Lab on a Chip › work › wp-content › uploads › 2017 › 11 › Clark_LOC_2017.pdfSorting of cells based on specific sequences followed by transcriptome analysis Not reported

Lab on a Chip

CRITICAL REVIEW

Cite this: DOI: 10.1039/c7lc00241f

Received 6th March 2017,Accepted 5th May 2017

DOI: 10.1039/c7lc00241f

rsc.li/loc

Finding a helix in a haystack: nucleic acidcytometry with droplet microfluidics

Iain C. Clark and Adam R. Abate *

Nucleic acids encode the information of life, programming cellular functions and dictating many biological

outcomes. Differentiating between cells based on their nucleic acid programs is, thus, a powerful way to

unravel the genetic bases of many phenotypes. This is especially important considering that most cells exist

in heterogeneous populations, requiring them to be isolated before they can be studied. Existing flow cy-

tometry techniques, however, are unable to reliably recover specific cells based on nucleic acid content.

Nucleic acid cytometry is a new field built on droplet microfluidics that allows robust identification, sorting,

and sequencing of cells based on specific nucleic acid biomarkers. This review highlights applications that

immediately benefit from the approach, biological questions that can be addressed for the first time with it,

and considerations for building successful workflows.

1 Introduction

Across all of biology cells are categorized by their phenotypes.These characterizations are useful because phenotypes deter-mine the roles of different cells in phenomena like nitrogen

cycling (microbes), immune defense (T cells), or sensorystimulation (neurons). Often cells with interesting and com-plex phenotypes exist within heterogeneous populationswhere identifying and studying them can be challenging. Sep-arating cells with particular phenotypes, therefore, is essen-tial for developing a mechanistic understanding of their un-derlying biology. Fluorescence-activated cell sorting (FACS)was developed to accomplish this feat, characterizing largenumbers of cells and isolating subpopulations based on

Lab ChipThis journal is © The Royal Society of Chemistry 2017

Iain C. Clark

Iain Clark graduated from Cor-nell University in 2003 with a B.S. in Biological Engineering. Af-terward he joined the PeaceCorps, working on sustainableagriculture in the Paraguayancountryside. In 2009 he com-pleted a masters in BiosystemsEngineering at UC Davis, wherehe studied waste-to-energy sys-tems and anaerobic microbiol-ogy. He completed his PhD inEnvironmental Engineering atUC Berkeley in 2014, working

under John Coates on the molecular genetics of bacterial chloro-xyanion respiration. He is currently a postdoc in the Abate Lab atUniversity of California, San Francisco, where he is developingnew methods to isolate and study rare cells based on their nucleicacid content. He is interested in droplet microfluidics, highthroughput single cell analysis, molecular genetics, and morebroadly, in leveraging engineering in biology.

Adam R. Abate

Adam Abate graduated fromHarvard College in 2002 with anA. B. in Physics. He received amasters in physics from UCLA in2004, before moving to the Uni-versity of Pennsylvania where, in2006, he received his Ph.D. inPhysics studying the physics ofsoft materials with Douglas Du-rian. He returned to Harvard fora postdoc in the lab of DavidWeitz, working on soft matterphysics, microparticle synthesis,and biological applications of

microfluidics. He is now an Associate Professor at the Universityof California, San Francisco. His research interests include high-throughput biology with microfluidics, protein engineeringthrough directed evolution, and biophysics.

Department of Bioengineering and Therapeutic Sciences, University of California,

San Francisco, San Francisco, CA, USA. E-mail: [email protected];

Tel: 415 476 9819

Publ

ishe

d on

11

May

201

7. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

San

Fran

cisc

o on

25/

05/2

017

18:2

2:51

.

View Article OnlineView Journal

http://crossmark.crossref.org/dialog/?doi=10.1039/c7lc00241f&domain=pdf&date_stamp=2017-05-24

http://orcid.org/0000-0001-9614-4831

http://dx.doi.org/10.1039/c7lc00241f

http://pubs.rsc.org/en/journals/journal/LC

Lab Chip This journal is © The Royal Society of Chemistry 2017

Tab

le1

Dropletworkflowsthat

dem

onstrate

componen

tsofnucleicac

idcy

tometry

Sampletype

Objective

Celllysis

Detection

Sortingmethod

Analysis

Ref.

E.coliwithT4or

ΦX17

4Viral

hosten

rich

men

tThermocycling

Taq

Man

PCR

Droplet

sorting

qPCRof

sorted

drops

Lim,2

017

(ref.1

0)

Lambd

aDNA,

ΦX17

4DNA

Enrich

men

tof

rare

molecules

Non

e(purified

DNA)

PCRmultiplexedTaq

Man

orintercalatingdye

FACSof

dou

ble

emulsions

qPCRof

sorted

dou

ble

emulsions

Suko

vich

,20

17(ref.1

1)

Bacteriop

hag

eT4

andΦX17

4Sequ

ence

specific

enrich

men

tof

virus

Thermocycling

PCRwithintercalatingdye

FACSof

dou

ble

emulsions

qPCRof

sorted

dou

ble

emulsions

Lance,2

016

(ref.1

2)

Human

PC3prostate

cancercellsRaji

B-ly

mph

ocytecells

Sortingof

cellsba

sedon

specific

sequ

encesfollow

edby

tran

scriptom

ean

alysis

Not

repo

rted

Taq

Man

RT-PCR

Droplet

sorting

SinglecellqP

CR,

RNA-seq

onpo

oled

sort

Pellegrino,

2016

(ref.1

3)

SV40

viralpa

rticles

Wholegenom

esequ

encing

ofrare

virus

Thermocycling


Droplet

sorting

Wholegenom

eam

plificationfollo

wed

byIlluminasequ

encing

Han

,201

5(ref.1

4)

E.coliΔtolA,E

.coli

LpoA

K16

8ARaremicrobe

sequ

encing

withou

tcu

lturing

Thermocycling

Taq

Man

PCR

Droplet

sorting

Sangersequ

encingof

lpoA

mutant

Lim,2

015

(ref.1

5)

DNAfrom

lymph

oblast

cells

Enrich

men

tan

dsequ

encing

ofrare

molecules

Non

e(purified

DNA)

Taq

Man

PCR

Droplet

sorting

Illuminasequ

encing

Eastburn,

2015

(ref.1

6)

Viral

genom

icRNA

Sequ

ence

viralrecom

binan

tsNon

e(purified

RNA)

Taq

Man

RT-PCR

Droplet

sorting

Sangersequ

encingof

amplicon

sTao

,201

5(ref.1

7)

Human

DU14

5prostate

cancercells

RajiB-ly

mph

ocytecells

High-throug

hpu

tdetection

andsortingof

cells

basedon

specificsequ

ences

Tween-20

proteinaseK

Taq

Man

PCRan

dRT-PCR

Droplet

sorting

SNPan

alysis,Illumina

sequ

encing,

qRT-PCR

Eastburn,

2014

(ref.1

8)

BC-1

(CRL-22

30)B-cell

withEps

tein–B

arrvirus

SNPan

dCNVdetection

Cells

fixed

Rollin

gcircle

amplification

ofgenom

ictarget

Non

eAnalysis

offluorescent

droplet

imag

esKon

ry,2

013

(ref.1

9)

Human

PC3prostate

cancer,Raji

B-ly

mph

ocyteRNA

High-throug

hpu

tmultiplex

RT-PCR

Purified

RNA

Taq

Man

RT-PCR

Non

eMicroscop

yEastburn,

2013

(ref.2

0)

Human

PC3prostate

cancer,Raji

B-ly

mph

ocytecells

High-throug

hpu

tmultiplex

RT-PCRon

singlecells

Tween-20

proteinaseK

Taq

Man

RT-PCR

Non

eDroplet

cytometry

Eastburn,

2013

(ref.2

1)

S.typh

imurium

SL13

44,

E.coliK12

,en

vironmen

talsamples

Singlecellan

alysis

ofen

vironmen

talsamples

(qPC

R,W

GA)

Thermocycling


Wellrecovery

qPCR,W

GA,Illumina

sequ

encing

Leun

g,20

12(ref.2

2)

Human

cells(KatoIII

andMDA-M

B-231

)Singlecelldigital

RT-PCR

1%TritonX-100

Aga

rose-cap

ture

ofPC

Ram

plicon

and

intercalatingdye

Non

eFlow

cytometry

Zhan

g,20

12(ref.2

3)

Lab on a ChipCritical review

Publ

ishe

d on

11

May

201

7. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

San

Fran

cisc

o on

25/

05/2

017

18:2

2:51

. View Article Online



specific protein biomarkers. However, it is only possiblewhen the biomarkers exist, are known, and can be identifiedby fluorescence, which is often not the case.

