Comparison of RNA yield from small cell populations sorted by flow cytometry applying different isolation procedures




RNA from sorted cell populations is crucial in many instances. We therefore compared four current protocols for RNA isolation, with regard to mRNA yield and purity. Moreover, we examined the effects on RNA recovery caused by different storage reagents.


Small populations of K562 cells or PMBC were sorted into the lysing reagent and subjected to RNA extraction, employing either phase separation extraction using an acidic guanidinium–isothiocyanate reagent (TriFast™ reagent), the silica-gel membrane-based spin-column technology (RNeasy Mini-/Micro-Kit™), or the isolation via paramagnetic oligo(d)T-beads (μMACS™). Cells designated for delayed RNA isolation were kept either in RNAlater™, Qiagen Buffer RLT™, TriFast™ or PrepProtect™, or simply frozen after pelleting from PBS. The mRNA yield was determined by quantitative RT-PCR.


Performing unpaired two-tailed t-tests revealed that RNA was extracted in significantly higher amounts using magnetic bead isolation. This method also allowed best discrimination of induced IL2 gene expression. In contrast, phase separation extraction showed the highest rate of failures. Intermediate storage reduced RNA yield. Contamination by genomic DNA was detected in several samples subjected to phase separation or silica-gel membrane-based spin-column extraction.


Our results reveal advantages and disadvantages of RNA isolation procedures for small numbers of sorted cells and, therefore, facilitate the decision for the most appropriate protocol in a particular experimental context. © 2007 International Society for Analytical Cytology

An important application of flow cytometry-based cell sorting is the generation of distinct sample populations of usually rather limited size. Effective processing of the rare material is highly desired to gain valid experimental results. Downstream analyses frequently involve examinations of gene expression profiles, i.e., they use sorted cells as a source of mRNA. However, RNA is very sensitive to degradation by ubiquitous RNases. In case of sorted cells, this problem is further enhanced due to shearing forces that stress cells when they pass the flow cytometer nozzle. Elaborate procedures have been developed for RNA isolation, but nevertheless, working with it is a practical challenge. Protocols for RNA extraction in special experimental contexts have been evaluated concerning their efficiency in each case (1, 2), and optimized protocols for distinct purposes have been published recently (3, 4). In this study, we compared four current standard protocols for RNA isolation conducted directly after cell sorting, with regard to yield and purity of mRNA of a housekeeping gene. Furthermore, using the IL2 gene as a model, we investigated whether the expression rates determined for an inducible gene vary depending on the RNA extraction procedure. Finally, we wanted to know how amount and quality of the recovered mRNA are influenced by the effects of different stabilization reagents. The techniques applied were phase separation extraction via an acidic guanidinium-isothiocyanate (GITC)-containing extraction reagent (PSG), the silica-gel membrane-based spin-column technology (SSC) and magnetic bead isolation followed by on-column cDNA synthesis (MBI).

To avoid a methodical bias due to existing technical knowledge, all experiments in this investigation were carried out by an advanced graduate student who had not received special training in any of the protocols applied. Moreover, we performed the whole set of procedures on housekeeping gene mRNA detection, separately on cultured cells (K562) and primary cells (PBMC), to take into account biological differences between cell types that are likely to influence RNA yield.


Cell Preparation and Flow Cytometry-Based Sorting

K562 cells harvested by centrifugation or PBMC separated from whole blood on a Ficoll-gradient (Amersham, Uppsala, Sweden) were resuspended in PBS and sorted into the lysing- or storage reagent, as described later, with the MoFlo™ flow cytometer (DakoCytomation, Ft. Collins, CO). For the sorting procedure, cells were gated with a combined live gate, according to scatter characteristics and dead cell exclusion via DAPI stain at 100 ng/ml final concentration (Sigma, Taufkirchen, Germany). T-cells were discriminated via CD3 (clone SK7) and CD45 antibody surface staining (clone 2D1; both obtained from BD Biosciences, San Jose, CA), and sorting was performed in 1.5-ml plastic tubes at room temperature without external cooling.

