Monitoring the Single-Cell Stress Response of the DiatomThalassiosira pseudonana by Quantitative Real-Time ReverseTranscription-PCR

Xu Shi, Weimin Gao, Shih-hui Chao, Weiwen Zhang,* Deirdre R. Meldrum

Center for Biosignatures Discovery Automation, Biodesign Institute, Arizona State University, Tempe, Arizona, USA

Directly monitoring the stress response of microbes to their environments could be one way to inspect the health of micro-organisms themselves, as well as the environments in which the microorganisms live. The ultimate resolution for such anendeavor could be down to a single-cell level. In this study, using the diatom Thalassiosira pseudonana as a model species,we aimed to measure gene expression responses of this organism to various stresses at a single-cell level. We developed asingle-cell quantitative real-time reverse transcription-PCR (RT-qPCR) protocol and applied it to determine the expres-sion levels of multiple selected genes under nitrogen, phosphate, and iron depletion stress conditions. The results, for thefirst time, provided a quantitative measurement of gene expression at single-cell levels in T. pseudonana and demonstratedthat significant gene expression heterogeneity was present within the cell population. In addition, different expression pat-terns between single-cell- and bulk-cell-based analyses were also observed for all genes assayed in this study, suggestingthat cell response heterogeneity needs to be taken into consideration in order to obtain accurate information that indicatesthe environmental stress condition.

It was generally assumed in the field of microbiology thatmicrobial cells growing under the same conditions have a

uniform population (1). Based on this assumption, microbiol-ogists in the past decades have been analyzing average values atthe population level to describe microbial behaviors. However,recent studies showed that even isogenic cells exhibit notablediversity and a significant cell-to-cell difference that is an orderof magnitude greater than previously thought for any micro-bial population (2). For isogenic populations, gene expressionheterogeneity could arise from a stochastic process in the ex-pression of individual genes. The amplitude of such stochastic-ity in gene expression is further complicated by many factors,such as regulatory dynamics, transcription rate, and geneticfactors of the cells (3–15). These stochasticities, once amplified,could offer the opportunity to generate long-term heterogene-ity at the cellular level in a microbial population. Therefore,more attention has recently been paid to exploring the hetero-geneity of a small number of cells, even at the single-cell level(16). For example, Lenz et al. dissected and captured subsets ofcells from vertical strata within Pseudomonas aeruginosa bio-films and quantified mRNA transcripts as well as 16S rRNAusing quantitative real-time reverse transcription-PCR (RT-qPCR) (17). By dissecting the heterogeneity, cells with abnor-mal gene expression patterns can be identified. These atypicalexpression patterns may indicate potential environmentalproblems earlier than conventional population-level analysisand improve regulation efficiency, since the stress responsealways starts from a small number of cells, including even asingle cell.

Under adverse conditions such as nutrient deficiency orother environmental stresses, microorganisms can trigger pro-tective response mechanisms for survival. Concurrently, manyregular physiological activities such as photosynthesis may berepressed under these stresses. Directly monitoring the stressresponse of microorganisms to their environments could be

one way to inspect the health of microorganisms themselves, aswell as the environments in which they live. Under laboratoryconditions or biotechnological industry settings, direct moni-toring of the stress response of cultured microorganisms can becarried out easily. However, the same endeavor could pose alarge challenge with many environmental microorganisms.This is because many environmental microbial species/phylo-types often live together as consortia, the cell density of indi-vidual species/phylotypes can be very low, and over 99% ofthem cannot be cultivated under laboratory conditions (18).Thus, many of them are not accessible to conventional meth-ods, which typically require large numbers (�105 to 106) ofcells. In such situations, the pursuit of analysis methods target-ing a few or single microbial cells, which are directly recoveredfrom environments without further cultivation, is necessary.Diatoms are a group of unicellular phytoplankton (19, 20) thatare present in widespread niches, from inland lakes to openoceans (21, 22). It was reported that diatoms contribute up to30 to 40% of the primary productivity in the oceans (23, 24).They play such significant roles in the global carbon cycle thatit is essential to understand what environmental stresses theyare susceptible to and how they respond in order to maintain

Received 4 November 2012 Accepted 31 December 2012

Published ahead of print 11 January 2013

Address correspondence to Weiwen Zhang, wwzhang8@tju.edu.cn, or Deirdre R.Meldrum, deirdre.meldrum@asu.edu.

* Present address: Weiwen Zhang, School of Chemical Engineering & Technology,Tianjin University, Tianjin, People’s Republic of China.

Supplemental material for this article may be found at http://dx.doi.org/10.1128/AEM.03399-12.

Copyright © 2013, American Society for Microbiology. All Rights Reserved.

doi:10.1128/AEM.03399-12

1850 aem.asm.org Applied and Environmental Microbiology p. 1850–1858 March 2013 Volume 79 Number 6

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/a

em o

n 09

Feb

ruar

y 20

22 b

y 21

8.10

3.20

1.16

3.

the primary productivity in oceans. Furthermore, we can mon-itor the stress conditions of the oceans and other water bodiesby direct using native diatoms as biosensors or bioreportors.This is in contrast to the case for other similar research whichrequires introducing foreign species to achieve a similar objec-tive (25).

Nitrogen is an essential element for living organisms and isrequired for the biosynthesis of macromolecules such as aminoacids. It has been reported that the availability of nitrogen inoceans varies drastically on spatial and temporal scales due tophysical and biological processes, and nitrogen has been con-sidered a major limiting nutrient for primary production in theoceans (19, 26). Phosphate is another important element in-volved in many aspects of cellular metabolism, such as ATPsynthesis. It was reported that photosynthesis was disrupted bylow-level phosphorus (27, 28). Iron is a key component offerredoxin, an iron-sulfur protein that controls electron trans-fer (29), and its limitation and restriction of primary produc-tivity have been reported for some ocean regions (30, 31). Be-cause of the short residence time of bioavailable iron (32) andthe extremely low concentration of iron in the surface water,which is only 0.07 nM/kg (33), phytoplankton growth and pri-mary productivity are restricted in vast high-nutrient, low-chlorophyll (HNLC) regions of the Southern Ocean, the equa-torial Pacific, and the North Pacific (30, 34). Nevertheless, dueto their key roles in the ecology and biogeochemistry of theoceans, it is important to further understand the mechanismthat diatoms use to deal with various environmental stresses. Inthis study, Thalassiosira pseudonana (35), a typical centric dia-tom, was applied as a model system to measure the stress re-sponse of microorganisms to their environment at the single-cell level using single-cell RT-qPCR, based on our previousefforts (36, 37). In contrast to previously published single-cellanalyses on mammalian or prokaryotic cells, working with di-atoms has its own particular challenges due to their small size(�5-�m diameter) and protective frustules. We quantitativelymeasured the expression of six genes in single T. pseudonanacells, each with three technical replicates. The single-cell resultsrevealed significant heterogeneity in terms of stress responseswithin the T. pseudonana population. The study provides thefirst quantitative gene expression evidence for the responseheterogeneity of the diatom T. pseudonana to environmentalstresses. Our work also demonstrates the possibility of applyingnative habitants as biosensors to monitor environmental stressconditions.