Phenotypes are ultimately a manifestation of a genetic pro-gram expressed under specific environmental conditions. As aresult, the clearest biomarker of a phenotype is often the pres-ence of specific sequences encoded in the genome or trans-criptome of the cell. Cancer cells with unique transcriptionalregimes,1–3 cells that harbor integrated provirus,4 anduncultivable microbes, can be uniquely identified by their RNAor DNA content, but often nothing else. Given the centrality ofnucleic acids to all living things, and the unique ability of spe-cific sequences to distinguish important phenotypes, a tool forisolating cells based only on nucleic acids is essential.

Nucleic acid cytometry is an approach for identifying andsorting nucleic acid sequences free in solution or packaged incells or viruses. This is analogous to an experimental BLAST se-quence search and represents a new and powerful techniquefor studying subpopulations of biological importance. Thetechnology uses droplet microfluidics to encapsulate cellularmaterial in water droplets suspended in oil, creating discretereactors. Within each reactor, a soluble detection assay pro-duces a signal that identifies the presence of a specific se-quence in a cell, virus, or molecule. Positive droplets, identifiedby their fluorescence, are sorted and their contents released forfurther analysis. This technology is a valuable alternative to cellsorting based on surface markers, expanding sorting capabili-ties from proteins to nucleic acids. Different technical ap-proaches to nucleic acid cytometry are possible using dropletmicrofluidics; the devices, detection assay, and sorting methodmay vary depending on the application. Indeed, many differentmethods for full and partial nucleic acid cytometry have beenreported in the literature (Table 1). We do not attempt to re-view techniques for microfluidic device construction or design,as these topics have been covered extensively.5–9 The goal ofthis review is to highlight applications in biology that benefitfrom this approach, not to review the different workflows anddevices possible. We provide examples of how nucleic acid cy-tometry can be used in targeted comparative genomics, viralgenome sequencing, non-coding RNA biology, and rare speciesenrichment. In addition, we discuss practical considerationsfor implementation, which draws on our experience developingworkflows.

2 The workflow: digital, high-throughput, and flexible

The first step of nucleic acid cytometry is to partition a sam-ple containing a mixture of nucleic acids into aqueous drop-lets, such that the droplets encapsulate one or no target mol-ecules. A soluble assay is performed in all droplets,interrogating for specific sequences of interest. The simplestembodiment of this operation is digital droplet PCR (ddPCR).Droplets with a target have a higher effective concentrationthan the bulk sample, enabling reliable single copy detection.Upon amplification, positive droplets become fluorescent,T

able

1(continued

)

Sampletype

Objective

Celllysis

Detection

Sortingmethod

Analysis

Ref.

E.coliO15

7:H7,

E.coliK12

cells

Sequ

ence

specific

detection

ofrare

pathog

ens

Thermocycling

Aga

rose-cap

ture

ofmultiplexPC

Ram

plicon

sNon

eFlow

cytometry

Zhu,2

012

(ref.2

4)

Lymph

oblast

cells

Highthroug

hpu

tsequ

encing

ofmultipletargetsin

asingle-cell

SDS,

proteinase

Kin

agarose

hyd

rogels

Beadcapture

offluorescentPC

Ram

plicon

FACSof

bead

sSa

ngersequ

enced

b-actingenean

dch

romosom

altran

slocationt(14

;18)

Novak

,201

1(ref.2

5)

E.coli

Screen

forsequ

ence

motifs

inlibrary

Heat96

°C,1

0minutes


Com

plex

object

parametric

analyzer

and

sorter

Sangersequ

encingof

plasmid

insert

Walser,20

09(ref.2

6)

Human

lymph

ocytes,

E.coliK12

Prod

uction

ofSa

ngeran

dpy

rosequ

encingcompa

tible

amplicon

sfrom

singlecells

Heat95

°C,1

0minutes

Beadcaptureof

fluo

rescen

tPC

Ram

plicon

FACSof

bead

sSa

ngersequ

encing

Kumaresan

,20

08(ref.2

7)

Lab on a Chip Critical review

Publ

ishe

d on

11

May

201

7. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

San

Fran

cisc

o on

25/

05/2

017

18:2

2:51




allowing them to be counted, and providing an absolute mea-sure of the starting target concentration.28,29 The number ofpartitions that can be generated determines the sensitivity ofddPCR for quantifying rare sequences. With ultrahigh-throughput microfluidic techniques, millions of picoliterdroplets can be generated in under an hour, providing overfive decades of DNA quantitation and, thus, significantlyhigher sensitivity than common qPCR. Consequently, ddPCRis displacing qPCR in many instances, particularly in theclinic where absolute quantitation without a standard curveis valuable.

Droplet technologies are not limited to purified DNA: anynucleic acid can serve as a target, provided it can be detectedvia a fluorescence assay. For example, cells, viruses, mRNA,non-coding RNA, or spliced transcripts can all serve as tem-plates for ddPCR, making the approach general. Sample prep-aration depends on the starting material, but the concept isthe same: a heterogeneous sample of nucleic acid entities ispartitioned into droplets and target molecules are detectedwith an assay. Any assay that produces a sequence specificfluorescent signal can be used. Polymerase chain reactionor isothermal amplification techniques (loop mediated,recombinase polymerase, rolling circle, etc.) are usually re-quired to yield sufficient DNA for detection. Amplification iscoupled to a fluorescent signal using hydrolysis probes(TaqMan), molecular beacons, or intercalating dyes (SYBR).Designing the detection assay requires only knowledge ofshort segments of target sequence, and the DNA oligomerprobes are constructed in a robust and inexpensive chemicalsynthesis process. This makes detection of specific cell typesand viruses far simpler than with antibody labeling, whichdepends on the availability of specific and sensitive antibodies.

But microfluidic technologies have developed far beyondthe capability of just partitioning samples into droplets – they

can also reliably sort at kilohertz rates. This adds a new andvaluable layer of capability to digital nucleic acid detection:the ability to isolate positive droplets and analyze their con-tents, thereby allowing correlation of a detected nucleic acidwith whatever material is co-encapsulated in the droplet. Thisopens new avenues in biology, such as isolating genomic mu-tations that correlate with expression of a cancer gene, deter-mining the host of an uncultivable virus, or measuring corre-lations between nucleic acids sequences existing within thesame cell, even if they are on different chromosomes. Nucleicacid cytometry thus consists of two components: digital drop-let detection of target nucleic acids and fluorescence acti-vated droplet sorting (Fig. 1).

3 Cell-free nucleic acid cytometry

While current sequencing technologies have incredible capac-ity that grows each year, they are nevertheless unable to char-acterize all relevant nucleic acids in most biological samples.For example, modern high-throughput sequencers providehundreds of billions of base pairs of sequence information,but a gram of soil can contain over a thousand times thisamount.30 Given the staggering diversity of most samples,the only reasonable approach to answering most questions isto focus sequencing on the relevant nucleic acids. Nucleicacid cytometry affords a powerful solution to this problem,allowing specific molecules to be isolated based on the pres-ence of “keyword” subsequences. Molecules containing thiskeyword are enriched by sorting, discarding unwanted readsfrom abundant and uninteresting molecules, and providingfar deeper coverage of new and interesting ones. Eastburnet al. (2015)16 used this approach to enhance coverage of fivetarget regions in the human genome, which improved variantcalling with low input DNA. Additional applications of this

Fig. 1 Nucleic acid cytometry is a general method for isolating nucleic acids based on the presence of a keyword sequence. The workflowencapsulates material, detects specific DNA or RNA sequences, and sorts droplets containing those sequences. Sorted droplets are amenable tonumerous downstream analysis techniques.


Publ

ishe

d on

11

May

201

7. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

San

Fran

cisc

o on

25/

05/2

017

18:2

2:51




approach include sequencing integrated provirus, studyinggenetic variation of specific chromosomal regions, targetedcomparative genomics of bacteria, and enrichment of specificsequences from cell-free DNA in blood.

3.1 Targeted comparative genomics

Microbes catalyze a staggering array of chemical transforma-tions in the environment including mercury methylation,degradation of xenobiotic compounds, and carbon, nitrogen,and sulfur cycling.31–33 Identifying the genetic bases for thesephenotypes is a first step towards developing a mechanisticunderstanding of them. Microbial transformations generallyrely on multiple proteins that work together, and are oftenencoded in close proximity on the genome. Looking for con-served regions in bacteria with the same phenotype is there-fore useful. Often a single enzyme in the pathway has beenpurified and sequenced, providing a starting point. Se-quences surrounding this are compared to find accessorygenes that may also be important; when large numbers of ge-nomes are available, a greater picture of the genetic diversityand ecological role can be inferred. For example, denitrifica-tion is encoded by, at a minimum, a series of catalytic en-zymes for the sequential reduction of nitrate to nitrogengas.31 Surrounding these genes are detoxification systems forremoving reactive nitrogen species, and sensors and tran-scription factors for regulating expression in the absence ofoxygen. Finding these systems together in multiple organ-isms, especially distantly related ones, supports the hypothe-sis that they are functionally involved in nitrate respiration.