For each of the two cell species, 10 series of samples, including doubles of each 500, 1,000, 5,000, and 10,000 cytometrically counted cells and a single reference of 100,000 cells, were created. For direct and delayed PSG, cells were sorted into 500 μl of GITC buffer. In SSC, 350 and 100 μl of buffer RLT™ supplied with the two variants of the SSC kit were used when cells were immediately processed. We also employed 350 μl of this buffer when we tested it as a storage medium. In parallel, we prepared cell samples in 500 μl of RNAlater™ (Ambion, Austin, TX) and 500 μl of PBS from where cells were pelleted prior to freezing for delayed RNA extraction, applying SSC. For MBI, we sorted cells into 500 μl of lysis-/binding buffer and 300 μl of the supplier's RNA protection reagent, respectively. For analysis of IL2 gene induction, 5 ng/μl PMA and 0.5 μg/ml ionomycin (both from Sigma) were added to heparinized whole blood and incubated at 37°C. After 2 h, 1 μg/μl GolgiPlug™ (BD Biosciences) was added. Stimulation was stopped after 6 h by addition of 50 μl/ml 20 mM EDTA (pH 7.2) to the samples and 15-min incubation at room temperature. PBMC were separated by density gradient centrifugation, and T-lymphocytes were sorted out by flow cytometry as described. The sample series created of stimulated and unstimulated T-lymphocytes included populations of 100, 1,000, 10,000, and 100,000 cells, and were subjected only to direct RNA isolation procedures.

Sample Storage, RNA Isolation, and cDNA Synthesis

PSG was conducted employing the peqGold TriFast™ reagent (Peqlab, Erlangen, Germany), SSC via the Qiagen RNeasy™ Mini and Micro Kit™ (Qiagen, Valencia, CA), and MBI using the μMACS™ technology (Miltenyi Biotech, Bergisch-Gladbach, Germany). Delayed application of the last method employed Miltenyi's PrepProtect™ RNA stabilization reagent. K562 cells and PBMC designated for delayed RNA preparation were stored for 1 week at −80 or −20°C, respectively, when RNAlater™ was used as a storage reagent. All RNA extraction procedures followed manufacturers' instructions. In both variants of SSC, the homogenization step was skipped due to the small cell number, and the on-column DNaseI digestion step was included. In case of MBI, cDNA synthesis was performed in one step directly on the μMACS™ columns, all other RNA preparations were subjected to reverse transcription, employing the Sensiscript-RT Kit™ (Qiagen). Cells were recovered from RNAlater™ by 5-min centrifugation at 3,000g and aspiration of the supernatant. Similarly, cells were prepared for RNA isolation from PrepProtect™ by centrifugation at 5,000g for 3 min.

Quantitative Real Time and Standard PCR Applications

To determine the mRNA yield, 1 μl of cDNA obtained from K562 or T-cell extracts, and 2 μl of PBMC derived cDNA or of selected RNA solutions were subjected to quantitative PCR analysis on the ABIPrism 7700 sequence detector™ (Applied Biosystems, Weiterstadt, Germany). Volumes of cDNAs synthesized in one step throughout MBI were equivalently risen. Reactions were set up using the QuantiTect® SYBR®Green PCR Kit™ (Qiagen) and the following pair of primers targeting GAPDH: (GAPDH-1: 5′-CTC CTC CAC CTT TGA CGC TG-3′, GAPDH-2: 5′-ACC ACC CTG TTG CTG TAG CC-3′). IL2 gene expression was determined by amplifying the target gene using the primers IL2 fw (5′-GGA TGC AAC TCC TGT CTT GC-3′) and IL2 rv (5′-GTG GCC TTC TTG GGC ATG TA -3′), and the normalizer gene GAPDH employing the primer pair GAPDH-1 and GAPDH-3 (5′-TCC TCT TGT GCT CTT GCT GG-3′). The temperature profile consisted of a 15-min denaturation step at 95°C and 45 cycles, including 15 s of denaturation at 94°C, 30 s annealing at 60°C, and 30 s extension at 72°C. Melting curves were measured subsequently. Each cDNA or RNA sample, no-template-control and as a calibrator 1 μl of a standard K562 cDNA, was run in triplicates (T-cell cDNA in duplicates). In standard PCR reactions, a β-actin fragment was amplified using the Ampli Taq®Polymerase™ (Applied Biosystems) and the primers β-actin-1 (5′-CCT TCC TGG GCA TGG AGT CCT-3′) and β-actin-3 (5′-AAT CTC ATC TTG TTT TCT GCG-3′), which span two introns of 207 bp total length. Here, 40 amplification cycles with an annealing temperature of 55°C were run on the PE 9600 thermal cycler (Applied Biosystems).

