6 mouse 5-7 days post viral infection. Tumors were measured by caliper
every 2-3 days to assess growth rate.
Cell Systems 8, 136–151.e1–e7, February 27, 2019 e4
SILAC Proteomics Sample PreparationBoth sgPrkar1a and NTC transduced IMC9 cells were grown in standard tissue culture conditions as described previously, with the
exception of the media used. sgPrkar1a cells were cultured with media containing lys6 and arg6 isotopes, deficient in the regular
forms of the amino acids. NTC cells were grown in standard media. Cells were cultured for one week, and > 5 passages to promote
full incorporation of the heavy isotopes. Cells were counted using a countess II machine and trypan blue staining. Protein was sub-
sequently extracted with RIPA buffer on ice for 30 minutes, using 200ml 1x buffer per million cells. Lysate was then spun down for
15 min at 14,000 g. Protease inhibitor cocktail (Roche) was then added to the extracted supernatant to prevent degradation. Protein
concentration was measured and then normalized by Bradford assay with BSA standard. Following Bradford lysates from heavy
sgPrkar1a cells and light NTC cells were mixed in a 1:1 ratio, followed by phosphopeptide enrichment. The flowthrough fractions
contain unmodified peptides, and the eluate fractions contain phosphopeptides, which were subjected to MassSpec analysis.
Proteomics AnalysisProteomic quantification and analysis of SILAC-labeled samples was performed using MaxQuant version 1.6.0.1 (Tyanova et al.,
2016). Specifically, raw LC-MS/MS data files were analyzed using the default Orbitrap and MS/MS instrument parameters for
Arg6/Lys6 ‘‘heavy’’ and Arg0/Lys0 ‘‘light’’ double SILAC-labeled peptides that had no more than three labeled residues. Peptide
searches were then performed using a reference panel, derived from the in silico trypsin digest of the UniProtC57BL/6Jmouse refer-
ence proteome (Proteome ID: UP000000589; last modified: May 16, 2017) (Bayona-Bafaluy et al., 2003; Church et al., 2009) with a
maximum of 2 missed cleavage sites. These peptide searches included carbamidomethyl (C) as a fixed modification, as well as
oxidation (M), acetyl (N-terminal), Gln/Glu -> pyro-Glu, phospho (STY) and deamination (NQ) as variable modifications. In addition,
each peptide was searched for a maximum of 6 variable modifications, a maximum of 200 modification combinations, and the
searches were set to include peptides with >= 7 residues, <= 4600 Da, and unspecific searches required a peptide length between
8 and 25 residues. Co-fragmented peptides from a MS/MS spectrum were identified and quantified. Missing patterns of incomplete
isotope pattern SILAC labeled pairs were re-quantified, based on the identified isotope pattern. The ‘‘Match between runs’ algorithm
was used to identify peptide features with amissingMS/MS spectrum or peptide sequence from another run, based on identifiedm/z
and RT information. For the protein-level analysis, protein group identification required >=1 razor/unique peptide, and protein group
quantification required at least 2 quantified peptide feature ratios between SILAC pairs. Protein ratio calculations included both razor
and unique peptides, as well as unmodified peptides and peptides with oxidation (M) and acetyl (N-terminal) modifications; however,
the unmodified counterparts/versions of these modified peptides were omitted from quantification. The protein group ratios were
calculated using the ‘‘advanced ratio estimation’’ method, which reports the median of the peptide feature ratios or the result of a
regression model for the intensities and log ratios if there is a correlation. Results from the peptide spectrum match (PSM), protein
group and phospho-site analyses had reverse and contaminant sequences removed and were filtered based on a 1% false-
discovery rate (FDR) that was calculated using the target-decoy strategy, in which decoy sequences were derived from reverse pro-
tein sequences. The PSM were also filtered based on the best spectrum match of a PSM (Andromeda score > 40), the difference
between the Andromeda score of the best hit with that of the second best sequence hit (delta-score > 8), and the posterior error
probability (PEP) of a false match (PEP < 0.01). The phospho-site level data were filtered as the PSM with the addition of a
localization probability threshold for post-translational modifications (probability > 0.75). Protein groups were also filtered based
on FDR (q value < 0.01).