A common challenge, however, is that this type of compar-ative genomics requires genome sequences from organismswith the phenotype of interest. Bulk sequencing of an envi-ronmental sample may completely miss sequences from rare,but important organisms. While abundant sequences canprovide much insight into an environment, they do not allowtargeted sequencing of cells with a particular phenotype. Forthis reason, in microbiology labs the primary strategy is toisolate, culture, and sequence the microbes with the pheno-type of interest. Enrichment by culturing, however, is inher-ently low throughput, and also suffers from cultivation bias,

since the vast majority of microbes cannot be cultured. Phe-notypes of microbes that cannot be cultured, cannot be stud-ied. A cultivation-independent approach to finding and se-quencing regions of interest would enable far more effectivecomparative genomics studies.

Nucleic acid cytometry provides an elegant solution,allowing any gene cluster in an environment to be recoveredwithout the need to culture the host microbes. Using a spe-cific conserved gene as a keyword search, intact molecules inwhich the target sequence is embedded can be recovered andsequenced. The resultant data is a metagenome focusedaround that keyword, in which thousands of variants can becharacterized and compared (Fig. 2). When applied to envi-ronmental DNA, nucleic acid cytometry is a transformativetool for targeted comparative genomics without the need toisolate, culture, and sequence entire genomes. Longer readsequencing, or single drop sorting can be used to preservesingle molecule resolution and facilitate the assembly of re-gions contiguous to detected sequence.

A valuable application of targeted comparative genomicsis mining environmental biochemical libraries for novel geneor pathway variants. This is an especially interesting ap-proach for tailoring microbial synthesis of natural products,including antibiotics.34,35 Chemical synthesis relies on petro-leum feedstocks, heavy metal catalysts, high temperatures,strong acids and bases, and harsh solvents. Many chemicalswith complex structures have no established synthesis path-way or suffer from low yield. In these circumstances, micro-bial synthesis has the potential to greatly improve the eco-nomics of product formation. Enzymes have tailoredsubstrate specificity, are amazingly efficient catalysts at roomtemperature, and can produce high purity products. Genomesequences available in databases can be mined to identify en-zymes in useful pathways. However, the pathways in these da-tabases are skewed towards abundant species. Nucleic acidcytometry addresses this limitation, allowing environmentalsequences to be search experimentally using a keyword thattargets a conserved gene. The resulting data is likely to con-tain a more diverse set of sequences, increasing the chancesof finding novel enzymes with the desired substrate, activity,or product.

Fig. 2 Targeted comparative genomics. An example workflow: cell-free nucleic acid cytometry isolates bacterial operons based on a conservedgene sequence. The enriched material is sequenced, generating reads targeted to a specific region of interest. Bacterial operons are denoted ascolored arrows, with regions of synteny highlighted in grey.


Publ

ishe

d on

11

May

201

7. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

San

Fran

cisc

o on

25/

05/2

017

18:2

2:51




3.2 Viral genome sequencing

Sequencing viruses is difficult due to their rarity and theneed to sequence single genomes separately. Currentmethods to obtain single genomes rely on serial dilution inwell plates, which is severely limited in throughput, providingjust tens of genomes per plate. To obtain a comprehensivepicture of viral diversity and evolution, orders-of-magnitudemore viral genomes must be sequenced. Nucleic acid cytome-try provides a powerful solution to this challenge. Tao et al.(2015)17 used this approach to identify, sort, and sequence re-combination sites after con-infection with murine norovirusstrains MNV-1 and WU20. A recent demonstration of cell-freeDNA sorting enriched for Lambda virus genomes in a back-ground of ΦX174, and used double emulsions to sort usingFACS.36 A similar workflow can be applied to virus that estab-lish latent reservoirs.

Many human viral infections – HIV, herpesviruses, Ep-stein–Barr, human cytomegalovirus, hepatitis B virus, andhuman papillomavirus – establish a silent reservoir that pre-vents eradication.37 This reservoir can be a barrier to diseasetreatment and is the major reason an HIV cure does not ex-ist.38 Cells harboring latent virus can only be identified bythe presence of the proviral sequence, and therefore eludeboth the immune system and our ability to characterize themat the molecular level. However, sequencing proviral genomesin patients with latent infection is an important part of un-derstanding disease progression and prognosis.39 For exam-ple, understanding sequence variation in patients can iden-tify mutations that enable evasion of host immunity and helptailor treatments. Using viral sequences as the keywords bywhich to recover all DNA molecules with integrated provirus,regardless of where in the genome they reside or in whichcell type, yields molecules that contain both full viral ge-nomes and the junctions between virus and host. This allowsinsertion sites to be mapped and related questions to beaddressed.40 As such, nucleic acid cytometry on cell-free DNAis a powerful tool for characterizing latent viral reservoirs.

4 Whole cell nucleic acid cytometry

DNA or RNA markers are often the best indicators of impor-tant phenotypes like cancer41 and pluripotency.42,43 Genome-wide analysis of cells with these nucleic acid markers canprovide critical insights into their underlying biology. How-ever, this requires that a cell's DNA and RNA be co-encapsulatedand preserved during the workflow. Experiments that use pu-rified nucleic acids are not amenable to full genome or trans-criptome sequencing of single cells because nucleic acids arefragmented and mixed during bulk purification. Whole cellanalysis solves this issue by compartmentalizing all cell con-tents in the droplets. After encapsulation, single-cell resolu-tion can be maintained by barcoding or sorting.

Barcoding of genomes44 or transcriptomes45–48 usingdroplet microfluidics uniquely tags cellular DNA or RNA witha nucleotide sequence. After sequencing, this tag is used to

group reads originating from single cells. In silico clusteringof cells based on expression allows subpopulations to bestudied individually and in the context of the larger popula-tion. This approach, although high throughput (>10 000cells), is poorly suited to studying rare cells. Sequencing largenumbers of cells reduces sequencing depth per cell, andmeans that the majority of reads do not contain relevant in-formation. Barcoding also relies on the in silico identificationof a cell through sequencing. This makes correlating RNAand DNA sequences within the same single cells difficult, un-less the genome and transcriptome are barcoded together,which has never before been described in a high-throughputformat.

Nucleic acid cytometry allows the study of cell populationstoo rare to detect with barcoding workflows. Sequencing ofsorted cells yields reads that pertain only to cells of interest,greatly improving coverage of rare molecules. In addition,RNA or DNA can serve as the detection target, and thegenome or transcriptome can be sequenced, providing flexi-bility in how cells are identified and studied. The develop-ment of this technology began with high-throughput se-quence-based detection and counting.20,21,23,24,27 Zhang et al.(2012)23 and Zhu et al. (2012)24 used agarose to encapsulatecells and capture PCR amplicons that were detectable withflow cytometry. Eastburn et al. (2013)21 developed a workflowfor detecting RNA from whole cells in droplets using RT-PCR.Analysis of drops has expanded from counting of positives tothe sorting and analysis of their contents. Novak (2011)25 de-veloped a method for the PCR-based detection of mammaliancells. Positive sorting and Sanger sequencing of targets con-firmed known genomic mutations. Detection of RNA usingRT-PCR, followed by sorting from a background of leuko-cytes, enabled targeted sequencing.18 Subsequent workshowed that sorted cells could be pooled and RNA-sequencedto elucidate genome-wide transcriptional changes associatedwith cancer.13 Given the flexibility of TaqMan assays, this ap-proach can also be adapted to detect and sort based on alter-natively spliced transcripts or the presence of non-codingRNA.

4.1 Finding cells with alternatively spliced transcripts

Alternative splicing of RNA generates a large number of pro-tein isoforms that modify localization, activity, and protein–protein interactions.49 As such, splicing greatly impacts thefunctional biology of a cell, and is associated with both nor-mal cell lineage differentiation and disease.50–52 For example,splice variants of the neuron-restrictive silencer factor (NRSF)are overexpressed in small cell lung cancer, leading to a stopcodon and truncated isoform that has been proposed as aclinical biomarker.53 Nucleic acid cytometry can find andstudy cells expressing this NRSF variant. TaqMan assays candistinguish alternatively splice variants using either exonjunction spanning primers or probes.54 Sorting cells withthese exon junctions can elucidate the underlying geneticbasis and functional consequences of splicing on a genome-wide scale. This is useful for studying splice variants


Publ

ishe

d on

11

May

201

7. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

San

Fran

cisc

o on

25/

05/2

017

18:2

2:51




associated with an important phenotype, but for which nomechanistic understanding has been established.

