Pooled CRISPR screening with lentiviral sgRNA libraries has proven to be a very effective approach to identify genes functionally required to generated particular phenotypes, and RNA-seq is an effective method to detect underlying changes in gene expression generating particular phenotypes. With the advent of droplet microfluidic platforms that enable single-cell molecular analysis on a large scale, distinct expression profiles produced by genetic disruption in a pooled CRISPR knockout screen can be characterized and correlated.

Combining Barcode Libraries with Targeted Gene Expression Profiling for Single-Cell Genetic Analysis

Combining Barcode Libraries with Targeted Gene Expression Profiling for Single-Cell Genetic Analysis

Alex Chenchik¹, Andrei Komarov¹, Paul Diehl¹, Mikhail Makhanov¹, and Costa Frangou¹.¹Cellecta, Inc., Mountain View, CA USA

Abstract DriverMap Expression Profiling Assay Workflow

Multiplex PCR withAnchor primers

Directional1st and 2nd

StrandSynthesis

cDNA

GSPAmplicon

80-200 nt Gene-speci�c fragment

Next-Gen Sequencing (NGS)

FwdGSP

GSPAnchor1

GSP

GSP

Anchor1

Anchor2

RevGSP

GSP

Anchor2GSP

GSP

Index1P7

Anchor2

Anchor1

P7 Index1

P5

Index2

Index2 P5

Total RNA (from cells, tissue, blood, biopsy)

RTReaction Total RNA +

CalibrationStandard

AAAA....

N6N6

The DriverMap work�ow leverages the power of quantitative PCR with Next-Gen Sequencing (NGS). Experimentally validated primers amplify speci�c �xed-length regions of all protein-coding genes in a multiplex reaction. The number of reads of each of the resulting amplicons, as determined by NGS, provides a highly quantitative linear measurement of the abundance of each transcript across a range of 5 orders of magnitude. De�ned amplicons also greatly facilitate alignment and downstream analysis.

NGS read levels detected by RNA-Seq and DriverMap assay for selected high-abundant (10K-100K copies per sample), medium-abundant (1,000-10,000 copies per sample), and low-abundant transcripts (100-1,000 copies per sample) in 50ng of total RNA from seven common cancer cell lines.

DriverMap Sensitivity vs. RNA-Seq

PSME3 PSME3FAT1 FAT1TNFRSF10B TNFRSF10BS100A6 S100A6COPS5 COPS5MUC1 MUC1RCAN1 RCAN1PNMA1 PNMA1TOX4 TOX4CHD4 CHD4CSNK1G3 CSNK1G3RAF1 RAF1MAPK1 MAPK1CSDE1 CSDE1NCOR1 NCOR1HPRT1 HPRT1ITCH ITCHCRAT CRATSP1 SP1FOXA1 FOXA1ASPM ASPMIDH2 IDH2ACSL3 ACSL3CHUK CHUKRALGPS2 RALGPS2KRAS KRASTRIP10 TRIP10FANCI FANCICNOT1 CNOT1SMARCB1 SMARCB1NCOR2 NCOR2S100P S100PSTAG1 STAG1ZNF292 ZNF292

20191817161514131211109876543210

HTR2B HTR2BRORA RORAEPHB2 EPHB2ARHGAP26 ARHGAP26BATF BATFIL6R IL6RVIM VIMGBP1 GBP1DNM1 DNM1UBASH3B UBASH3BTRERF1 TRERF1MASP2 MASP2ST6GAL1 ST6GAL1LCN2 LCN2NLRC5 NLRC5EGFR EGFRBCOR BCORGABBR1 GABBR1ICOSLG ICOSLGINPP4A INPP4AREPS2 REPS2FLT3LG FLT3LGANTXR1 ANTXR1DEPDC5 DEPDC5UTY UTYIL11RA IL11RACTSS CTSS

FUT7 FUT7BEND5 BEND5CCR4 CCR4RUNX3 RUNX3NIPSNAP3B NIPSNAP3BIFNA10 IFNA10NTRK1 NTRK1RNASE2 RNASE2SLCO5A1 SLCO5A1SLAMF1 SLAMF1CCR10 CCR10GLI2 GLI2CYTIP CYTIPCDH5 CDH5GZMM GZMMFGF2 FGF2ABCB1 ABCB1BNC2 BNC2MAGEA11 MAGEA11NCAM1 NCAM1CD180 CD180FPR1 FPR1DOCK2 DOCK2NLRP3 NLRP3PCSK5 PCSK5PDGFRA PDGFRAPTPRC PTPRCIL13RA2 IL13RA2LHCGR LHCGRCD52 CD52IL17F IL17FZBTB32 ZBTB32IFNB1 IFNB1CSF2RB CSF2RBHAL HALGPR171 GPR171SERPING1 SERPING1IL1A IL1AFLT1 FLT1AZU1 AZU1HLA-DPA1 HLA-DPA1MMP9 MMP9CXCL6 CXCL6CD7 CD7IGJ IGJFYB FYB