MATERIALS AND METHODSCell culture. Thalassiosira pseudonana (CCMP1335) cells were obtainedfrom the National Center for Marine Algae and Microbiota (NCAM), andwere grown in f/2 medium at 24 � 1°C (38, 39) under a constant-lightcondition (30 �mol photons m�2 s�1 irradiance measured using a LiCor[Lincoln, NE] instrument) without mixing. The total volume for eachgrowth condition was 20 ml within a 125-ml flask. Cells at mid-exponen-tial phase were harvested by centrifugation at 1,500 � g for 5 min at 4°Cand used to inoculate f/2 medium with or without nitrogen (NaNO3,8.82 � 10�4 M), phosphate (NaH2PO4, 3.62 � 10�5 M), and iron(FeCl3 · 6H2O, 1.17 � 10�5 M) depending on the condition of starvation.Artificial seawater was prepared using chemicals of analytical purity andbased on the formula of Kester et al. (40) and used instead of filterednature seawater for f/2 medium.

Sampling and RNA extraction. For bulk-cell-based analysis, 1 ml cellculture was collected by centrifugation at 1,500 � g for 5 min at 4°C. A3900 hemocytometer (Hausser Scientific, Horsham, PA) was used tocount the cell number directly. An RNeasy Minikit (Qiagen, Valencia,CA) was used to extract RNA from the bulk cells. For single-cell-basedanalysis, a micromanipulator developed in our center (41, 42) was used topick cells from the diluted cell population and load them into individualEppendorf microtubes. This micromanipulator uses a piezoelectric actu-ated diaphragm to dispense/aspirate picoliter-level liquid through a30-�m capillary. Owing to the low flow rates, single cells suffer very littleshear stress, which will minimize the effects on their gene expression pro-file. Thirty individual cells from each growth condition were picked. A ZRFungal/Bacterial RNA MicroPrep kit (Zymo Research, Irvine, CA) wasused to extract RNA from single cells, and the total RNA was eluted into afinal volume of 6 �l in Eppendorf microtubes.

cDNA synthesis. A SuperScript VILO cDNA synthesis kit (Invitrogen,Carlsbad, CA) was used to synthesize cDNA. For cDNA synthesis frombulk-cell RNA, the total reaction volume was 20 �l containing 2 �l 10�SuperScript enzyme mix, 4 �l 5� VILO reaction mix, and 14 �l of elutedRNA. To increase the relative concentration of single-cell mRNA forcDNA synthesis preparation, the total reaction volume was decreased to10 �l, which contained 1 �l of specific primer mixture, 1 �l 10� Super-Script enzyme mix, 2 �l 5� VILO reaction mix, and 6 �l of eluted RNA.After cDNA synthesis, 10 �l diethyl pyrocarbonate (DEPC)-treated water(Ambion, Austin, TX) was added to make the final volume of 20 �l beforethe mixture was used as the template for quantitative PCR analysis.

Quantitative PCR. Primers for RT-qPCR were designed usingPrimer-BLAST (http://www.ncbi.nlm.nih.gov/tools/primer-blast/index.cgi?LINK_LOC�BlastHome). To differentiate PCR productsfrom primer dimers, we selected primers which will generate ampli-cons with sizes around 170 to 220 bp (37). qPCR was performed usingExpress SYBR GreenER qPCR SuperMix kits (Invitrogen, Carlsbad,CA) on an ABI StepOne real-time PCR system for bulk-cell analysisand an ABI 7900HT real-time PCR system for single-cell analysis (Ap-plied Biosystems, Foster, CA). The temperature for qPCR was 10 minat 95°C for an initial hot start, and this was followed by 40 cycles of 15s at 95°C for denaturing, 50 s at 60°C for annealing and extension, and10 s at 75°C for signal detection. There was also another melting curveanalysis step which was set to be the default condition based on thereal-time PCR system. For PCRs, 1 �l of each primer at 4 �M, 5 �l ofmaster mixture, 0.1 �l ROX reference dye, 0.9 �l DEPC-treated water,and 2 �l cDNA were combined. Technical triplicates of PCR analysiswere performed for each gene. Reaction mixtures without cDNA tem-plates served as negative controls. Expression levels of target geneswere normalized against an internal control actin gene.

Data analysis. To describe the distribution variation of single-cellgene expression levels among cells, nonparametric statistic tests which donot require normal distribution of data sets were applied (43). Kolmogo-rov-Smirnov and Kruskal-Wallis analysis of variance (ANOVA) tests wereused to analyze the relationship between four different groups of RT-qPCR measurements using the OriginPro 8.1 software (OriginLab Cor-poration, Northampton, MA). Principal-component analysis (PCA) wasconducted using the SPSS Statistics 20 package (IBM, Armonk, NY) todetermine the possible control variances.

RESULTSGrowth of T. pseudonana under stress conditions. T. pseudo-nana growth was determined by counting the cell number with ahemocytometer directly. Figure 1 showed the growth-time curvesof T. pseudonana under control and three stress conditions. Theresults showed that the initial increases in cell numbers over days1 to 4 were roughly exponential for all conditions, although thegrowth under the nitrogen and phosphate depletion conditionswas at a relatively low rate. After day 4, the cultures under controland iron depletion conditions still maintained exponential growth

Single-Cell Analysis of Thalassiosira pseudonana

March 2013 Volume 79 Number 6 aem.asm.org 1851

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/a

em o

n 09

Feb

ruar

y 20

22 b

y 21

8.10

3.20

1.16

3.

for another 24 h. After day 5, both cultures of the control and irondepletion conditions reached stationary phase, while the cellnumbers under the nitrogen and phosphate depletion conditionsdeclined. The results showed that all three depletion conditionscaused significant decreases in cell growth, with phosphate andnitrogen affected the most. In these cases, the cell number reachedonly 10 to 28% of the peak cell numbers. The slow growth of T.pseudonana under these stress conditions was consistent with pre-vious reports (27, 31). Cells at mid-exponential phase were col-lected for RT-qPCR analysis (Fig. 1).