Phosphorylation Site Motif AnalysisHypo- and hyper-phosphorylation site (hypo-PS and hyper-PS, respectively) motifs were identified by a motif discovery analysis
performed using the Modification Motifs (MoMo) tool from the MEME software suite version 4.12.0 (Bailey et al., 2009). The motif
discovery analysis used default settings with the Motif-X algorithm (Schwartz and Gygi, 2005) and included PS sequences, adjusted
to a 15 residue width with the PS of interest at the center, and the background sequence database was the UniProt mouse proteome,
which was used as the initial proteomics reference panel. The identified PS motifs were then compared to kinase phosphorylation
motifs using the Tomtom motif comparison tool (Gupta et al., 2007) from the MEME software suite. The motif comparison analysis
was performed using the default settings and kinase phosphorylation motif data from the PhosphoNetworks database (Hu et al.,
2014a; Hu et al., 2014b; Newman et al., 2013).
Western BlotNTC and sgPrkar1a cells were cultured in regular media as above for 48 hours. Total protein samples from were collected, and pro-
tease inhibitor cocktail was added to prevent degradation. Protein concentration was measured and then normalized by Bradford
assay with BSA standard. Western was performed using a standard molecular biology protocol with antibodies detailed in Key
Resources Table.
RNA-seq Differential Expression AnalysisStrand-specific single-end RNA-seq read files were analyzed to obtain transcript level counts using Kallisto (Bray et al., 2016), with
the settings –rf-stranded -b 100. Differential expression analysis between sg-NTC and sg-Prkar1a treated cells were then performed
using Sleuth (Pimentel et al., 2017), with default settings.
e5 Cell Systems 8, 136–151.e1–e7, February 27, 2019
Pathway Enrichment AnalysisFor GSEA (Subramanian et al., 2005) we considered the entire set of genes in the differential expression analysis and sorted the genes
by their fold change values without regard to p-values. This ranked gene list was used for GSEA pre-ranked analysis, in reference to
the C5-all pathway database. We considered a pathway as statistically significant if the q-value was less than 0.1. Using an adjusted
p-value cutoff of 0.05, and a fold change threshold of ±1 (the ‘‘b’’ value from Sleuth output), we determined the set of transcripts
that were significantly upregulated or downregulated with Prkar1a deletion. Converting these transcripts to the gene level, we
then used the resultant gene sets for DAVID functional annotation analysis (Huang et al., 2009a, 2009b), where we only considered
GO categories. We considered a GO category as statistically significant if the Benjamini-Hochberg adjusted p-value was less
than 0.05.
ELISANTCand sgPrkar1a cells were cultured in regularmedia as above for 48 hours. Supernatant protein samples fromwere collected, and
protease inhibitor cocktail was added to prevent degradation. IL-6 ELISA was performed using a commercial kit (RnD).
TIL Isolation for scRNA-seqCells were cultured and prepared for subcutaneous injection as above. Five million cells were injected into each mouse, and tumors
was harvested for scRNA-seq experiments. Tumor tissue was immediately placed in PBS containing 2% FBS to maintain cell
viability. Tumor tissue was then manually cut into small pieces using a razor in a 10cm dish containing 5ml PBS containing 2%
FBS. Tumor tissue was then suspended into 15-20ml PBS and loaded into gentleMACs tubes. Collagenase IV was added in a
1:100 dilution to each sample, prior to loading into the gentleMACs tissue dissociation machine. Tissue was then dissociated for
30’ at 37 C. After dissociation samples were passed through a 100uM filter and re-suspended into 40mL of PBS. Samples were
then spun down at 400g for 5 min, and supernatant was discarded. Cell pellets then were washed with 40mL PBS, and spun
down again at 400g for 5 min and supernatant was discarded. 3mL ACK lysis buffer was added to each cell pellet, gently pipetting
to mix, then incubated for 3-5 min. After incubation samples were re-suspended in 40mL PBS and passed through a 40uM filter to
remove debris. Cells were then washed again, and stained with anti-CD45.2 antibody for 30’ using a 1:100 dilution (1e6 cells per mL).