MCF7

RNA-Seq DriverMap™ hGW19K

HighlyAbundant

MediumAbundant

LowAbundant

231 436 453 468 BR3 T47D ZR75 231 436 453 468 BR3 T47D ZR75

Reads(log2)

MDA-MB-231 cells were transduced with the CellTracker XP barcoded sgRNA library targeting key regulators of motility, tumor suppressors, epithelial-to-mesynchymal (EMT). Transduced cells, cells forming primary tumors, and cells from secondary metastasis in lung were analyzed for representation of sgRNA-barcode constructs by NGS.

Identification of Metastasis-Producing Genes

RNA Input and Reproducibility

The correlation (R-squared values) of detected genes between human universal RNA and total brain RNA remains highly consistent across a 10 pg to 50 ng range of starting total RNA.

10 p

g

50 p

g

200

pg

780

pg

3.1

ng

12.5

ng

50 n

g

10 p

g

50 p

g

200

pg

780

pg

3.1

ng

12.5

ng

50 n

g

10 pg50 pg

200 pg780 pg3.1 ng

12.5 ng50 ng10 pg50 pg

200 pg780 pg3.1 ng

12.5 ng50 ng

Hum

an U

nive

rsal

Human Brain

Hum

an B

rain

Human Universal

P-value

Conclusions

We have developed a set of novel tools for single-cell functional genomics:

1. Clonal barcode libraries for in vitro and in vivo cell labeling.

– CellTracker XP barcode libraries compatiblewith scRNAseq

2. Pre-made and custom barcoded sgRNA and shRNAlibraries compatible with the Perturb-seq geneticscreen protocol.

– Custom CellTracker XP sgRNA or shRNA librariescompatible with scRNAseq

3. DriverMap targeted RNA-seq technology for single-cellexpression profiling in sorted cells

– Genome-wide cell and custom targeted assays for any gene set

The above tools allow changes in gene expression to be directly correlated to gene knockout, knockdown, or activation—depending on the type of CRISPR or shRNA constructed used to make the library – and enable novel insights into single-cell and clonal phenotypes in heterogenous populations.

DriverMap expression profiles of primary and metastasized xenograft mammary tumors. Human breast adenocarcinoma cells (MDA-MB-231) were implanted into mouse mammary fat pads of three animals. After several weeks, primary tumors formed in the mammary tissue at the injection sites, and secondary metastasized tumors formed in the lungs. Panel A shows the full DriverMap profiles of the preimplanted cells, implanted primary tumors, and lung tissue containing the metastasized tumors from the animals. Analysis of the data show clear activation of signature pathways in lung tissue containing the metastatic tumors but not in the primary tumors. Panel B shows profiles for three of these pathways relevant to metastasis: cell migration, adhesion, and proteolysis active.

(A)

(B)

Fat PadInjection

Site Tumor

MetasticTumorin Lung

Profiles of Xenograft Breast Cancer Tumors

Expression Profiling in PDX Models

DriverMap Advantage:

• Directly use RNA isolated from bulktumor sample – no human cell puri�cation

• Speci�c ampli�cation of human mRNAs inthe background of mouse transcripts

Human breast adenocarcinoma cells (MDA-MB-231) were implanted into mouse mammary fat pads of three animals. Primary tumors formed in the mammary tissue at the injection sites, and later secondary metastasized tumors formed in the lungs.

Discovery of Metastasis-Promoting Genes

Panel A: Enrichment of subset of sgRNAs targeting metastasis-related genes in lung metastasis samples versus primary tumor and transduced MDA-MB-231 cells.

Panel B: PCA analysis demonstrates signi�cant clustering of genes promoting distant metastasis.

(A) (B)

Single-Cell Profiling with DriverMap for Screening Applications

Since the sensitivity of the PCR-based DriverMap assay enables comprehensive RNA profiling from very small amounts of RNA, we were able to adapt the protocol to work directly with RNA captured on oligo-dT plates (Panel A). Cells lysates can be placed into a multi-well plate with oligo-dT coated wells. Wells are then washed to remove unbound material and the captured mRNA used as input material for DriverMap RNA expression profiling. In this test experiment single cells from a mixed culture of Jurkat and HEK293 cells were added to each well, then cells were lysed, wells washed, and reagents added to run DriverMap single-tube RT-PCR and multiplex amplification.