Primer evaluation. A total of 82 pairs of PCR primers weredesigned and evaluated for 39 different target genes. Since themajor goals of this study were (i) to evaluate the possibility ofusing single cells as biosensors, (ii) to determine the response het-erogeneity of T. pseudonana to various important environmentalfactors (i.e., nitrogen, phosphate, and iron limitation), and (iii)also to compare the results with those previously obtained at thebulk-cell level, the target genes included some of the genes withdemonstrated functions in photosynthesis, iron transportation,and stress responses. Although most of the primers (78 out of 82)functioned well with bulk-cell RNA, only one pair of primers eachwas obtained for nine genes after the evaluation process (see Fig.S1 and Data S1 in the supplemental material). The relatively lowsuccess rate for primer selection reflected the different perfor-mance between bulk-cell-based and single-cell-based RT-qPCRanalyses and also the difficulty of measuring gene expression at thesingle-cell level. The successful primer sets and their correspond-ing gene targets were as follows: psaA, photosystem I (PS I) P700chlorophyll a apoprotein A1 (forward primer, CGGTTCTGCATCTTCAGCATACGGC; reverse primer, GTGCTAAACCAACGGCACGACCT); psaF, photosystem I reaction center subunit (for-ward primer, TGTGGCGCAGATGGCTTACCTC; reverseprimer, TGCACTCGTACTTACTGCGCGTA); psbA, photosys-tem II protein D1 (forward primer, CCACATGGCTGGTGTTGCTGGT; reverse primer, CGACCAAAGTAACCGTGTGCAGCT);psbC (forward primer, TCATCTGCACAAGGTCCAACTGGT;reverse primer, AGCAGCACGACGTTCTTGCCA); psbC, photo-system II reaction center protein (forward primer, TCATCTGCACAAGGTCCAACTGGT; reverse primer, AGCAGCACGACGTTCTTGCCA); hsp90, heat shock protein (forward primer, AGGCTCTTACGGCCGGGGCGGA; reverse primer, AAGACCCGCCA

GCCTCGGAAGCC); rbcL, ribulose-bisphosphate carboxylase(forward primer, AGGCTCTTACGGCCGGGGCGGA; reverseprimer, TGTAGATAACTTGACGACCTGCGCC); actin gene(forward primer, CCGTAGTGAACGCCTATCGTGGC; reverseprimer, CCATCGTCTCGCTGCGGCTG); tubulin gene (forwardprimer, GGACGCTACGTTCCTCGTGCC; reverse primer, GCTCTCGGCCTCCTTCCTCACA); and 18S rRNA gene (forwardprimer, TGCCAGTAGTCATACGCTCGTCTCA; reverse primer,CCTTCCGCGAACAGTCGGGTAT). The primers that func-tioned well at the bulk-cell level but not at the single-cell level arealso provided in Table S1 in the supplemental material.

Enhanced cDNA synthesis by addition of target-specificprimers. cDNA synthesis typically employs random primerswhich generate the least bias in the resulting cDNA (44). However,since we were using total RNA rather than purified mRNA as thestarting template, most of the cDNA synthesized through the ran-dom primers will be rRNA-derived cDNA, which could furthercomplicate the single-cell gene expression (44). To address thisissue and to enhance the yield of cDNA derived from targetmRNA, primers specific to the target genes were added to thereverse transcription reaction mixture so that more mRNA of thetarget genes would be converted to cDNA (45). To ensure detec-tion sensitivity and reproducibility of single-cell qPCR, cDNAfrom each T. pseudonana cell was used to detect a maximum ofthree different genes, each with three technical replicates. In thecDNA synthesis step, 1 �l of primer mixture containing threetarget gene-specific primers (reverse primers which are comple-mentary to the mRNA sequence) was added. The final concentra-tion of each target-specific reverse primer in the qPCR mixtureswas 4 nM. To demonstrate the effects of adding target gene-spe-cific primers on single-cell analysis, we evaluated the single-cellRT-qPCR of three genes, psbC, the actin gene, and the 18S rRNAgene. In this experiment, we diluted the RNA isolated from bulkcells (�106 cells/ml) to the level of a single cell, which is approxi-mately 50 fg/�l (46). During cDNA synthesis, a primer mixturecontaining target gene-specific primers was added to 6 replicates,while another 6 replicates contain only random cDNA synthesisprimers. The results showed that except for the 18S rRNA gene,addition of the target-specific primers could significantly decreasethe quantification cycle (Cq) values by 2 to 4 cycles, which is a 4- to16-times-higher yield of target cDNA than for control samples forthe psbC and actin genes, suggesting the target-specific primers inthe cDNA synthesis reaction were able to improve the yield oftarget cDNA significantly (Fig. 2). No effect was observed for the18S rRNA gene, probably because it is one of the most abundantgenes in the total RNA (47). However, even for the 18S rRNAgene, our results showed that addition of target-specific primerscan improve the qPCR reproducibility by decreasing the standarddeviation of Cq values from 0.2 to 0.1 cycle (Fig. 2). The resultsdemonstrated that adding target-specific primers to the cDNAsynthesis reaction mixture was a useful approach which can im-prove the performance of qPCR, especially for the genes withlarger Cq values.

Selection of internal reference gene. In order to ensure thatthe gene expression across different conditions or analytical plat-forms is quantitatively comparable, expression measurementsneed to be normalized against an internal reference gene (48).While several internal reference genes have been demonstrated inthe bulk-cell-based RT-qPCR analysis, so far limited informationis available regarding the constant expression of these internal

FIG 1 Growth of T. pseudonana cells under various conditions. The arrowindicates the sampling time for gene expression analysis.

Shi et al.

1852 aem.asm.org Applied and Environmental Microbiology

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/a

em o

n 09

Feb

ruar

y 20

22 b

y 21

8.10

3.20

1.16

3.

reference genes across individual cells (49–51). For single-cell-based analysis, relative activities of each target gene against thereference gene were acquired by the ��Cq method (52, 53). Basedon previous studies, we selected three genes, i.e., the tubulin gene,the 18S rRNA gene, and the actin gene (54–56), as candidate ref-erence genes for further evaluation. To simplify the selection pro-cess, only control and iron depletion growth condition were used.A total of 12 cells from control and iron depletion conditions werepicked and subject to expression determination for the tubulingene, the 18S rRNA gene, and the actin gene. The Cq measure-ments for a total of 24 cells (i.e., 12 from control conditions and 12from iron depletion conditions) are presented in Fig. 3. The re-sults showed that the standard deviations (SD) of the Cq values forthe tubulin gene, the 18S rRNA gene, and the actin gene among all24 cells were 0.89, 2.9, and 0.39 cycles, respectively. The actin genehad the smallest variance among cells and was thus selected as aninternal control for our further analysis. The result was also con-sistent with that of Kustka et al. that the expression of the actingene was constitutive under all iron concentration conditions(57). The results also showed that even for the 18S rRNA andtubulin genes, which were widely used as internal controls in var-ious bulk-cell-based RT-qPCR analyses, significant cell-cell heter-ogeneity existed.

