T cell differentiation dynamics profiled by massively parallel single cell RNA capture and highly multiplexed PCR

Christina Chang1, Eleen Y. Shum1, Nidhanjali Bansal1, Joanna Dreux1, Gretchen Lam1,

Jue Fan1, H. Christina Fan1

1. BD Genomics, 4040 Campbell Ave, Suite 120, Menlo Park, CA 94025

Introduction

• While flow cytometry of cell surface markers has aided the characterization of immune subsets, restrictions in the number of available antibodies and channels can greatly limit the gene profiling resolution in single cells;

• Many immune cells, such as T cells, have defined subtypes that can only be defined by a combination of cell surface markers and intracellular gene markers (i.e. transcription factors);

• Single cell RNA sequencing is an emerging tool to profile gene expression, but can be costly because of the high sample number and its associated sequencing costs;

• Here, using BD™ Resolve we compared 2 approaches to single cell RNA sequencing on in vitro differentiated T cell subsets – by assaying the whole transcriptome (WTA) and by targeting a panel of crucial T cell genes (Targeted), subsampling 2000 cells from the same flowcell for each Th subset;

• We found that both assays were able to differentiate distinct T cell subtypes and activation states by cluster analysis. But with much fewer sequencing reads than whole transcriptome sequencing, targeted assays is a cost effective, high throughput method for understanding T cell dynamics at the single cell level.

CD4Isolationfromhumanwholeblood Thelpercellactivationanddifferentiation

5’3’

Univ = Universal sequence MI = Molecular index

CL = Cell labeldT = oligo(dT)

mRNA

5’

poly AdTMICLUnivbead

3’

3’ 5’

5’mRNA

5’

poly A

dTMICLUnivbead3’

3’

cDNA

Resolve PCR 1 primer *

Universal oligo

dTMICLUnivbead cDNA

cDNA

Multiplex PCR

Multiplex PCR 2

cDNA

cDNASequencing adapter

Final amplification

Read 1

Read 2

Sequencing adapter and library index

Resolve PCR 1 primer *

Universal oligo

Library forward primerResolve PCR 2 primer *

Library reverse primer

* Panel-specific

dTMICLUniv

Universal oligo

dTMICLUniv

dTMICLUniv

cDNAsynthesisonbead HighlyMultiplexTargetedAmplification

Reverse transcription

C

PairONEcellwithONEbarcodedbeadinamicrowell

LysecellstohybridizemRNAontobarcodedcaptureoligos onbead

cDNAsynthesis Sequenceandconstructsinglecellgeneexpressionprofiles

mRNABead

LysedCell

BarcodedBeadCell

Microwell

Targeted(multiplexedPCR)

WTA

~10kcells

Subsample~2000cells

~2000cells

Librarypreparation

B

Procedure

• CD4+ T cells were differentiated in vitro for 11 days with different lineage-polarizing cytokines and placed in the BD Resolve system for single cell capture, mRNA barcoding, and NGS library preparation (Figure 1A-C);

• Single cells from 3 in vitro differentiation conditions were captured in 3 BD Resolve cartridges and 2000 cells each were subsampled for library preparation of WTA or Targeted Assay from the same flow cell to control for sample variabilities (Figure 1C);

• Sequencing data from all 3 differentiation conditions were pooled together to perform data visualization and analysis.

A

Figure 1. Single cell capture and library preparation of in vitro differentiated CD4 T helper cell subsets. A) Total CD4+ T cells were isolated from human whole blood using RosetteSep Human CD4+ T cell enrichment kit (StemCell Technologies) and plated at 1 million cells per ml. CD4 T cell were activated with Human T-Activator CD3/CD28 Dynabeads (Thermo Fisher) and cultured with lineage-polarizing cytokines and neutralizing antibodies. After 11 days of culture, cells were harvested for single cell capture with BD Resolve; B) Workflow of single cell capture and lysis on the BD Resolve cartridge; and C) Library preparation for highly multiplex targeted amplification of single cells.

Results

• Both WTA and Targeted assays were able to distinguish between different in vitro differentiated CD4 T cell subsets (Figure 2A-B);

• Many reads and MIs in the WTA assay are dedicated to high expressing genes such as mitochondrial RNA and ribosomal RNA, driving overall single cell sequencing cost to be higher per cell to achieve optimal coverage of genes of interest (Figure 2C-D);

• Resolve Targeted Assay with a 312-gene T cell panel can still distinguish between different T cell subsets even though it has 10-fold less sequencing footprint (Figure 2C,2E).

• With comparable sequencing depth and cell number, targeted assays were able to detect many T cell relevant genes with higher sensitivity compared to WTA (Figure 2F).

• Multiple levels of analysis were performed to understand differential gene expression between in vitro differentiated Th0, Th1, and Th17 cells;

• As expected, Th1 cells have high expression of IFNG and Th17 cells have high RORC, which is the master regulator of Th17 cells (Figure 3B-C);

• Other top differentially expressed genes are outlined in Table 1.