Prior to FACS, cells were washed 2x in PBS. Cells were then sorted based on CD45.2 (pan-immune cells) andGFP (cancer cells), and
mixed as 90%:10% for single cell transcriptomics analysis. Unique molecular identifier (UMI) and single cell barcoded library prep-
aration and sequencing was done using a slightly modified approach based on previously established methods (Gierahn et al., 2017;
Macosko et al., 2015).
Pre-processing the Single-Cell RNA-seq DataDrop-seq (version 1.12) (Macosko et al., 2015) tools were used to preprocessing all Seq-Well (Gierahn et al., 2017) single-cell RNA-
seq data unless stated otherwise. First, the paired-end read fastq files were merged and converted to Bam files by Picard-
FastqToSam (Picard version 2.9.0). From the Bam files, cellular and unique molecular identifier barcodes were extracted, and reads
were filtered to exclude those with single-base quality scores below 10. The sequences were then trimmed to remove adapters and
polyA tails, and the files were converted back to fastq format by Picard-SamToFastq. The sequences were then mapped by STAR
aligner (version 2.5.3a) (Dobin et al., 2013) to the mm10 mouse reference genome, provided by the McCarroll lab (GEO dataset:
GSE63472) (Macosko et al., 2015). The aligned data were sorted by Picard-SortSam, and the cell and UMI barcode metadata
were attached using Picard-MergeBamAlignment. Gencode annotations were added to the aligned reads, and digital gene expres-
sion (DGE) files were prepared for the 1000 most frequently occurring cell barcodes. For data generated by Chromium Single-cell
3’ RNA-sequencing (10X Genomics), preprocessing was performed using the Cellranger software count-pipeline (10X Genomics),
aligning to the mm10 mouse genome reference with default settings. The pre-processed 10X and DGE matrices were then imported
using the Seurat package (Satija et al., 2015) for the R programming language (version 3.4.4), and the data were filtered to include
genes detected in >5 cells, and cells with >200 detected genes and <5% mitochondrial genes. The data were then normalized to
10,000 transcripts/cell and log-transformed, using a pseudo-count of 1 (ln[x + 1]).
Filtering and Visualization of scRNA-seq DataAll single cell RNA-seq data analysis pipelines were performed using the Seurat package for R (Macosko et al., 2015) (Satija et al.,
2015) (Butler et al., 2018). For single-dataset analyses of NTC-Rag1-/-, sgPrkar1a-Rag1-/-, sgPrkar1a-Nu/Nu and sgPrkar1a-
C57BL/6 tumor micro-environments, variable genes were identified using the FindVariableGenes Seurat function, in which genes
are split into 20 equal-width bins by mean expression, then filtered for those with high dispersion (>0.5 standard deviations from
the average dispersion (log variance/mean ratio) within a bin) and high/low mean expression (>8 or <0.0125) (Satija et al., 2015).
The variance in these variably expressed genes was then assessed by principal components analysis (PCA), and principal
components were dimensionality reduced by t-distributed stochastic neighbor embedding (t-SNE) (van der Maaten and Hinton,
2008) via the Barnes-Hut approximation implementation (van der Maaten, 2014) or PHATE (Moon et al., 2017). Next, cell clustering
was performed using shared nearest neighbor (SNN) algorithm (Waltman and van Eck, 2013). The t-SNE, PHATE and SNN tech-
niques were each performed using the number of principal components that explains majority of the data variance, based on the
‘‘elbow’’ in plots of the partial variance described by each PC (Not shown). For the integrated analysis of sg-Prkar1a and sg-NTC-
treated Rag1-/- cell samples, dimensionality reduction was performed using canonical correlation analysis (CCA). The correlation
Cell Systems 8, 136–151.e1–e7, February 27, 2019 e6
structure of the two datasets was then aligned using the dynamic time warping method, as described previously (Butler et al., 2018).