Single-cell gene expression under stress conditions. To es-tablish a baseline for single-cell-based analysis, we first performeda bulk-cell-based RT-qPCR for the selected target genes underthree stress conditions. The relative activity of each gene was de-rived from the Cq value, which was normalized first by cell numberand then by the activity of the control growth condition. Theresults showed that except for the hsp90 gene under iron depletioncondition, all other genes were downregulated by the stresses (seeFig. S2 in the supplemental material). Upregulation of the hsp90gene under iron depletion condition was also reported by Thama-trakoln et al., who applied a combined genome-wide and targetedcomparative transcriptomic analysis with diagnostic biochemistryand in vivo cell staining as a platform to identify the suite of genesinvolved in acclimation to iron and associated oxidative stress inT. pseudonana (20). In another study, Allen et al. also found thatthe hsp90 gene was upregulated under iron starvation stress con-dition in a pennate diatom, Phaeodactylum tricornutum (58). Both

the psaA gene, encoding photosystem I P700 chlorophyll II a apo-protein A1, and the psaF gene, encoding a photosystem I reactioncenter subunit, were downregulated under three nutrient defi-ciency conditions. Similar results of a PS I decrease under ironlimitation were also reported by Allen et al. (58). Compared withthe psaA and psaF genes of PS I, the psbA and psbC genes of PS IIwere downregulated more under all nutrient depletion condi-tions, suggesting that photosystem II may be more vulnerable tonutrient depletion conditions than photosystem I; this is consis-tent with the results of Mock et al., who analyzed whole-genomeexpression profiling under several different growth conditions,such as Fe depletion, N depletion, Si depletion, and high temper-ature (59). The rbcL gene was downregulated significantly undernitrogen starvation and phosphate depletion conditions but wasdownregulated only slightly under iron depletion condition. In arecent study, Allen et al. reported that downregulation of severalproteins, such as phosphoribulokinase (PRK) and two enzymessupplying substrate for RubisCO, will lead to a decrease of carbonfluxes toward RubisCO under Fe stress in P. tricornutum (58). Inaddition, comparison of gene expression patterns showed thatalthough T. pseudonana and P. tricornutum were divergent �90million years ago and had vast differences in genome structure(60), they may still share a similar fundamental response mecha-nism to iron starvation. Other than these results, Pearson correla-tion coefficients under different conditions (see Tables S2 to S5 inthe supplemental material) indicated that psaA and psaF were al-ways negatively correlated under different nutrient depletion con-ditions, suggesting that the two genes were regulated by a similarmechanism but in opposite directions under different nutrientdepletion conditions, which was rational since both of them be-long to photosystem I. For psbA and psbC, no such correlation wasfound, which suggested that possibly the regulation mechanismswere different for photosystem II and photosystem I.

For single-cell level analysis, the ��Cq method was used tocalculate the relative expression of each gene against the referenceactin gene. Figure 4 and Fig. S3 in the supplemental material showthe result of qPCR analysis of 6 genes under control and three

FIG 3 Evaluation of three internal control candidates under control and irondepletion conditions. Cq is the qPCR quantification cycle, the fractional cyclenumber where fluorescence increases above the threshold. Twelve cells fromthe control condition and 12 cells from the iron depletion condition were usedto evaluate the consistency of the internal control genes.

FIG 2 Effects of adding target-specific primers. Cq is the qPCR quantificationcycle, the fractional cycle number where fluorescence increases above thethreshold. Six in-parallel reactions (with and without addition of specificprimers) were run.

Single-Cell Analysis of Thalassiosira pseudonana

March 2013 Volume 79 Number 6 aem.asm.org 1853

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/a

em o

n 09

Feb

ruar

y 20

22 b

y 21

8.10

3.20

1.16

3.

stress conditions. For each condition, 30 individual cells werepicked and analyzed. Reactions with large variations betweentechnical replicates and/or with multiple peaks observed in themelting curves were considered failed reactions and were excludedfrom further analysis. Overall, the success rate for qPCRs was ap-proximately 93%. The reproducibility of the qPCR was derivedfrom the SD of the technical replicates of each cell. Based on ourresults, the hsp90, psbA, psbC, and actin genes all had small averageSD values among all samples, which were 0.2041 cycle (0.75% ofaverage Cq values), 0.2109 (0.75%), 0.2116 (0.72%), and 0.2148(0.74%), respectively. For the genes with larger Cq values, al-though the average SD values were almost doubled, to 0.3847(1.2%), 0.4048 (1.2%), and 0.422 (1.3%) for the psaF, psaA, andrbcL genes, respectively, they were still in the relatively low-varia-tion range. In general, our single-cell-level qPCR protocol wasrobust and able to generate reproducible data.

The RT-qPCR results showed that gene expression varied sig-nificantly between individual cells, suggesting significant cell-cellheterogeneity in the T. pseudonana population (Fig. 4), consistentwith the previous conclusions that stochasticity of transcriptioncontributed significantly to the level of heterogeneity within aclonal population and that this heterogeneity cannot be revealedby snap-shot measurements of bulk cells (4, 5, 61). Comparison ofthe distribution patterns between conditions can be achieved byusing the Kruskal-Wallis ANOVA test (62). The results showedthat 4 genes, not including the psaF gene, exhibited independent

expression distribution patterns under four growth conditions(Table 1). The P value for the psaF gene was 0.06841, which wasclose to the cutoff (i.e., � 0.05), indicating that there were stillsome differences for the psaF gene under the four conditions.

Bulk-cell-based analysis showed that the psaA gene had ahigher expression level under the phosphate depletion conditionthan under the nitrogen depletion condition (see Fig. S2 in thesupplemental material). However, a reverse pattern was observedfrom the single-cell-based analysis. A similar pattern betweenbulk- and single-cell analyses was also observed for the psaF gene.For psbA genes, the nitrogen depletion condition had the lowestactivity among four growth conditions, which was only 10% ofthat under the control condition. This may be due to the insuffi-cient supply of inorganic nitrogen as found by Kolber et al. (63),who showed that nitrogen limitation could lead to substantial

FIG 4 Gene expression distributions of selected genes under four different growth conditions. P values were determined by using the nonparametric two-sampleKolmogorov-Smirnov test between each depletion and control conditions ( � 0.05). The x axis shows the relative activity of a specific gene compared with theactivity of the control actin gene in the same cell, and the y axis shows the number of cells that have the same relative activity.

TABLE 1 P values of Kruskal-Wallis tests at the 95% confidence level

Gene P value

psaA 8.23229E�15psaF 0.06841psbA 0.00255psbC 9.47652E�5hsp90 7.70247E�4rbcL 4.40972E�8

Shi et al.