Figure 2. Comparison between WTA vs Targeted single cell sequencing methods on in vitro differentiated T cells. A) tSNE plot of WTA on the same set of T cell samples as described in Figure 1.To achieve faster tSNE computing time, top 548 genes with highest dispersion index from Resolve WTA were selected for tSNE; B) tSNE plot of Resolve Targeted assay; C) (left) Comparison of sequencing reads required for WTA versus a Targeted assay (right) Higher MI return per read obtained in a Targeted assay; D) Breakdown of MI assigned to gene in the T cell panels versus other genes in T cell WTA, note the high percentage of MIs dedicated to genes with low variability; E) Correlation plots comparing sum of molecules detected in WTA vs Targeted assay for each gene in the T cell panel; F) Bar graphs showing number of molecules detected for various critical T cell genes in Targeted assay vs WTA analysis of Th1 cells.

Figure 3. Analysis of T cell subsets using the BD Resolve Targeted Assay. A) tSNE plot from 2B overlaid with a heatmap highlighting expression of key markers up-regulated in each differentiation condition; B) Boxplot analysis comparing expression level of key genes CD4, IFNG, RORC, and TBX21 between different T cell subsets.

Table 1. Top differentially expressed genes represented in different T cell subsets.

Th1 Th17Th0

-

50

100

150

200

250

300

Targeted WTA

MIs

Det

ecte

d pe

r 100

0 Re

ad

Molecular Index Count Efficiency for T Cell Panel Genes

-

10

20

30

40

50

60

70

Targeted WTA

1000

Rea

ds P

er C

ell

Reads Per Cell Molecular Index Count

Genes with low variability 67%

Mitochondrial and ribosomal genes 24%

Genes in T cell panel 8%Highly variable genes 1%

A

E F

-80 -60 -40 -20 0 20 40 60 80

coord 1

-150

-100

-50

0

50

100

150

coor

d 2

CD45314 cells (98.3%)40952 mols (0.6%)

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

log1

0(nu

mbe

r of m

olel

cule

per

cel

l)

-80 -60 -40 -20 0 20 40 60 80

coord 1

-150

-100

-50

0

50

100

150

coor

d 2

IFNG2831 cells (52.4%)

255354 mols (3.5%)

0

0.5

1

1.5

2

2.5

3

log1

0(nu

mbe

r of m

olel

cule

per

cel

l)

-80 -60 -40 -20 0 20 40 60 80

coord 1

-150

-100

-50

0

50

100

150

coor

d 2

RORC1150 cells (21.3%)2830 mols (0.0%)

0

0.2

0.4

0.6

0.8

1

1.2

log1

0(nu

mbe

r of m

olel

cule

per

cel

l)

-80 -60 -40 -20 0 20 40 60 80

coord 1

-150

-100

-50

0

50

100

150

coor

d 2

TBX213475 cells (64.3%)17566 mols (0.2%)

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

log1

0(nu

mbe

r of m

olel

cule

per

cel

l)

A

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

log10(number of molecules per cell)

CD4

Th0

Th1

Th17P

0 0.5 1 1.5 2 2.5 3

log10(number of molecules per cell)

IFNG

Th0

Th1

Th17P

0 0.2 0.4 0.6 0.8 1 1.2 1.4

log10(number of molecules per cell)

RORC

Th0

Th1

Th17P

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

log10(number of molecules per cell)

TBX21

Th0

Th1

Th17P

B

B C D

Th0 Th1 Th17

Gene log10 (pvalue) # of Cells Expressed Gene log10 (pvalue) # of Cells Expressed Gene log10 (pvalue) # of Cells Expressed

CST7 -109.9 1919 IFNG -inf 1172 CCL20 -inf 1150

IL13 -107.58 427 CSF2 -242.17 1083 CXCL13 -inf 1779

GZMB -77.24 2117 CCL3 -201.74 1051 DUSP4 -166.66 1575

DPP4 -60.76 2020 TBX21 -134.6 1125 LGALS3 -160.93 1678

EGR3 -57.58 836 CCL4 -106 1088 CCL5 -148.85 1936

CD27 -46.94 1596 GZMB -38.21 1138 FOXO1 -78.05 1780

GNLY -46.52 997 IL12RB2 -35.52 1060 RORC -45.56 984

HAVCR2 -43.88 2160 NKG7 -30.44 1032 TLR2 -39.22 429

PRF1 -41.33 1531 PRDM1 -29.01 889 CXCL10 -34.96 1541

FYB -40.96 2086 HAVCR2 -28.76 1154 CCR6 -32.95 246

Results (continued)

Conclusion

• Two approaches to study single cell gene expression on BD Resolve assay reveal distinct gene expression profiles in CD4 cells that were cultured in different lineage polarizing conditions;

• Multiplex PCR-based targeted sequencing approach using BD Resolve allows gene expression profiling of thousands of single cells with significantly lower sequencing cost requirements.

For Research Use Only. Not for use in diagnostic or therapeutic procedures. Trademarks are the property of their respective owners.© 2017 BD. BD, the BD Logo and all other trademarks are property of Becton, Dickinson and Company.

www.bd.com/genomics