Statistical Data Analysis and Data Interpretation

To calculate CT values, the threshold was adjusted to the same value for all quantitative PCR assays. Samples showing either an ambiguous melting curve or a CT value above 40 were considered empty and not taken into statistical analyses. Relative IL2 expression levels were calculated using the ΔΔCT method. Cells were grouped into populations consisting of very low (100/500–1,000), low (5,000–10,000), and high (100,000) numbers of K562 cells or PBMC. The different RNA isolation techniques were compared performing unpaired two-tailed t-tests. The level of significance was set to P ≤ 0.05.


MBI Technique and One-Step cDNA Synthesis Is the Most Reliable RNA Extraction Method

In order to assess the overall reliability of the different RNA isolation procedures, we calculated the share of successfully extracted samples for each technique. It revealed that only when employing MBI, RNA could be recovered from all populations of cells (Table 1). Using SSC, 100% success could only be gained for K562 in direct application, but when SSC was applied to PBMC, almost half of quantitative PCR assays gave invalid data. PSG-treated samples showed the highest rate of failures, i.e. no cDNA signal could be detected in 11–67% of samples in direct application and 37–89% with delayed extraction. Table 1 gives a complete overview of the success rates we achieved conducting different RNA extraction protocols to small populations of sorted cells.

Table 1. Successful RNA Extraction from K562 cells and PBMC Applying Different Protocols
RNA extraction procedureK562PBMC
Total qPCR assaysExcludedData yield (%)Total qPCR assaysExcludedData yield (%)
  1. RNA was isolated successfully more frequently from K562 cells than from PBMC, as indicated by quantitative PCR assays targeting the houskeeping gene GAPDH. MBI proved to be the most reliable procedure for direct or delayed RNA extraction. In contrast, amplification products were detected in lowest frequencies employing PSG technique.

  2. PSG, phase separation extraction via an acidic guanidinium–isothiocyanate (GITC)-containing extraction reagent; SSC, silica-gel membrane-based spin-column technology; MBI, magnetic bead isolation followed by on-column cDNA synthesis.

PSG direct27389271833
SSC (Mini™ kit) direct270100271159
SSC(Micro™ kit) direct270100271256
MBI direct210100210100
Buffer RLT™->SSC2729327774

High RNA Yield Is Obtained by Direct Application of MBI and SSC

To determine the efficiency of RNA extraction, quantitative real time PCR was performed, because low CT values correlate with high RNA yield. The housekeeping gene GAPDH was amplified, and mean CT values were measured for each method.

We performed this analysis separately for K562 and PBMC populations containing very low (500–1,000), low (5,000–10,000), or high (100,000) numbers of cells. In direct application, we found that, in very small and small populations of both cell types, RNA yield using PSG was significantly lower than the amount of RNA gained with SSC or MBI, as indicated by a higher CT value of PSG-treated samples (Figs. 1a and 1b). MBI was also significantly more effective than SSC in all groups of cells, except for very small numbers of PBMC. Here, we did not find a significant advantage of magnetic beads over the Mini™ version of SSC (Fig. 1b). Comparison of the two variants of SSC (RNeasyMini™ versus RNeasyMicro™ Kit) suggests equal power of both methods (Figs. 1a and 1b). However, even for very small numbers of cells, with every assay, RNA yield was high enough to produce distinct bands in standard PCR analysis targeting β-actin (Fig. 1c). In summary, these observations suggest that basically all procedures applied in this study are suitable for RNA extraction from small numbers of cells, yet there is a considerable variation in RNA yield, depending on the method applied.

Figure 1.