4.2 Finding cells expressing specific non-coding RNA

There is increasing evidence of the importance of non-codingRNA (ncRNA) in normal cell function and disease.55,56 Micro-RNAs (miRNA) post-transcriptionally silence protein-codinggenes by promoting mRNA degradation and inhibiting theinitiation of translation.57 miRNA can target multiplemRNAs, making it difficult to understand the effect on cellu-lar pathways on a genome-wide scale. The ability to performtranscriptomics on cells enriched for a specific miRNA wouldfacilitate an understanding of normal miRNA function, andcases where miRNA dysregulation causes disease. Nucleicacid cytometry enables this by sorting cells based on expres-sion of target miRNAs. It is also applicable to long noncodingRNAs (lncRNA) that have been ubiquitously detected in se-quencing data, but for which defining functional roles is of-ten challenging. For example, although functional roles havebeen established for some lcRNA, including in X chromo-some inactivation, allelic-specific expression, and regulationof pluripotency, the majority have no known function.56

Comparing the transcriptomes of cells enriched for specificlcRNA to those without could significantly advance discoveryof lcRNA function (Fig. 3).

4.3 Finding extracellular vesicles with specific cargo

Another area where nucleic acid cytometry benefits biology isin the study of extracellular vesicles (EV), which are classifiedbased on their biogenesis and size as exosomes, micro-vesicles or apoptotic bodies. Exosomes can transmit proteins,DNA and RNA between cells; miRNA cargo has been shownto influence host transcription upon fusion, providing amechanism of long-range cell-to-cell communication.58 Exo-some subpopulations have distinct cargo depending on their

source and are therefore likely to impact recipients differ-ently.59 For example, exosomes from cancer may promote cellsurvival, proliferation, and tumorigenesis60,61 while onesfrom B-cells and dendritic cells can promote T cell immuneresponse.62–64

A current challenge is to unravel the heterogeneity of EVs;bulk purification and sequencing approaches lose the linkbetween EV cargo and their originating cells. Sorting ap-proaches for understanding this diversity, including the func-tional significance of minor subpopulations, rely on FACS ofsubmicron particles. This approach has several limitations,including the technical challenges of detecting submicronparticles using FACS,65 and the inability to sort EVs based onmiRNA content. miRNA is a major constituent of exosomes,and can play an important role in their mechanistic action af-ter cargo delivery.66,67 An interesting application of nucleicacid cytometry would be to separate exosomes based on thepresence of cancer-specific miRNA biomarkers.68 This hasthe significant advantage of single molecule detection and isnot subject to sorting challenges based on exosome size; thedrop maker used for exosome encapsulation uniquely deter-mines the sorted drop size, resulting in easily detectable(>20 μm) fluorescent drops. Bulk exosome sequencing couldfirst be used to find markers,69 with subsequent dropletsorting allowing for marker-specific subpopulations to be se-quenced separately, preserving the correlation between DNAand RNA cargo. In this manner, tumor specific mutationscan be related to exosome RNA, allowing the molecular dis-section of tumor heterogeneity without the need for biopsy(Fig. 4).

4.4 Targeted genome sequencing of environmental microbes

Sequencing of environmental samples has revealed massivephylogenetic diversity.70 Many of the microbes identifiedusing these techniques cannot be grown in the lab, and

Fig. 3 Sorting and sequencing cells based on specific RNA expression. Whole cell nucleic acid cytometry encapsulates single cells in droplets anddetects RNA using reverse transcription PCR. Sorting and RNA-seq on positive drops is used to understand the transcriptional landscape of cellswith disease-specific transcription, alternative splicing, or non-coding RNA.


Publ

ishe

d on

11

May

201

7. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

San

Fran

cisc

o on

25/

05/2

017

18:2

2:51




shotgun metagenomics fails to adequately reconstruct theirgenomes, either because they are exceedingly rare, or becauseassembly is difficult in complex samples. Although meta-genomics can provide insight into the cultivation conditionsrequired for different phylogenetic subsets, the majority ofmicrobes identified by sequencing remain uncultivable. With-out cultivation, genomes from many taxonomic subsets can-not be purified, preventing deep sequencing and assembly.

As a proof of concept, Lim et al. (2015)15 sorted E. colibased on the presence of tolA from population of E. coliΔtolA. The sorted cells contained two point mutations in theouter membrane lipoprotein LpoA (K16A), which was con-firmed by sequencing. This approach lays the groundworkfor isolating unique subpopulations from environmentalsamples, providing a means of enriching genomes from taxo-nomic groups without the need to culture. For example, 16Ssequencing of an environmental sample might reveal a novelbacterial clade. Using primers that target this group, cells aresorted, massively enriching their genomes for sequencing(Fig. 5). This approach greatly facilitates the genomics ofnovel uncultivable clades. Since this can be applied to enrichfor organisms based on any sequence keyword, any gene canbe used as a biomarker for sorting. For instance, a functionalgene involved in antibiotic resistance could be used to se-quence and understand the genomes of drug-resistant bacte-

ria. Similarly, the presence of a viral sequence can be used tounderstand the host-range of phage or to genome sequencespecific subsets of the viral community.10

5 Environmental virus sequencing

Viruses are the most abundant biotic unit on earth,71,72 yettheir genomic diversity is mostly unexplored due to technicallimitations in cultivation, isolation, sequencing, and classifi-cation. Moreover, compared to bacterial sequences generatedvia metagenomics, of which 10% may be novel, over 60% ofviral metagenomes have no sequence equivalents in currentdatabases.73 No single sequence, like ribosomal DNA in bac-teria, is conserved in all viral genomes, which makes diversityestimates difficult and necessitates full genome sequencing.Viral metagenomics has greatly expanded the genomes in se-quence databases and therefore diversity estimates,74 butchallenges remain in de novo assembly of genomes fromsamples with incredible sequence variability and uneven cov-erage.75 Often it is desirable to complement metagenomicsapproaches with single virus sequencing, in which all readsoriginate from a single genome. This increases coverage, im-proves assembly, and facilitates the prediction of open read-ing frames. Isolation of single viral particles with FACS allowsfor full genome sequencing,76,77 but is biased to abundant

Fig. 4 Sorting and sequencing of extracellular vesicles. EVs are sorted based on the presence of miRNA and genome sequenced, revealing thegenotype of their cellular origin.

Fig. 5 Targeted sequencing of microbial genomes. Novel microbial clades are chosen for enrichment based on phylogenetic affiliation. Sortingand sequencing is used to understand the genomics of these uncultivable clades.


Publ

ishe

d on

11

May

201

7. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

San

Fran

cisc

o on

25/

05/2

017

18:2

2:51




species or those that can be labeled with antibodies. This isinadequate given that most viruses are known only by thepresence of contigs generated during environmental sequenc-ing. Isolating viruses based on sequence keywords can drasti-cally increase the number of genomes that can be studied.This approach has recently been demonstrated by enrichingSV40 genomes from a heterogeneous sample14 and T4 ge-nomes from a background of ΦX174.78

Viruses represent a vast genetic reservoir of virulence fac-tors with human health implications, and metabolic genesthat can influence the biogeochemical cycling of nutrients bybacterial hosts.79,80 A better understanding of the viral se-quence space is needed to predict the origin of environmen-tal pathogens and to understand the role of phage in the evo-lution and function of their microbial hosts. Using nucleicacid cytometry, it is possible to detect pathogenic factors orbacterial metabolic genes in virus, which would allow for agreater understanding of how viral gene reservoirs transferpathogenic islands and manipulate bacterial sulfur and nitro-gen cycling.74

6 Building successful workflows

Applications of nucleic acid cytometry are numerous becauseany nucleic acid can be used to detect a subpopulation andany co-encapsulated nucleic acid can be analyzed. However,the design and execution of experiments is nontrivial, and abalance between the experimental goals and technical hur-dles must be reached. Key decisions when planning an exper-iment include: the choice of starting material, sample prepa-ration and handling, the detection method, sortingresolution and throughput, and downstream analysis. Eachof these considerations is highly coupled, necessitating thedevelopment of an integrated biological and microfluidicworkflow.

6.1 Sample selection: purified nucleic acids or whole cells?

The choice of starting material is largely governed by thenucleic acid detection target and the analysis goals. In manycases, a simplified workflow that encapsulates purifiednucleic acids can be used. Using cell-free nucleic acids sim-plifies the workflow by avoiding cell handling and obviatingthe need for in-drop cell lysis. This approach is sufficientwhen small fragments (<1 Mbase) and their surroundingcontext are being studied.16 When using purified nucleicacids as the starting material, careful quantification of con-centration is required to achieve appropriate loading of tar-gets per drop. As in digital droplet PCR, controlled fragmen-tation is used to evenly load droplets with nucleic acids andovercome sample viscosity effects. Using purified DNA orRNA is appealing for biological applications where thedetected sequence is contiguous to the analyzed sequence,for example when a gene is detected and its adjoining operonis sequenced.