1854 aem.asm.org Applied and Environmental Microbiology

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/a

em o

n 09

Feb

ruar

y 20

22 b

y 21

8.10

3.20

1.16

3.

decreases in photosynthetic energy conversion efficiency and lossof PS II protein D1, which is encoded by the psbA gene. For psbCgenes, the results from the single-cell level were consistent with theresults from bulk-cell analysis. Although the bulk-cell results in-dicated that phosphate depletion had about 2-times-higher activ-ity than under the nitrogen depletion condition for the hsp90gene, the single-cell-level results indicated that the activities weresimilar to each other, while the upregulation of the hsp90 geneunder the iron depletion condition at the bulk cell level was con-firmed by single-cell-level results which indicated that low Feavailability indeed triggered stress on T. pseudonana. For the rbcLgene, the results showed that iron depletion had no effect on rbcLexpression while nitrogen depletion affected rbcL expression,based on both single-cell- and bulk-cell-based analyses, consistentwith previous work with the marine diatom P. tricornutum (64).

PCA of single-cell RT-qPCR data. With the aid of powerfulstatistical tools, more intrinsic information can be extracted fromsingle-cell-based data sets. For instance, besides the independencetest based on results of response distributions, principal-compo-nent analysis (PCA) also can be applied to analyze the relationshipbetween different growth conditions (Fig. 5). PCA can provide asimple plot that shows the most important two factors that affectthe samples of each growth condition. PCA analysis of psaF

showed that the nitrogen depletion condition had no significanteffect on gene expression in single cells compared with the controlcondition. For psbA and rbcL, the PCA results showed that the fourgrowth conditions were distinguished from each other. These re-sults agreed well with the distribution analysis. For the psaA andhsp90 genes, the P value generated from the Kolmogorov-Smirnovtest suggested that there were no significant differences betweeniron depletion and control conditions (Fig. 4). However, based onthe PCA analysis, they had a similar score for component 1 but aslightly different score for component 2 for different nutrient de-pletion conditions, which indicated that the expression of thesetwo genes under nutrient depletion conditions was similar butdistinct from that under the control condition. In addition, forpsbC, the distribution analysis showed that iron depletion andcontrol conditions were similar to each other, but the PCA resultsshowed that expression of the psbC gene under iron depletion andcontrol conditions was not controlled in the same way.

DISCUSSION

The responses of diatoms to various nutrient deficiencies havebeen evaluated at the population level (20, 59). However, becausethey are planktonic microorganisms, the cell-cell heterogeneity ofdiatoms in terms of responses to environmental factors could besignificant and have never been documented. In this study, wemade the first attempt to measure the expression of selected genesof the model diatom T. pseudonana when they were subject tonitrogen, phosphate, and iron depletion conditions. The resultsshowed significant heterogeneity, which shed light on potentialenvironmental problems. Opposite expression patterns werefound for the psaA, psaF, psbA, and hsp90 genes between single-cell-based and bulk-cell-based analyses. The abnormal cells maybe an indicator of potential environmental problems and suggestthat further investigations would be possibly buried under theaverage value without single-cell-level analysis.

In order to use T. pseudonana as a sensor by using single-cellRT-qPCR, several issues need to be addressed. The first is thesensitivity of the sensor, which is equivalent to single-cell RT-qPCR sensitivity. Since the sensitivity of our technology can godown to a single-cell level, it has the capability to analyze someimportant and/or uncultured environmental samples.

Our results showed that as the copy number of transcripts of agene decreased, the SD of RT-PCR technical replicates increasedaccordingly. To overcome the issue, small reaction volumes,which increase the local template concentration, are preferred. Inthe study, we used 10-�l reaction volumes instead of conventional20-�l reaction volumes for qPCR. Currently, 10 �l is the smallestvolume with which we can obtain consistent and reliable results inthe tube/microtiter plate-based qPCRs. In order to further de-crease the reaction volume, chip-level devices which can decreasethe volume to several microliters (65) or even to the picoliter level(66–68) will be more attractive. In addition, we also addressed theissue of low levels of starting material by increasing the cDNAyields of specific targets through adding target-specific primers.Ståhlberg et al. evaluated 4 different primer strategies, i.e., randomhexamers, oligo(dT), gene-specific primers, and gene-specificprimer mixtures, on five different genes, and the results showedthat gene-specific primer mixtures had an overall advantage basedon the yield and SD of qPCR results for several different genes(45). In order to simplify the whole process, considering that thereverse primer of qPCR is complementary to the mRNA sequence

FIG 5 Principal-component analysis (PCA) of single-cell-based analysis ofselected genes. Component 1 and component 2 are the top two indicators thatcould explain the pattern of the gene expression data set. Dots represent eachof the conditions we tested in this study (C, control; N, nitrogen depletion; Fe,iron depletion; P, phosphate depletion). The distances between dots representsimilarity in terms of gene expression pattern between various growth condi-tions. A longer distance indicates less similarity.

Single-Cell Analysis of Thalassiosira pseudonana

March 2013 Volume 79 Number 6 aem.asm.org 1855

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/a

em o

n 09

Feb

ruar

y 20

22 b

y 21

8.10

3.20

1.16

3.

and may bind to specific mRNA during the cDNA synthesis step,which will increase the cDNA yield of specific targets, we addedthe reverse primer directly rather than using the specially designedspecific primers that are complementary to the mRNA sequence asdescribed by Ståhlberg et al. (45). The results showed that addingtarget-specific primers in the cDNA synthesis step could increasethe quality and yield of target cDNA by about 10-fold on average.

The second issue is how to interpret RT-qPCR results in aquantitative way so that the result can be used as an indicator ofenvironmental stress conditions. The use of reference genes is im-portant in order to normalize qPCR results, and much researchhas been done on the selection of reference genes for various bulk-cell-based analysis (51, 69). However, considering gene expressionstochasticity in single cells, the reliability of employing these genesfor single-cell gene expression is still unclear. In this study, weselected and validated the actin gene as an internal reference basedon its better performance than other candidate genes, and expres-sion heterogeneity of the actin gene was still observed betweenindividual T. pseudonana cells. To fully address the heterogeneityissue, an alternative internal control strategy, such as using a mol-ecule that is artificially incorporated into the sample as an RNAspike (44, 51, 70), may be necessary and worth further develop-ment.

There are still technical challenges for using microbial geneexpression at the level of a small number of cells as an environ-mental sensor. For example, the targeted microbe is in the mixturewith other microbes, yet further manipulation such as cell sortingor cultivation will alter gene expression levels. How to successfullydetermine the atypical gene expression patterns from a few cellsamong a larger number of background normal cells is another bigchallenge. To overcome this, a feasible approach is to performhigh-throughput single-cell level analysis on the microbiota andthen extract the targeted information using postprocessing on theacquired data.