RNA yield from different isolation techniques applied directly after cell sorting. RNA yield was determined by quantitative RT-PCR targeting GAPDH. The graph shows mean CT values calculated from triplicate quantitative PCR assays of each sample for very low (500–1,000>), low (5,000–10,000), and high (100,000) numbers of (a) K562 cells and (b) PBMC. Samples showing CT values above 40 or melting curves indicating contamination by genomic DNA were excluded from the analysis. Error bars represent SD. Compared to the other techniques applied, RNA could be extracted in significantly higher amounts (P ≤ 0.05) using MBI. ▪ indicates the only lack of significance. (c) β-actin standard PCR confirms efficient RNA isolation. Five hundred K562 cells or a population of total PBMC were processed directly after sorting, using different RNA extraction procedures. The obtained RNA was reversely transcribed, and the cDNA products were subsequently subjected to standard PCR targeting β-actin. PSG, phase separation extraction via an acidic GITC-containing extraction reagent; SSC, silica-gel membrane-based spin-column technology, MBI, magnetic bead isolation; NTC, no-template-control.

MBI Allows Best Discrimination of Induced IL2 Gene Expression

We determined the rates of IL2 expression relative to GAPDH in PMA-stimulated and untreated T–lymphocytes, applying different RNA isolation procedures (Fig. 2). Mean expression rates were calculated from CT values gained in qPCR assays for sample sets, including populations of 100, 1,000, and 10,000 T-lymphocytes. Notably, in the smallest population of unstimulated cells, IL2 mRNA could only be detected in SSC Micro™ processed samples. In stimulated cells, highest levels of IL2 were yielded from cDNAs prepared by MBI, and expression rates obtained for the PSG- or SSC-derived cDNAs showed less variation between the methods. The two versions of the SSC technique also appeared to be of equal power concerning induced gene expression analysis. Taken together, our findings indicate that all methods applied are suitable to detect changes in the expression rates of an inducible gene. Yet our data also suggest that best discrimination is achieved using the MBI procedure.

Figure 2.

Relative expression levels of IL2 in unstimulated and PMA-stimulated T-lymphocytes, using different RNA extraction procedures. Populations of 100, 1,000, 10,000, and 100,000 PMA stimulated or untreated T-lymphocytes were sorted out from total PBMC by flow cytometry and subjected to PSG, a variant of SSC and MBI. Samples were assayed for IL2 and GAPDH by quantitative PCR and relative IL2 levels were calculated for each cell number using the ΔΔCT method. The graph shows the mean IL2 expression rates for the pooled fractions of 100–10,000 cells. Error bars represent SD. All methods allow clear discrimination of induced and baseline gene expression rates, with MBI showing a tendency to yield the highest IL2 levels in stimulated cells and best discrimination between the stimulated and unstimulated condition. PSG, phase separation extraction via an acidic GITC-containing extraction reagent; SSC, silica-gel membrane-based spin-column technology; MBI, magnetic bead isolation

Intermediate Storage Decreases RNA Yield

Since immediate RNA isolation from sorted cells is not always possible, we investigated the effects on RNA yield caused by the application of different storage reagents. RNA yield was generally lower in procedures including storage than for direct applications. For PSG, the differences of immediate RNA isolation compared to delayed extraction after sample storage in GITC-buffer were not significant with both cell types. In contrast we found that freezing cells in the corresponding SSC lysis buffer led to significant reduction of RNA yield in K562 cells as well as PBMC. Prior to RNA isolation on the magnetic column when cells had been kept in the manufacturer's RNA protection reagent, a significantly lower efficiency was only observed with K562 cells (Figs. 3a and 3b).

Figure 3.

RNA yield from different isolation techniques after intermediate storage of cell samples. K562 cells or PBMC were sorted into different RNA storage reagents and subjected to different RNA extraction procedures 1 week later. RNA yield was determined by quantitative RT-PCR targeting GAPDH. Mean CT values were calculated from triplicate quantitative PCR assays of each sample for the complete set of sorted (a) K562 or (b) PBMC samples subjected to the same extraction procedure. Significant (P ≤ 0.05) differences in RNA recovery for direct and delayed application of a distinct RNA extraction procedure are indicated by *. Mean CT values from triplicate quantitative PCR assays of each sample were calculated for very low (500–1,000) and low (5,000–10,000) numbers of (c) K562 cells and (d) PBMC. Error bars represent SD. * indicates significantly higher RNA yield from MBI than from SSC after intermediate storage in the proprietary lysis buffer. PSG, phase separation extraction via an acidic GITC-containing extraction reagent; SSC, silica-gel membrane-based spin-column technology; MBI, magnetic bead isolation.