Whole cells are used when the detected sequence is dis-tant from the analyzed sequence or when the detected se-

quence is a physically distinct molecule from the analyzed se-quence. For example, when the detected sequence is on onechromosome and the analyzed sequence is on another, orwhen the detected sequence is DNA and the analyzed se-quence is RNA. In such cases, encapsulating whole cells pre-serves the cellular link between these two molecules. In the-ory, whole cell experiments can be used to detect any nucleicacid, and determine any other measurable property of thedetected cell. For example, cells can be sorted based on a ge-nomic mutation and transcriptome sequenced or cells can besorted based on a transcript and genome sequenced.

6.2 A cultured approach: tips for cell handling

Careful cell handling is an important part of successful ex-periments. Cells damaged by aggressive centrifugation andre-suspension, or by prolonged incubation at room tempera-ture lyse and release nucleic acids into solution. This resultsin droplets that contain a detection target but no cell. This ismore pronounced when the target is RNA, because RNA copynumber can be high and is more readily released when mem-brane integrity is compromised. In more extreme cases,strands of precipitated DNA can clog the microfluidic deviceinlet. To minimize lysis before encapsulation, it is best to re-duce residence time in the syringe and keep cells cold duringinjection.

Cell staining is used to identify cell-containing drops dur-ing sorting. Eukaryotic cells are stained using cell permeantcalcein dye that is hydrolyzed by intracellular esterases. Afterstaining, cells are re-suspended in a density matching solu-tion (OptiPrep) to prevent settling and maintain the desiredencapsulation frequency. The concentration of OptiPrep isadjusted based on the cell's density.

In whole cell experiments, cells are lysed to release nucleicacids for detection. Lysis methods depend on cell type anddetection target. While many virus and extracellular vesiclesmay not require lysis,14 larger mammalian cells do.21 Deter-gents, salts and proteases are commonly employed. There aretwo major considerations: both the lysis agent used and thereleased cellular material can inhibit detection or post-sortmolecular biology. Cell lysis with Proteinase K followed byheat inactivation, dilution of cell lysate and drop splitting,has been used to lyse and mitigate inhibition.21 Dilutionfollowed by drop splitting limits the material available forlater sequencing, which is especially important in single cellapplications. Reverse transcription, even in the presence ofinactivated cell lysate, is sub-optimal and may limit coveragein transcriptomic applications.47

An alternative to diluting cellular material is to use beadsor hydrogels to capture nucleic acids, allowing for the re-moval of lysis buffers and inhibitors.25,46 Hydrogels provideflexibility in cases where particularly harsh lysis is required,for example when using enzymes to break down peptidogly-can or release DNA from chromatin. Hydrogels are porous,allowing bulk biological reactions if nucleic acids remain cap-tured. However, hydrogels also permit leakage of cellular ma-terial and the potential for cross-contamination.


Publ

ishe

d on

11

May

201

7. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

San

Fran

cisc

o on

25/

05/2

017

18:2

2:51




In any droplet workflow, cross contamination can occur atseveral stages. Cell lysis in the syringe, co-encapsulation dueto cell–cell adhesion or improper re-suspension, dropletmerger during thermocycling or re-injection, and false posi-tive sorts can all contribute to cross contamination. Crosscontamination can be measured with two-population experi-ments, where the two cell types can be uniquely differenti-ated, such as mouse and human cells. Two cell experimentsprovide an estimate of the workflow's ability to isolate targetnucleic acids. When RNA is detected using reverse transcrip-tion PCR, the presence of RNA positive drops that do not con-tain a calcein stained cell is an indicator of cross-contamination during cell handling or drop making.

6.3 Detectomics: assay selection for downstream omics

Sequence specific single molecule nucleic acid detection isdifficult, and amplification of a target sequence using poly-merase chain reaction or isothermal techniques is generallyused to yield a detectable signal. If RNA is the target, a re-verse transcription step is added prior to amplification. De-tection of amplified material relies on probes that bind theamplicon, or dyes that intercalate DNA or detect a byproductof amplification. TaqMan hydrolysis assays have been widelyadopted because they add an additional layer of specificity;primer dimers and non-specific amplification are much lesslikely to produce a false positive signal than intercalatingdyes. Validated TaqMan qPCR assays are widely reported inthe literature and are available from several manufacturers.In general, TaqMan assays are straightforward to design anduse, but our experience has been that robust assays in bulkdo not always translate to droplet format. Often several as-says must be screened for adequate signal to noise separa-tion. Multiplexing of TaqMan assays is common in droplets,and allows for detection of multiple targets.81 Other detectionstrategies, including FRET and molecular beacons have alsobeen successfully implemented,82–85 but may require morecareful design and experimentation to achieve a specific andsensitive signal.

It is important to consider downstream analysis duringthe selection of a detection method. In particular,thermocycling at high temperature in the presence of diva-lent cations damages RNA, which may disrupt RNA sequenc-ing applications. For single-cell work, preserving RNA qualityduring detection is important. RNAses can significantly de-grade RNA; it is best to lyse and inactivate cellular materialquickly after encapsulation. DNA is also fragmented duringthermocycling; in single cell sorting applications this de-creases the length of material that can be amplified aftersorting. For example, if a 10 kilobase provirus is the target,DNA fragmentation could necessitate stitching smalleramplicons to cover the full region. Isothermal methods in-cluding loop-mediated isothermal amplification,86 andrecombinase polymerase amplification87 are better suited insuch cases, especially when processing single cells. A minorissue is drop coalescence during thermocycling, which can

make sorting unreliable. Several tactics can reduce coales-cence. Avoid handling tubes with gloves, which tend to carryand transfer charge. Reduce temperature and cycles as muchas possible and use FC-40 oil instead of HFE-7500 duringthermocycling. PCR additives like PEG 6k and Tween-20 im-prove droplet stability, but should be evaluated with respectto detection performance. Droplet filters that selectively re-tain large drops can also be beneficial when drops are re-injected for sorting.

6.4 Droplet sorting: the good and the bad

Some of the first applications of droplet sorting were in en-zyme evolution, which used bulk homoginization to formdouble-emulsions that could be FACS sorted.88,89 However,emulsions generated by bulk techniques are polydisperse,which translates to increased assay variability and bias insequencing results. Because of this, microfluidics is nowoften used to generate extremely uniform oil-in-wateremulsions.90–92 Microfluidic sorting can isolate dropletsbased on fluorescence, absorbance, or electrochemical prop-erties. Solenoid valves,93 acoustic waves,94,95 piezoelectric ac-tuation,96 thermocapillary valves,97 and dielectrophoresis(DEP)98 have all been used to actively sort droplets, but DEPis the most common due to its speed and the simplicity ofthe requisite devices, particularly when using conductive liq-uid electrodes.99

A system for DEP sorting of fluorescent droplets isdepicted in (Fig. 6). Lasers are aligned onto the microfluidicchannel where they excite fluorophores in droplets. Fluores-cence is collected using photomultiplier tubes, which convertphotons into electrical signals that are processed with a com-puter. When the fluorescence reaches a threshold value, ahigh voltage AC signal is applied to an on-chip electrode, cre-ating a DEP force that pulls the droplet into a collectionchannel.

Dielectrophoretic droplet sorters generally run at less than1 kHz, although careful design has pushed this limit to 30kHz.100 Sort speeds are constrained by the ability to accu-rately detect positive droplets, to apply a force strong enoughto deflect a droplet's path without splitting it, and to selec-tively apply a sort force without influencing neighbors. Theability to detect positive droplets depends on their signal tonoise ratio, which requires careful assay optimization. At fastspeeds, the droplets spend significantly less time in the exci-tation window, making accurate fluorescence measurementschallenging. In addition, the decision to sort relies on real-time analysis of photomultiplier tube signals, which can pro-duce millions of digital data points per second. Analyzingthis data in real time and basing sorting decisions on it re-quires intricately programmed high speed electronics, usuallyfield-programmable gate arrays.100

The number of negative droplets that are incorrectlysorted (false positives) and the number of positive dropletsthat fail to sort (false negatives) are highly dependent on theemulsion quality and sorter design. Several common factors


Publ

ishe

d on

11

May

201

7. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

San

Fran

cisc

o on

25/

05/2

017

18:2

2:51




can greatly influence sorting error. False positives increasewith aperiodic droplet spacing, which leads to drops enteringthe sort junction in close succession and interfering witheach other's sorting. Controlled spacing of emulsions is gen-erally the best approach to prevent aperiodicity and improvesorting efficiency.100 Faster spacer oil can increase the dis-tance between drops and assuage aperiodic effects, but alsomakes the duration of dielectrophoretic pulsing shorter. Volt-age can be increased to a point, but eventually the high fieldstear droplets apart, electrocoalesce droplets upstream, andeven cause a device to fail due to arcing between positive andnegative electrodes.