Finally, although our results demonstrated that with properselection of gene targets and optimization of RT-PCR condition,gene expression measurements at single-cell resolution will allowmonitoring of the marine environmental health, possibly at anearly stage of potential environmental problems to minimize thecost of environmental remediation, currently the technologyworks well only with highly expressed genes, which limits the se-lection of gene targets. In the future, further development andoptimization of the molecular biology protocol and integrationwith chip-level real-time PCR devices (65) will generate a chip-level sensor instrument for monitoring marine environmentalhealth in a fast and effective way to overcome the remaining tech-nical challenges. At the same time, fundamental microbiologyquestions about heterogeneity within an isogenic population willbe answered as well.

ACKNOWLEDGMENTS

We thank ASU’s NEPTUNE fund for funding to Deirdre Meldrum for thesupport of this research. Weiwen Zhang is currently funded by a grantfrom the National Natural Science Foundation of China (project no.31170043).

REFERENCES1. Brehm-Stecher BF, Johnson EA. 2004. Single-cell microbiology: tools,

technologies, and applications. Microbiol. Mol. Biol. Rev. 68:538 –559.2. Raser JM, O’Shea EK. 2005. Noise in gene expression: origins, conse-

quences, and control. Science 309:2010 –2013.

3. Banerjee B, Balasubramanian S, Ananthakrishna G, Ramakrishnan TV,Shivashankar GV. 2004. Tracking operator state fluctuations in gene ex-pression in single cells. Biophys. J. 86:3052–3059.

4. Bengtsson M, Hemberg M, Rorsman P, Ståhlberg A. 2008. Quantifica-tion of mRNA in single cells and modelling of RT-qPCR induced noise.BMC Mol. Biol. 9:63.

5. Blake WJ, Kaern M, Cantor CR, Collins JJ. 2003. Noise in eukaryoticgene expression. Nature 422:633– 637.

6. Colman-Lerner A, Gordon A, Serra E, Chin T, Resnekov O, Endy D,Pesce CG, Brent R. 2005. Regulated cell-to-cell variation in a cell-fatedecision system. Nature 437:699 –706.

7. Elowitz MB, Levine AJ, Siggia ED, Swain PS. 2002. Stochastic geneexpression in a single cell. Science 297:1183–1186.

8. Hasty J, Pradines J, Dolnik M, Collins JJ. 2000. Noise-based switchesand amplifiers for gene expression. Proc. Natl. Acad. Sci. U. S. A. 97:2075–2080.

9. McAdams HH, Arkin A. 1997. Stochastic mechanisms in gene expres-sion. Proc. Natl. Acad. Sci. U. S. A. 94:814 – 819.

10. Newman JRS, Ghaemmaghami S, Ihmels J, Breslow DK, Noble M,DeRisi JL, Weissman JS. 2006. Single-cell proteomic analysis of S. cerevi-siae reveals the architecture of biological noise. Nature 441:840 – 846.

11. Pedraza JM, van Oudenaarden A. 2005. Noise propagation in genenetworks. Science 307:1965–1969.

12. Rosenfeld N, Young JW, Alon U, Swain PS, Elowitz MB. 2005. Generegulation at the single-cell level. Science 307:1962–1965.

13. Strovas TJ, Sauter LM, Guo X, Lidstrom ME. 2007. Cell-to-cell hetero-geneity in growth rate and gene expression in Methylobacteriumextorquens AM1. J. Bacteriol. 189:7127–7133.

14. Swain PS, Elowitz MB, Siggia ED. 2002. Intrinsic and extrinsic contri-butions to stochasticity in gene expression. Proc. Natl. Acad. Sci. U. S. A.99:12795–12800.

15. Lidstrom ME, Meldrum DR. 2003. Life-on-a-chip. Nat. Rev. Microbiol.1:158 –164.

16. Walling MA, Shepard JRE. 2011. Cellular heterogeneity and live cellarrays. Chem. Soc. Rev. 40:4049 – 4076.

17. Lenz AP, Williamson KS, Pitts B, Stewart PS, Franklin MJ. 2008.Localized gene expression in Pseudomonas aeruginosa biofilms. Appl. En-viron. Microbiol. 74:4463– 4471.

18. Rajilic-Stojanovic M, Smidt H, De Vos WM. 2007. Diversity of thehuman gastrointestinal tract microbiota revisited. Environ. Microbiol.9:2125–2136.

19. Falkowski PG, Katz ME, Knoll AH, Quigg A, Raven JA, Schofield O,Taylor FJR. 2004. The evolution of modern eukaryotic phytoplankton.Science 305:354 –360.

20. Thamatrakoln K, Korenovska O, Niheu AK, Bidle KD. 2012. Whole-genome expression analysis reveals a role for death-related genes in stressacclimation of the diatom Thalassiosira pseudonana. Environ. Microbiol.14:67– 81.

21. Bennett JR, Cumming BF, Ginn BK, Smol JP. 2010. Broad-scale envi-ronmental response and niche conservatism in lacustrine diatom commu-nities. Global Ecol. Biogeogr. 19:724 –732.

22. Mann D, Droop S. 1996. Biodiversity, biogeography and conservation ofdiatoms. Hydrobiologia 336:19 –32.

23. Maheswari U, Jabbari K, Petit JL, Porcel BM, Allen AE, Cadoret JP, DeMartino A, Heijde M, Kaas R, La Roche J, Lopez PJ, Martin-Jézéquel V,Meichenin A, Mock T, Schnitzler Parker M, Vardi A, Armbrust EV,Weissenbach J, Katinka M, Bowler C. 2010. Digital expression profilingof novel diatom transcripts provides insight into their biological func-tions. Genome Biol. 11:R85. doi:10.1186/gb-2010-11-8-r85.

24. Nelson D, Treguer P, Brzezinski M, Leynaert A, Queguiner B. 1995.Production and dissolution of biogenic silica in the ocean-revised globalestimates, comparison with regional data and relationship to biogenicsedimentation. Glob. Biogeochem. Cycle 9:359 –372.

25. Ripp S, Nivens DE, Ahn Y, Werner C, Jarrell J, Easter JP, Cox CD,Burlage RS, Sayler GS. 2000. Controlled field release of a bioluminescentgenetically engineered microorganism for bioremediation process moni-toring and control. Environ. Sci. Technol. 34:846 – 853.

26. Hockin NL, Mock T, Mulholland F, Kopriva S, Malin G. 2012. Theresponse of diatom central carbon metabolism to nitrogen starvation isdifferent from that of green algae and higher plants. Plant Physiol. 158:299 –312.

27. Bucciarelli E, Sunda WG. 2003. Influence of CO2, nitrate, phosphate,and silicate limitation on intracellular dimethylsulfoniopropionate in

Shi et al.