Comparing several conditions of cell storage after sorting, we found that the different modes of preservation were almost equal for very low numbers of both cell types. However, in small cell populations, RNA yield was higher with PrepProtect (Figs. 3c and 3d).

Taken together, our data indicate that isolation of RNA from small numbers of cells should directly follow flow cytometric cell sorting when maximum yield is desired. If this is not possible, RNA loss can be limited to a certain degree, depending on the storing reagent and extraction protocol applied for delayed RNA isolation.

The Phase Separation Extraction and Spin Column Methods Are Susceptible to Contamination by Genomic DNA

In order to asses the purity of differently prepared RNA solutions, we performed quantitative PCR assays on selected samples of RNA extracts and subjected these products to agarose gel electrophoresis. Using intron spanning primers allowed distinguishing genomic DNA amplificates from the desired cDNA product upon gel visualization as well as melting curve analysis. Notably, genomic DNA was detected in several PSG- and SSC-treated samples, despite inclusion of a DNase digestion step in the spin-column procedure even with very low numbers of cells (data not shown). In contrast, genomic DNA amplification was detected in none of the samples that had been processed by MBI. Table 2 lists the samples processed via PSG or a variant of SSC that contained genomic DNA. Taken together, these findings point out one-step on-column cDNA synthesis as the most suitable procedure to gain highly pure cDNA.

Table 2. Content of Genomic DNA in Differently Prepared RNA and cDNA Solutions
RNA extraction procedureTotal qPCR assaysPercentage of samples with detectable genomic DNA content (%)
  1. Applying PSG or SSC techniques, up to 50% of RNA and up to 25% of cDNA solutions analysed by standard or quantitative PCR showed contamination with genomic DNA.

  2. PSG, phase separation extraction via an acidic guanidinium–isothiocyanate (GITC)-containing extraction reagent; SSC, silica-gel membrane-based spin-column technology using either the Qiagen RNeasy™ Mini or Micro kit.

SSC (Mini™)522161310
SSC (Micro™)16542511


In our evaluation study, we compared four current protocols for RNA isolation—the phase separation extraction via an acidic GITC-containing extraction reagent, two alternative variants of the SSC and MBI followed by on-column cDNA synthesis with regard to quantity and quality of RNA and cDNA, recovered from small populations of sorted cells.

Our results suggest that MBI most closely meets the criteria of a universally applicable protocol for this purpose: It is highly reliable when directly employed as well as after intermediate sample storage and yields large amounts of pure product. However, currently available technical equipment allows parallel processing of only four samples. Additionally, it is important to state that this magnetic technology is strictly limited to mRNA isolation. It is also the most expensive method investigated in our study.

In contrast, the phase separation method is the cheapest procedure among all techniques and is also the state of the art protocol if microRNA processing is required (5–7). However, our results point out minor suitability of the method with regard to reliability and yield of RNA. Nevertheless, it may be highly effective in the hands of “trained” users.

Yet, the PSG protocol is rather time-consuming compared to the SSC technique. Therefore, SSC presents the method of choice, if high sample throughput is designated to be combined with easy handling. In this context, it is important to state that without extensive practice of the protocol, both variants of SSC we tested are equivalently effective. Moreover, the SSC principle is highly susceptible to co-isolate genomic DNA as a side product.

With regard to expression levels of an inducible gene, our findings indicate a tendency for MBI to yield higher expression rates upon induction than the other procedures applied. However, all other RNA isolation protocols used in this study also allowed detection of remarkable changes of target gene levels in response to stimulation.

Concerning the best conditions for storing cells prior to RNA extraction, our findings suggest that isolation should directly follow sample preparation. Thus, the critical steps of cell recovery from a dense protection medium are avoided. If direct sample processing is not possible, sorting the cells into the lysing reagent or application of the PrepProtect reagent appears to be the best choice.

In conclusion, our findings illustrate specific difficulties and advantages of distinct RNA extraction procedures for scientists who have not previously specialized on a particular technique. Thus, our data allow objective choice of the most appropriate protocol to isolate RNA from low numbers of sorted cells.