Panel B: The PCR products resulting after DriverMap ampli�cation.

Panel C: DriverMap expression pro�les for each well.

Panel D: Analysis of a subset of cell-specific genes enabled clear differentiation of the cell type in each of the wells. 10 pg of control RNA isolated from each type was included in the assay as a control.

Single Cell Sorting Amplify/Transfer

Cultivated CellsDissociated Tissue

70-250 bpamplicons

Oligo-dT Plate

Capture polyA+RNA

Multiplex RT-PCR

Index Plate

Add Indexes and i5/i7 sequences for NGS

NGS

bp

Combine/Purify

Single-Cell Profiling with DriverMap

Jurkat HEK29310pg pooled RNA10pg pooled RNA

DriverMap Single-Cell Profiling HeatMap

ExpressionScale

Highly expressed genes in HEK293

Highly expressed genes in Jurkat

Jurkat HEK29310pg Jurkat RNA10pg HEK293 RNA

expression level

NGS products amplified from single cells by DriverMap

(A) (C)

(B)

(C)

Robust transcriptome-based methods to analyze tissue samples, characterize tumor microenvironments, and identify biomarkers are required to improve prognosis and treatment of diseases such as cancer. However, challenges with current RNA profiling approaches include high sample requirements, poor sensitivity, low dynamic range, and limited throughput. These problems can be addressed with transcript targeting methods, such as Cellecta’s DriverMap™ Expression Profiling System, which provides the sensitivity of multiplex RT-PCR with the throughput and digital readout depth of next-generation sequencing (NGS). The straightforward DriverMap RT-PCR approach incorporates 19,000 validated primer pairs to amplify all human protein-coding genes in a single multiplex reaction that generates a genome-wide transcript profile directly from as little as 10 pg of total RNA. The PCR-based protocol also obviates the need for globin and rRNA depletion from whole blood samples.Genome-wide, validated DriverMap primers also provide a platform for targeted gene expression analysis to specifically look at low expressed genes in diverse, difficult-to-assay samples such as single cells. For this application, we have adapted the DriverMap assay to work with conventional oligo dT-molecular indexing and in conjunction with single-cell barcoding to allow molecular analysis of single-cell and clonal phenotypes in heterogenous populations. This application can also be used with CRISPR sgRNA libraries to obtain single-cell resolution of gene knockouts or other genetic perturbations.

Problem: The average mammalian cell has approximately 200,000 mRNA and approximately 180,000 mRNA (80%) are from several hundred highly expressed genes.

• Even with 2 million reads/cell of RNA-Seq(10x depth) almost all reads are from justa few hundred highly expressed genes.

• Targeted expression profiling, like theDriverMap Assay, is necessary to reliablydetect expression of ca. 95% of thetranscriptome.

Challenges with single-cell(sc) RNA-seq:

– Noisy, snapshot data– Low sensitivity of gene detection (500-2,000)– Complex protocol– High RNA-seq NGS cost

Advantages of DriverMap Targeted RNA-seq:

– Detect 2-5-fold more genes– Simple protocol– 10-fold less cost of NGS

PrimaryTumor

Local Spread

DistantMetastasis

Perturb-Seq Methodology Enables the detection of changes in transcriptional phenotypes induced by activation or repression of target genes by sgRNAs/shRNAs. Doesn’t require isolation of phenotype-specific cells and could be applied to discover novel mechanisms and drug targets in vitro and in vivo for cell:

• Di�erentiation • Motility• Immune Response • Apoptosis• Activation • Drug Resistance• Senescence

Applications

CellTracker XP Barcoded Libraries developed for cell tracking and Perturb-Seq applications. These libraries combine pooled CRISPR (and shRNA) genetic screens with single-cell expression analysis. In all libraries barcoded effector constructs are expressed as polyA+ RNAs from vector. As a result, the sgRNA (or shRNA) and barcode can be detected in the RNA expression data. When combined with single-cell expression profiling, changes in gene expression can be directly correlated to gene knockout, knockdown, or activation—depending on the type of CRISPR or shRNA constructed used to make the library.

CellTracker XP RNA-Seq and Perturb-Seq Libraries

Single-Cell Expression Profiling: Genome-wide vs Targeted RNA-seq

References: 1. Adamson, et al., Cell (2016): 167 (7): 1867–1882. PMID 279847332. Dixit, et al., Cell (2016). 167 (7): 1853–1866. PMID 279847323. Datlinger, et al., Nature Methods (2017). 14 (3): 297–301. PMID 28099430

* Perturb-Seq Library design covered by Chenchik, A., et al., U.S. Patent 9,429,565 August 30, 2016