The aligned, integrated data were visualized by t-SNE and clustered using SNN, as before.
Cluster Analysis of scRNA-seq DataCluster-specific marker genes were selected by differential expression analyses for each cluster, relative to all other clusters, using
the Model-based Analysis of Single-cell Transcriptomics (MAST) (Finak et al., 2015) with the Surat R package (Satija et al., 2015),
filtering the genes to include those with an adjusted-p < 0.01 and within the top absolute fold-change (95th percentile). The
cluster-specific markers were then used to compare each cluster population with known immune cells from the Immunological
Genome Project data (ImmGen version 1; GEO dataset: GSE15907) (Heng et al., 2008). More specifically, the marker expression
was averaged among cluster/cell type, log-normalized (as described above), and compared between datasets by Kendall
correlation analysis (fast approximationmethod by pcaPPRpackage), after which, the correlation coefficients from each comparison
were scaled and visualized by heatmap (pheatmap R package). In addition, this correlation heatmap data were grouped by ImmGen
(Heng et al., 2008) cell categories, and the specific cell types were arranged within these categories by unsupervised hierarchical
clustering.
Tumor cell markers were determined by comparing the NTC-treatedRag1-/- tumor cell bulk RNA-seq transcriptional data with that
of the ImmGen innate immune populations, using log-normalized, group-mean, gene-level expression values. The tumor markers
were selected as the top 50 genes, based on the tumor/immune cell transcriptional fold-change, that were detected in each
scRNA-seq dataset. Four representative tumor cell markers were chosen, after filtering for the genes, based on the detected
expression in single-cell data clusters (>= 0.50 detection) and by cluster-specific fold-change (fold-change >= 1.5). Note that these
representative genes were among the top 20 tumor markers, and were only selected for t-SNE/PHATE visualization purposes.
Lastly, all differential expression analyses were performed by MAST (Finak et al., 2015).
Flow CytometryFlow cytometry analysis of immune populations was performed using a BD FACSAria cytometer. The cell surface antibody staining
protocol for the analysis involved blocking Fc receptors with TruStain FcX anti-mouse CD16/32 for 10 min at 4C (1ug/1e6 cells
in 100uL staining buffer (0.5% (w/v) BSA, 2mM EDTA in PBS, pH 7.4); Biolegend), washing cells, staining cells with the appropriate
fluorophore-conjugated antibodies for 30 minutes at 4C (1uL antibody/1e6 cells in a total volume of 100uL staining buffer), washing
cells, and then resuspending in PBS, prior to FACS. Antibodies are listed in Key Resources Table.
Blinding StatementInvestigators were blinded for tumor sizemeasurement and histology scoring, but not blinded for screen data analysis, proteomics or
RNA-seq.
QUANTIFICATION AND STATISTICAL ANALYSIS
Standard data analysis (Non-NGS) were performed using regular statistics, where NGS data were analyzed with specific pipelines
described in methods under separate sub-headlines. Data comparison between two groups was performed using a two-tailed un-
paired t-test or non-parametric Wilcox test. p values and statistical significance were estimated for all analyses. Prism (GraphPad
Software Inc.) and RStudio were used for these analyses.
DATA AND SOFTWARE AVAILABILITY
Reagent AvailabilityReagents including plasmids and libraries are available to the academic community through Addgene. Other reagents used in this
study are described in Key Resources Table.
Code AvailabilityAll custom scripts used to process and analyze the data are available upon reasonable request.
Data Availability and AccessionsPhosphoproteomics data were deposited into PRIDE database under PRIDE: PXD011837. The mRNA-seq data were deposited into
NCBI SRA under GEO: PRJNA506858. Single-cell RNA-seq were deposited into GEO under GEO: PRJNA509929.
e7 Cell Systems 8, 136–151.e1–e7, February 27, 2019