Droplet monodispersity is paramount for accurate sortingwith current devices. A polydisperse emulsion results infalse positive and false negative sorts. Gating on signalwidth can reject moderately sized coalesced droplets. Typi-cally, a histogram of droplet durations shows multiple peaksfor single, double, and triple-coalesced droplets. Howeverdroplets above a certain size will split at the sort junction,with half entering the positive channel. Another commonfailure with polydisperse emulsions occurs when small drop-lets catch up to larger ones in the reinjection channel.When the larger droplet enters the waste channel, it causesa local increase in pressure that forces the smaller trailingdrop to take the alternative flow path into the sort channel.Adding links between the waste and sort outlets equalizespressure fluctuations,90,100 but cannot entirely prevent thiseffect. False negatives are mostly a result of applying an in-adequate dielectrophoretic force by using the wrong fre-quency, amplitude, or pulse width. At faster speeds, it isbeneficial to detect upstream of the electrode to allow longersorting pulses.

Droplet sorters are customizable, making them ideal foradvanced applications. However, they are also difficult andexpensive to build, and require experience to operate well.This represents a barrier to broad adoption of nucleic acid cy-tometry. It is possible to couple the advantages of dropletswith the ease of use and availability of fluorescence-activatedcell sorting (FACS) instruments using double emul-sions.36,78,101 These water-in-oil-in-water (W/O/W) emulsionscan be loaded onto FACS instruments directly. FACS is a ma-ture technology and instruments are often located in core fa-cilities where they are calibrated daily and well maintained.However FACS sorting of double emulsions also has chal-lenges. Shearing and drop destruction at larger sizes can oc-cur; generally <30 μm double-emulsions are used to avoidthis problem.102 FACS is also susceptible to sorting errors, al-though built-in systems reject doublets. In contrast, micro-fluidic sorters don't exclude droplets that enter the sort junc-tion in tandem. However, microfluidic sorters have severalunique advantages over FACS. The sorting junction is directlyvisible under the microscope at all times, and sorting eventscan be confirmed by triggering a high-speed camera.Whereas shearing and clogging are hidden during FACS, theoperator can observe and correct such anomalies duringmicrofluidic sorting.

Fig. 6 Microfluidic droplet sorting allows for the isolation drops. A) Dropletsorters use an inverted microscope to align lasers onto a microfluidicchannel. B) Drops containing fluorophores are excited as they pass throughthe laser line. Fluorescent signals are collected in real time usingphotomultiplier tubes. C) Gating of positive drops is based on the intensity,shape, and width of the detected signal. Large coalesced drops producewide signals and can be discarded. D) Sorting is achieved using a highvoltage AC signal that generates a dialectrophoretic force on water droplets.


Publ

ishe

d on

11

May

201

7. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

San

Fran

cisc

o on

25/

05/2

017

18:2

2:51




6.5 Pooled versus single-droplet experiments

Sorted droplets can be recovered individually or pooled to-gether. Which mode to use depends on the experimentalgoals and has implications for handling and downstreamprocessing. Pooling discards single-droplet resolution, buthas several clear advantages. It eliminates the need for exces-sive post-sort amplification of material, is more tolerant ofdegradation, and is generally more compatible with standarddownstream molecular biology techniques. This was demon-strated in a recent example, where standard RNA-sequencingwas applied to a sorted pooled subpopulation.13 Single dropsorting, while providing single-cell resolution, has severalchallenges. Positive drops must be reliably sorted, and theircontents released and amplified. Unlike bulk collection strat-egies that can tolerate false negatives or incomplete breakingof drops, these problems compound quickly in single dropsorting. In addition, single drops containing single cells haveonly picograms of DNA and RNA that must be amplified. Thehuman genome, with roughly 3.2 × 109 base pairs, has only6.6 picograms of DNA per diploid cell. Concentrations thislow cannot be visualized on gels or made into libraries for se-quencing, necessitating amplification. While this challenge isnot unique to droplets, it does mean that single dropletworkflows are much more sensitive to loss of material andcontamination. Automation of droplet collection into platesusing a mechanical stage is an additional engineering prob-lem that is necessary for reproducible single drop collection.

6.6 Downstream analysis: to the biology

After sorting, droplets must be broken to release their con-tents. If many droplets are sorted into one collection cham-ber, 1H,1H,2H,2H-perfluoro-1-octanol (PFO) is widely used torelease their contents. This is followed by efficient aqueousextraction, but often requires cleanup to remove PFO, whichis inhibitory to molecular biology. Methods such as electro-coalescence or gently heating in presence of an aqueous over-lay are more compatible with preservation of limited nucleicacids from single cells. If the detection step amplifies nucleicacids, the amplicon represents a potentially large fraction ofthe recovered nucleic acids and can interfere with down-stream reactions. Amplicons can be removed using bio-tinylated primers and streptavidin beads or by using dUTP inplace of dTTP during amplification with subsequent uracil-DNA-glycosylase digestion.14,16 Apart from issues of nucleicacid preservation and release of droplet contents, sorted ma-terial is amenable to virtually any type of molecular analysis,including epigenetic. For the human transcriptome, which issmall compared to the genome, it is possible to sequence theentire transcriptome at a reasonable cost. For human genomesequencing, either targeted amplification of multiple geno-mic locations or whole genome amplification can generateenough material for sequencing from single cells. However,the choice between these two methods is generally driven bycost and the objectives of the study, with researchers se-

quencing large numbers of small regions or small numbersof whole genomes.103

7 Conclusions

Cellular phenotypes are diverse and do not exist in isolation;in most cases cells are part of a heterogeneous population.Therefore, to understand unique subpopulations – cancercells in the blood, immune cells that harbor latent provirus,antibiotic resistant bacteria in the environment – requires arobust isolation technique. Detection and sorting based onsurface markers using FACS does not reliably isolate cellsbased on nucleic acid markers, especially single copy geno-mic targets, micro RNA, or alternatively spliced transcripts.Because important cells are often only identifiable based ontheir genomes or transcriptomes, a technology is needed thatuniquely identifies and sorts based on specific DNA or RNAsequences. Droplet microfluidics provides a tool for detectingnucleic acids, whether they be in cells, virus, or free in solu-tion, and for sorting this material for downstream analysis.We have presented examples of how this approach can be ap-plied in biology, including in the discovery of ncRNA func-tion, extracellular vesicle sequencing, targeted bacterial meta-genomics and environmental viral sequencing. These are justa few of the many areas that can benefit from sorting basedon nucleic acid sequence keywords.

References

1 M. A. Shipp, K. N. Ross, P. Tamayo, A. P. Weng, J. L. Kutok,R. C. T. Aguiar, M. Gaasenbeek, M. Angelo, M. Reich, G. S.Pinkus, T. S. Ray, M. A. Koval, K. W. Last, A. Norton, T. A.Lister, J. Mesirov, D. S. Neuberg, E. S. Lander, J. C. Asterand T. R. Golub, Nat. Med., 2002, 8, 68–74.

2 E. Huang, S. Ishida, J. Pittman, H. Dressman, A. Bild, M.Kloos, M. D'Amico, R. G. Pestell, M. West and J. R. Nevins,Nat. Genet., 2003, 34, 226–230.

3 A. S. Adler, M. Lin, H. Horlings, D. S. A. Nuyten, M. J. vande Vijver and H. Y. Chang, Nat. Genet., 2006, 38, 421–430.

4 G. Morissette and L. Flamand, J. Virol., 2010, 84,12100–12109.

5 B. Kintses, L. D. van Vliet, S. R. Devenish and F. Hollfelder,Curr. Opin. Chem. Biol., 2010, 14, 548–555.

6 A. B. Theberge, F. Courtois, Y. Schaerli, M. Fischlechner, C.Abell, F. Hollfelder and W. T. S. Huck, Angew. Chem., Int.Ed., 2010, 49, 5846–5868.

7 N. Shembekar, C. Chaipan, R. Utharala and C. A. Merten,Lab Chip, 2016, 16, 1314–1331.

8 M. T. Guo, A. Rotem, J. A. Heyman and D. A. Weitz, LabChip, 2012, 12, 2146–2155.

9 X. C. i. Solvas and A. deMello, Chem. Commun., 2011, 47,1936–1942.

10 S. W. Lim, S. T. Lance, K. M. Stedman and A. R. Abate,J. Virol. Methods, 2017, 242, 14–21.

11 D. J. Sukovich, S. T. Lance and A. R. Abate, Sci. Rep.,2016, 7, 1–9.


Publ

ishe

d on

11

May

201

7. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

San

Fran

cisc

o on

25/

05/2

017

18:2

2:51




12 S. T. Lance, D. J. Sukovich, K. M. Stedman and A. R. Abate,Virol. J., 2016, 13, 1–9.

13 M. Pellegrino, A. Sciambi, J. L. Yates, J. D. Mast, C. Silverand D. J. Eastburn, BMC Genomics, 2016, 17, 361.

14 H. S. Han, P. G. Cantalupo, A. Rotem, S. K. Cockrell, M.Carbonnaux, J. M. Pipas and D. A. Weitz, Angew. Chem.,2015, 127, 14191–14194.