1856 aem.asm.org Applied and Environmental Microbiology

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/a

em o

n 09

Feb

ruar

y 20

22 b

y 21

8.10

3.20

1.16

3.

batch cultures of the coastal diatom Thalassiosira pseudonana. Limnol.Oceanogr. 48:2256 –2265.

28. Rao IM, Arulanantham AR, Terry N. 1989. Leaf phosphate status, pho-tosynthesis and carbon partitioning in sugar beet. II. Diurnal changes insugar phosphates, adenylates, and nicotinamide nucleotides. PlantPhysiol. 90:820 – 826.

29. Abdel-Ghany SE, Ye H, Garifullina GF, Zhang L, Pilon-Smits EAH,Pilon M. 2005. Iron-sulfur cluster biogenesis in chloroplasts. Involvementof the scaffold protein CpIscA. Plant Physiol. 138:161–172.

30. Boyd PW, Jickells T, Law CS, Blain S, Boyle EA, Buesseler KO, CoaleKH, Cullen JJ, de Baar HJW, Follows M, Harvey M, Lancelot C,Levasseur M, Owens NPJ, Pollard R, Rivkin RB, Sarmiento J, Schoe-mann V, Smetacek V, Takeda S, Tsuda A, Turner S, Watson AJ. 2007.Mesoscale iron enrichment experiments 1993-2005: synthesis and futuredirections. Science 315:612– 617.

31. Lewandowska J, Kosakowska A. 2004. Effect of iron limitation on cells ofthe diatom Cyclotella meneghiniana Kützing. Oceanologia 46:269 –287.

32. Martin J, Coale K, Johnson K, Fitzwater S, Gordon R, Tanner S, HunterC, Elrod V, Nowicki J, Coley T, Barber R, Lindley S, Watson A, VanscoyK, Law C, Liddicoat M, Ling R, Stanton T, Stockel J, Collins C,Anderson A, Bidigare R, Ondrusek M, Latasa M, Millero F, Lee K, YaoW, Zhang J, Friederich G, Sakamoto C, Chavez F, Buck K, Kolber Z,Greene R, Falkowski P, Chisholm S, Hoge F, Swift R, Yungel J, TurnerS, Nightingale P, Hatton A, Liss P, Tindale N. 1994. Testing the ironhypothesis in ecosystems of the equatorial Pacific-ocean. Nature 371:123–129.

33. Johnson KS, Gordon RM, Coale KH. 1997. What controls dissolved ironconcentrations in the world ocean? Mar. Chem. 57:137–161.

34. Marchetti A, Schruth DM, Durkin CA, Parker MS, Kodner RB, Ber-thiaume CT, Morales R, Allen AE, Armbrust EV. 2012. Comparativemetatranscriptomics identifies molecular bases for the physiological re-sponses of phytoplankton to varying iron availability. Proc. Natl. Acad.Sci. U. S. A. 109:E317–E325.

35. Armbrust EV, Berges JA, Bowler C, Green BR, Martinez D, PutnamNH, Zhou S, Allen AE, Apt KE, Bechner M, Brzezinski MA, Chaal BK,Chiovitti A, Davis AK, Demarest MS, Detter JC, Glavina T, GoodsteinD, Hadi MZ, Hellsten U, Hildebrand M, Jenkins BD, Jurka J, Kapi-tonov VV, Kröger N, Lau WWY, Lane TW, Larimer FW, Lippmeier JC,Lucas S, Medina M, Montsant A, Obornik M, Parker MS, Palenik B,Pazour GJ, Richardson PM, Rynearson TA, Saito MA, Schwartz DC,Thamatrakoln K, Valentin K, Vardi A, Wilkerson FP, Rokhsar DS.2004. The genome of the diatom Thalassiosira pseudonana: ecology, evo-lution, and metabolism. Science 306:79 – 86.

36. Zeng J, Wang J, Gao W, Mohammadreza A, Kelbauskas L, Zhang W,Johnson RH, Meldrum DR. 2011. Quantitative single-cell gene expres-sion measurements of multiple genes in response to hypoxia treatment.Anal. Bioanal. Chem. 401:3–13.

37. Gao W, Zhang W, Meldrum DR. 2011. RT-qPCR based quantitativeanalysis of gene expression in single bacterial cells. J. Microbiol. Methods85:221–227.

38. Guillard RRL. 1975. Culture of marine invertebrates animals, p 29 – 60.Plenum Publishing Corp., New York, NY.

39. Guillard RRL. 1962. Studies of marine planktonic diatoms. I. Cyclotellanana Hustedt and Detonula confervacea Cleve. Can. J. Microbiol. 8:229 –239.

40. Kester DR, Duedall IW, Connors DN, Pytkowicz RM. 1967. Preparationof artificial seawater. Limnol. Oceanogr. 12:176 –179.

41. Anis Y, Holl M, Meldrum D. 2010. Automated selection and placementof single cells using vision-based feedback control. IEEE Trans. Automat.Sci. Eng. 7:598 – 606.

42. Anis Y, Houkal J, Holl M, Johnson R, Meldrum D. 2011. Diaphragmpico-liter pump for single-cell manipulation. Biomed. Microdevices 13:651– 659.

43. Siegel S. 1957. Nonparametric statistics. Am. Statistician 11:13–19.44. Bustin SA, Nolan T. 2004. Pitfalls of quantitative real-time reverse-

transcription polymerase chain reaction. J. Biomol. Technol. 15:155–166.45. Ståhlberg A, Håkansson J, Xian X, Semb H, Kubista M. 2004. Properties

of the reverse transcription reaction in mRNA quantification. Clin. Chem.50:509 –515.

46. Schmid A, Kortmann H, Dittrich PS, Blank LM. 2010. Chemical andbiological single cell analysis. Curr. Opin. Biotechnol. 21:12–20.

47. Valente V, Teixeira SA, Neder L, Okamoto OK, Oba-Shinjo SM, MarieSKN, Scrideli CA, Paçó-Larson ML, Carlotti Jr, CG. 2009. Selection of

suitable housekeeping genes for expression analysis in glioblastoma usingquantitative RT-PCR. BMC Mol. Biol. 10:17.

48. Heid CA, Stevens J, Livak KJ, Williams PM. 1996. Real time quantitativePCR. Genome Res. 6:986 –994.

49. Bustin SA. 2002. Quantification of mRNA using real-time reverse tran-scription PCR (RT-PCR): trends and problems. J. Mol. Endocrinol. 29:23–39.

50. Bustin SA, Benes V, Garson JA, Hellemans J, Huggett J, Kubista M,Mueller R, Nolan T, Pfaffl MW, Shipley GL, Vandesompele J, WittwerCT. 2009. The MIQE guidelines: minimum information for publication ofquantitative real-time PCR experiments. Clin. Chem. 55:611– 622.