15 S. W. Lim, T. M. Tran and A. R. Abate, PLoS One, 2015, 10,e0113549.

16 D. J. Eastburn, Y. Huang, M. Pellegrino, A. Sciambi, L. J.Ptáček and A. R. Abate, Nucleic Acids Res., 2015, 43, e86.

17 Y. Tao, A. Rotem, H. Zhang, S. K. Cockrell, S. A. Koehler,C. B. Chang, L. W. Ung, P. G. Cantalupo, Y. Ren, J. S. Lin,A. B. Feldman, C. E. Wobus, J. M. Pipas and D. A. Weitz,ChemBioChem, 2015, 16, 2167–2171.

18 D. J. Eastburn, A. Sciambi and A. R. Abate, Nucleic AcidsRes., 2014, 42, e128.

19 T. Konry, A. Lerner, M. L. Yarmush and I. V. Smolina,Technology, 2013, 01, 88–96.

20 D. J. Eastburn, A. Sciambi and A. R. Abate, PLoS One,2013, 8, e62961.

21 D. J. Eastburn, A. Sciambi and A. R. Abate, Anal. Chem.,2013, 85, 8016–8021.

22 K. Leung, H. Zahn and T. Leaver, Proc. Natl. Acad. Sci. U. S.A., 2012, 109, 7665–7670.

23 H. Zhang, G. Jenkins, Y. Zou, Z. Zhu and C. J. Yang, Anal.Chem., 2012, 84, 3599–3606.

24 Z. Zhu, W. Zhang, X. Leng, M. Zhang, Z. Guan, J. Lu andC. J. Yang, Lab Chip, 2012, 12, 3907–3913.

25 R. Novak, Y. Zeng, J. Shuga, G. Venugopalan, D. A. Fletcher,M. T. Smith and R. A. Mathies, Angew. Chem., 2011, 123,410–415.

26 M. Walser, R. Pellaux, A. Meyer, M. Bechtold, H.Vanderschuren, R. Reinhardt, J. Magyar, S. Panke and M.Held, Nucleic Acids Res., 2009, 37, e57.

27 P. Kumaresan, C. J. Yang, S. A. Cronier, R. G. Blazej andR. A. Mathies, Anal. Chem., 2008, 80, 3522–3529.

28 P. J. Sykes, S. H. Neoh, M. J. Brisco, E. Hughes, J. Condonand A. A. Morley, BioTechniques, 1992, 13, 444–449.

29 B. Vogelstein and K. W. Kinzler, Proc. Natl. Acad. Sci. U. S. A.,1999, 96, 9236–9241.

30 J. T. Trevors, Antonie Van Leeuwenhoek, 2010, 97, 99–106.31 D. A. Rodionov, I. L. Dubchak, A. P. Arkin, E. J. Alm and

M. S. Gelfand, PLoS Comput. Biol., 2005, 1, e55.32 I. A. C. Pereira, A. R. Ramos, F. Grein, M. C. Marques, S. M.

da Silva and S. S. Venceslau, Front. Microbiol., 2011, 2, 69.33 J. M. Parks, A. Johs, M. Podar, R. Bridou, R. A. Hurt, S. D.

Smith, S. J. Tomanicek, Y. Qian, S. D. Brown, C. C. Brandt,A. V. Palumbo, J. C. Smith, J. D. Wall, D. A. Elias and L.Liang, Science, 2013, 339, 1332–1335.

34 W. S. Mak, S. Tran, R. Marcheschi, S. Bertolani, J.Thompson, D. Baker, J. C. Liao and J. B. Siegel, Nat.Commun., 2015, 6, 10005.

35 X. Tang, J. Li, N. Millán-Aguiñaga, J. J. Zhang, E. C. O'Neill,J. A. Ugalde, P. R. Jensen, S. M. Mantovani and B. S. Moore,ACS Chem. Biol., 2015, 10, 2841–2849.

36 D. J. Sukovich, S. T. Lance and A. R. Abate, Sci. Rep.,2016, 7, 1–9.

37 P. M. Lieberman, Cell Host Microbe, 2016, 19, 619–628.38 R. F. Siliciano, Nat. Med., 2014, 20, 480–481.39 B. B. Simen, J. F. Simons, K. H. Hullsiek, R. M. Novak,

R. D. MacArthur, J. D. Baxter, C. Huang, C. Lubeski, G. S.Turenchalk, M. S. Braverman, B. Desany, J. M. Rothberg, M.Egholm, M. J. Kozal and Terry Beirn, Community Programsfor Clinical Research on AIDS, J. Infect. Dis., 2009, 199,693–701.

40 F. Maldarelli, X. Wu, L. Su, F. R. Simonetti, W. Shao,S. Hill, J. Spindler, A. L. Ferris, J. W. Mellors, M. F.Kearney, J. M. Coffin and S. H. Hughes, Science,2014, 345, 179–183.

41 E. T. Liu, Curr. Opin. Genet. Dev., 2003, 13, 97–103.42 L. Chen and G. Q. Daley, Hum. Mol. Genet., 2008, 17,

R23–R27.43 N. B. Ivanova, J. T. Dimos, C. Schaniel, J. A. Hackney, K. A.

Moore and I. R. Lemischka, Science, 2002, 298, 601–604.44 F. Lan, J. R. Haliburton, A. Yuan and A. R. Abate, Nat.

Commun., 2016, 7, 11784.45 A. M. Klein, L. Mazutis, I. Akartuna, N. Tallapragada, A.

Veres, V. Li, L. Peshkin, D. A. Weitz and M. W. Kirschner,Cell, 2015, 161, 1187–1201.

46 E. Z. Macosko, A. Basu, R. Satija, J. Nemesh, K. Shekhar, M.Goldman, I. Tirosh, A. R. Bialas, N. Kamitaki, E. M.Martersteck, J. J. Trombetta, D. A. Weitz, J. R. Sanes, A. K.Shalek, A. Regev and S. A. McCarroll, Cell, 2015, 161,1202–1214.

47 A. Rotem, O. Ram, N. Shoresh, R. A. Sperling, M. Schnall-Levin, H. Zhang, A. Basu, B. E. Bernstein and D. A. Weitz,PLoS One, 2015, 10, e0116328.

48 A. Rotem, O. Ram, N. Shoresh, R. A. Sperling, A. Goren,D. A. Weitz and B. E. Bernstein, Nat. Biotechnol., 2015, 33,1165–1172.

49 O. Kelemen, P. Convertini, Z. Zhang, Y. Wen, M. Shen, M.Falaleeva and S. Stamm, Gene, 2013, 514, 1–30.

50 J. Tazi, N. Bakkour and S. Stamm, Biochim. Biophys. Acta,2009, 1792, 14–26.

51 R. I. Skotheim and M. Nees, Int. J. Biochem. Cell Biol.,2007, 39, 1432–1449.

52 S. Oltean and D. O. Bates, Oncogene, 2014, 33, 5311–5318.53 J. M. Coulson, J. L. Edgson, P. J. Woll and J. P. Quinn,

Cancer Res., 2000, 60, 1840–1844.54 I. I. Vandenbroucke, J. Vandesompele, A. De Paepe and L.

Messiaen, Nucleic Acids Res., 2001, 29, e68.55 M. Esteller, Nat. Rev. Genet., 2011, 12, 861–874.56 J. T. Y. Kung, D. Colognori and J. T. Lee, Genetics,

2013, 193, 651–669.57 L. He and G. J. Hannon, Nat. Rev. Genet., 2004, 5, 522–531.58 H. Valadi, K. Ekström, A. Bossios, M. Sjöstrand, J. J. Lee

and J. O. Lötvall, Nat. Cell Biol., 2007, 9, 654–659.59 E. Willms, H. J. Johansson, I. Mäger, Y. Lee, K. E. M.

Blomberg, M. Sadik, A. Alaarg, C. I. E. Smith, J. Lehtiö, S. ElAndaloussi, M. J. A. Wood and P. Vader, Sci. Rep., 2016, 6,22519.


Publ

ishe

d on

11

May

201

7. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

San

Fran

cisc

o on

25/

05/2

017

18:2

2:51




60 M. Silva and S. A. Melo, Curr. Genomics, 2015, 16, 295–303.61 W. Xu, Z. Yang and N. Lu, J. Exp. Clin. Cancer Res.,

2016, 35, 156.62 G. Raposo, H. W. Nijman, W. Stoorvogel, R. Liejendekker,

C. V. Harding, C. J. Melief and H. J. Geuze, J. Exp. Med.,1996, 183, 1161–1172.

63 P. de Candia, V. De Rosa, M. Casiraghi and G. Matarese,J. Biol. Chem., 2016, 291, 7221–7228.

64 L. Zitvogel, A. Regnault, A. Lozier, J. Wolfers, C. Flament, D.Tenza, P. Ricciardi-Castagnoli, G. Raposo and S.Amigorena, Nat. Med., 1998, 4, 594–600.

65 T. G. Kormelink, G. J. A. Arkesteijn, F. A. Nauwelaers, G.van den Engh, E. N. M. Nolte't Hoen and M. H. M.Wauben, Cytometry, Part A, 2016, 89, 135–147.

66 J. Halkein, S. P. Tabruyn, M. Ricke-Hoch, A. Haghikia,N.-Q.-N. Nguyen, M. Scherr, K. Castermans, L. Malvaux, V.Lambert, M. Thiry, K. Sliwa, A. Noel, J. A. Martial, D.Hilfiker-Kleiner and I. Struman, J. Clin. Invest., 2013, 123,2143–2154.

67 M. L. Squadrito, C. Baer, F. Burdet, C. Maderna, G. D.Gilfillan, R. Lyle, M. Ibberson and M. De Palma, Cell Rep.,2014, 8, 1432–1446.

68 Y. H. Park, H. W. Shin, A. R. Jung, O. S. Kwon, Y.-J. Choi, J.Park and J. Y. Lee, Sci. Rep., 2016, 6, 30386.

69 F. A. San Lucas, K. Allenson, V. Bernard, J. Castillo, D. U.Kim, K. Ellis, E. A. Ehli, G. E. Davies, J. L. Petersen, D. Li,R. Wolff, M. Katz, G. Varadhachary, I. Wistuba, A. Maitraand H. Alvarez, Ann. Oncol., 2016, 27, 635–641.

70 C. J. Castelle, L. A. Hug, K. C. Wrighton, B. C. Thomas,K. H. Williams, D. Wu, S. G. Tringe, S. W. Singer, J. A.Eisen and J. F. Banfield, Nat. Commun., 2013, 4, 2120.

71 Ø. Bergh, K. Y. BØrsheim, G. Bratbak and M. Heldal,Nature, 1989, 340, 467–468.

72 C. H. Wigington, D. Sonderegger, C. P. D. Brussaard, A.Buchan, J. F. Finke, J. A. Fuhrman, J. T. Lennon, M.Middelboe, C. A. Suttle, C. Stock, W. H. Wilson, K. E.Wommack, S. W. Wilhelm and J. S. Weitz, Nat. Microbiol.,2016, 1, 15024.

73 R. A. Edwards and F. Rohwer, Nat. Rev. Microbiol., 2005, 3,504–510.

74 S. Roux, J. R. Brum, B. E. Dutilh, S. Sunagawa, M. B.Duhaime, A. Loy, B. T. Poulos, N. Solonenko, E. Lara, J.Poulain, S. Pesant, S. Kandels-Lewis, C. Dimier, M.Picheral, S. Searson, C. Cruaud, A. Alberti, C. M. Duarte,J. M. Gasol, D. Vaqué, T. O. Coordinators, P. Bork, S. G.Acinas, P. Wincker and M. B. Sullivan, Nature, 2016, 537,689–693.

75 R. Rose, B. Constantinides, A. Tapinos, D. L. Robertson andM. Prosperi, Virus Evol., 2016, 2, vew022.

76 L. Z. Allen, T. Ishoey, M. A. Novotny, J. S. McLean, R. S.Lasken and S. J. Williamson, PLoS One, 2011, 6, e17722.

77 J. M. Martínez, B. K. Swan and W. H. Wilson, ISME J.,2014, 8, 1079–1088.

78 S. T. Lance, D. J. Sukovich, K. M. Stedman and A. R. Abate,Virol. J., 2016, 13, 1–9.

79 P. G. Cantalupo, B. Calgua, G. Zhao, A. Hundesa, A. D.Wier, J. P. Katz, M. Grabe, R. W. Hendrix, R. Girones, D.Wang and J. M. Pipas, mBio, 2011, 2, e00180.

80 F. Rohwer and R. V. Thurber, Nature, 2009, 459, 207–212.81 D. Pekin, Y. Skhiri, J.-C. Baret, D. Le Corre, L. Mazutis, C.

Ben Salem, F. Millot, A. El Harrak, J. B. Hutchison, J. W.Larson, D. R. Link, P. Laurent-Puig, A. D. Griffiths and V.Taly, Lab Chip, 2011, 11, 2156–2166.

82 A. T.-H. Hsieh, P. J.-H. Pan and A. P. Lee, Microfluid.Nanofluid., 2009, 6, 391.

83 M. Srisa-Art, A. J. deMello and J. B. Edel, Anal. Chem.,2007, 79, 6682–6689.

84 T. D. Rane, H. C. Zec, C. Puleo, A. P. Lee and T.-H. Wang,Lab Chip, 2012, 12, 3341–3347.

85 L. M. Zanoli, M. Licciardello, R. D'Agata, C. Lantano, A.Calabretta, R. Corradini, R. Marchelli and G. Spoto, Anal.Bioanal. Chem., 2012, 405, 615–624.

86 T. D. Rane, L. Chen, H. C. Zec and T.-H. Wang, Lab Chip,2015, 15, 776–782.

87 Z. Li, Y. Liu, Q. Wei, Y. Liu, W. Liu, X. Zhang and Y. Yu,PLoS One, 2016, 11, e0153359.

88 K. Bernath, M. Hai, E. Mastrobattista, A. D. Griffiths, S.Magdassi and D. S. Tawfik, Anal. Biochem., 2004, 325, 151–157.

89 E. Mastrobattista, V. Taly, E. Chanudet, P. Treacy, B. T.Kelly and A. D. Griffiths, Chem. Biol., 2005, 12, 1291–1300.

90 J. J. Agresti, E. Antipov, A. R. Abate, K. Ahn, A. C. Rowat,J.-C. Baret, M. Marquez, A. M. Klibanov, A. D. Griffiths andD. A. Weitz, Proc. Natl. Acad. Sci. U. S. A., 2010, 107, 4004–4009.

91 L. Mazutis, J. Gilbert, W. L. Ung, D. A. Weitz, A. D. Griffithsand J. A. Heyman, Nat. Protoc., 2013, 8, 870–891.

92 F. Gielen, R. Hours, S. Emond, M. Fischlechner, U. Schelland F. Hollfelder, Proc. Natl. Acad. Sci. U. S. A., 2016, 113,E7383–E7389.

93 Z. Cao, F. Chen, N. Bao, H. He, P. Xu, S. Jana, S. Jung, H.Lian and C. Lu, Lab Chip, 2013, 13, 171–178.

94 T. Franke, S. Braunmüller, L. Schmid, A. Wixforth and D. A.Weitz, Lab Chip, 2010, 10, 789–794.

95 L. Johansson, F. Nikolajeff, S. Johansson and S. Thorslund,Anal. Chem., 2009, 81, 5188–5196.

96 J. Shemesh, A. Bransky, M. Khoury and S. Levenberg,Biomed. Microdevices, 2010, 12, 907–914.

97 C. N. Baroud, J.-P. Delville, F. Gallaire and R.Wunenburger, Phys. Rev. E: Stat., Nonlinear, Soft MatterPhys., 2007, 75, 046302.

98 K. Ahn, C. Kerbage, T. P. Hunt, R. M. Westervelt, D. R. Linkand D. A. Weitz, Appl. Phys. Lett., 2006, 88, 024104.

99 A. Sciambi and A. R. Abate, Lab Chip, 2014, 14, 2605–2609.100 A. Sciambi and A. R. Abate, Lab Chip, 2014, 15, 47–51.101 J. Yan, W.-A. Bauer, M. Fischlechner, F. Hollfelder, C.

Kaminski and W. Huck, Micromachines, 2013, 4, 402–413.102 A. Zinchenko, S. R. A. Devenish, B. Kintses, P.-Y. Colin, M.

Fischlechner and F. Hollfelder, Anal. Chem., 2014, 86,2526–2533.

103 C. Gawad, W. Koh and S. R. Quake, Nat. Rev. Genet.,2016, 17, 175–188.


Publ

ishe

d on

11

May

201

7. D

ownl

oade

d by

Uni

vers

ity o

f C

alif

orni

a -

San

Fran

cisc

o on

25/

05/2

017

18:2

2:51



Documents

Lab on a Chip › work › wp-content › uploads › 2017 › 11 › Clark_LOC_2017.pdfSorting of cells based on specific sequences followed by transcriptome analysis Not reported