51. Huggett J, Dheda K, Bustin S, Zumla A. 2005. Real-time RT-PCRnormalisation; strategies and considerations. Genes Immun. 6:279 –284.

52. Livak KJ, Schmittgen TD. 2001. Analysis of relative gene expression datausing real-time quantitative PCR and the 2(�Delta Delta C(T)) method.Methods 25:402– 408.

53. Pfaffl MW. 2001. A new mathematical model for relative quantification inreal-time RT-PCR. Nucleic Acids Res. 29:e45. doi:10.1093/nar/29.9.e45.

54. Goidin D, Mamessier A, Staquet MJ, Schmitt D, Berthier-Vergnes O.2001. Ribosomal 18S RNA prevails over glyceraldehyde-3-phosphate de-hydrogenase and beta-actin genes as internal standard for quantitativecomparison of mRNA levels in invasive and noninvasive human mela-noma cell subpopulations. Anal. Biochem. 295:17–21.

55. Kim B-R, Nam H-Y, Kim S-U, Kim S-I, Chang Y-J. 2003. Normalizationof reverse transcription quantitative-PCR with housekeeping genes inrice. Biotechnol. Lett. 25:1869 –1872.

56. Nailis H, Coenye T, Van Nieuwerburgh F, Deforce D, Nelis HJ. 2006.Development and evaluation of different normalization strategies for geneexpression studies in Candida albicans biofilms by real-time PCR. BMCMol. Biol. 7:25.

57. Kustka AB, Allen AE, Morel FMM. 2007. Sequence analysis and tran-scriptional regulation of iron acquisition gens in two marine diatoms. J.Phycol. 43:715–729.

58. Allen AE, LaRoche J, Maheswari U, Lommer M, Schauer N, Lopez PJ,Finazzi G, Fernie AR, Bowler C. 2008. Whole-cell response of the pen-nate diatom Phaeodactylum tricornutum to iron starvation. Proc. Natl.Acad. Sci. U. S. A. 105:10438 –10443.

59. Mock T, Samanta MP, Iverson V, Berthiaume C, Robison M, Holter-mann K, Durkin C, Bondurant SS, Richmond K, Rodesch M, Kallas T,Huttlin EL, Cerrina F, Sussman MR, Armbrust EV. 2008. Whole-genome expression profiling of the marine diatom Thalassiosira pseudo-nana identifies genes involved in silicon bioprocesses. Proc. Natl. Acad.Sci. U. S. A. 105:1579 –1584.

60. Bowler C, Allen AE, Badger JH, Grimwood J, Jabbari K, Kuo A,Maheswari U, Martens C, Maumus F, Otillar RP, Rayko E, Salamov A,Vandepoele K, Beszteri B, Gruber A, Heijde M, Katinka M, Mock T,Valentin K, Verret F, Berges JA, Brownlee C, Cadoret JP, Chiovitti, A,Choi CJ, Coesel S, De Martino A, Detter JC, Durkin C, Falciatore A,Fournet J, Haruta M, Huysman MJJ, Jenkins BD, Jiroutova K, Jor-gensen RE, Joubert Y, Kaplan A, Kröger N, Kroth PG, La Roche J,Lindquist E, Lommer M, Martin-Jézéquel V, Lopez PJ, Lucas S, Man-gogna M, McGinnis K, Medlin LK, Montsant A, Oudot-Le Secq MP,Napoli C, Obornik M, Parker MS, Petit JL, Porcel BM, Poulsen N,Robison M, Rychlewski L, Rynearson TA, Schmutz J, Shapiro H, SiautM, Stanley M, Sussman MR, Taylor AR, Vardi A, von Dassow P,Vyverman W, Willis A, Wyrwicz LS, Rokhsar DS, Weissenbach J,Armbrust EV, Green BR, Van de Peer Y, de Grigoriev IV. 2008. ThePhaeodactylum genome reveals the evolutionary history of diatom ge-nomes. Nature 456:239 –244.

61. Bengtsson M, Ståhlberg A, Rorsman P, Kubista M. 2005. Gene expres-sion profiling in single cells from the pancreatic islets of Langerhans re-veals lognormal distribution of mRNA levels. Genome Res. 15:1388 –1392.

62. Schmelz M, Schmidt R, Weidner C, Hilliges M, Torebjork HE,Handwerker HO. 2003. Chemical response pattern of different classes ofC-nociceptors to pruritogens and algogens. J. Neurophysiol. 89:2441–2448.

63. Kolber Z, Zehr J, Falkowski P. 1988. Effects of growth irradiance andnitrogen limitation on photosynthetic energy conversion in photosystemII. Plant Physiol. 88:923–929.

64. Greene RM, Geider RJ, Falkowski PG. 1991. Effect of iron limitation onphotosynthesis in a marine diatom. Limnol. Oceanogr. 36:1772–1782.

65. Shi X, Lin LI, Chen SY, Chao SH, Zhang W, Meldrum DR. 2011.

Single-Cell Analysis of Thalassiosira pseudonana

March 2013 Volume 79 Number 6 aem.asm.org 1857

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/a

em o

n 09

Feb

ruar

y 20

22 b

y 21

8.10

3.20

1.16

3.

Real-time PCR of single bacterial cells on an array of adhering droplets.Lab Chip 11:2276 –2281.

66. Beer NR, Hindson BJ, Wheeler EK, Hall SB, Rose KA, Kennedy IM,Colston BW. 2007. On-chip, real-time, single-copy polymerase chainreaction in picoliter droplets. Anal. Chem. 79:8471– 8475.

67. Kiss MM, Ortoleva-Donnelly L, Beer NR, Warner J, Bailey CG, ColstonBW, Rothberg JM, Link DR, Leamon JH. 2008. High-throughput quan-titative polymerase chain reaction in picoliter droplets. Anal. Chem. 80:8975– 8981.

68. Lindström S, Andersson-Svahn H. 2010. Overview of single-cell analyses:microdevices and applications. Lab Chip 10:3363–3372.

69. Czechowski T, Stitt M, Altmann T, Udvardi MK, Scheible W-R.2005. Genome-wide identification and testing of superior referencegenes for transcript normalization in Arabidopsis. Plant Physiol. 139:5–17.

70. Bower NI, Moser RJ, Hill JR, Lehnert SA. 2007. Universal referencemethod for real-time PCR gene expression analysis of preimplantationembryos. Biotechniques 42:199 –206.

Shi et al.

1858 aem.asm.org Applied and Environmental Microbiology

Dow

nloa

ded

from

http

s://j

ourn

als.

asm

.org

/jour

nal/a

em o

n 09

Feb

ruar

y 20

22 b

y 21

8.10

3.20

1.16

3.