Simultaneous genomic identification and profiling of a single cell using semiconductor-based next generation sequencing.

Combining single-cell methods and next-generation sequencing should provide a powerful means to understand single-cell biology and obviate the effects of sample heterogeneity. Here we report a single-cell identification method and seamless cancer gene profiling using semiconductor-based massively parallel sequencing. A549 cells (adenocarcinomic human alveolar basal epithelial cell line) were used as a model. Single-cell capture was performed using laser capture microdissection (LCM) with an Arcturus® XT system, and a captured single cell and a bulk population of A549 cells (≈ 10(6) cells) were subjected to whole genome amplification (WGA). For cell identification, a multiplex PCR method (AmpliSeq™ SNP HID panel) was used to enrich 136 highly discriminatory SNPs with a genotype concordance probability of 10(31-35). For cancer gene profiling, we used mutation profiling that was performed in parallel using a hotspot panel for 50 cancer-related genes. Sequencing was performed using a semiconductor-based bench top sequencer. The distribution of sequence reads for both HID and Cancer panel amplicons was consistent across these samples. For the bulk population of cells, the percentages of sequence covered at coverage of more than 100 × were 99.04% for the HID panel and 98.83% for the Cancer panel, while for the single cell percentages of sequence covered at coverage of more than 100 × were 55.93% for the HID panel and 65.96% for the Cancer panel. Partial amplification failure or randomly distributed non-amplified regions across samples from single cells during the WGA procedures or random allele drop out probably caused these differences. However, comparative analyses showed that this method successfully discriminated a single A549 cancer cell from a bulk population of A549 cells. Thus, our approach provides a powerful means to overcome tumor sample heterogeneity when searching for somatic mutations.

Introduction

Many areas of genomic research rely on pooled samples that include hundreds to millions of individual cells. When analyzing the genomic data of these samples, the results obtained are only average readouts. If these samples are mixtures or multi-clonal in nature, such as with tumor biopsies, then data interpretation may be hampered by low signal to noise ratios. Heterogeneity often limits data interpretation. Single-cell analysis has the potential to overcome this ambiguity in data interpretation. RNA sequencing to determine expression levels usually involves average values from bulk assays and single-cell analysis may obviate these heterogeneity issues. DNA sequence analysis also involves averaging (Shapiro et al., 2013).

Cancer research, in particular, would benefit from adopting single-cell analyses, as most tumor samples are mixtures of normal cells and cancer cells (Gerlinger et al., 2012). Recently, numerous next-generation sequencing (NGS) based studies have been conducted to provide a comprehensive molecular characterization of cancers to study tumor complexity, heterogeneity, and evolution (Shyr and Liu, 2013). Target enrichment methods for NGS are rapidly being developed and should be useful for cancer research by providing a powerful, cost effective method to study DNA and RNA in samples. Many PCR-based enrichment techniques are now available for this purpose (Mertes et al., 2011).

Currently, most cancer profiling still relies on average analyses, often because of methodological limitations. In these cases, genetic material is extracted from millions of cells. Despite the high sensitivity of modern NGS platforms, mutation frequencies of < 5% are difficult to detect even when using very high sequencing coverage (Harismendy et al., 2011). Thus, important somatic mutations may be missed due to the presence of contaminating wild-type cells or non-clonal contaminating cancer populations within the same sample (Swanton, 2012). However, research at the single-cell level enables unambiguous detection of rare variants and genetic characterization without this averaging effect of sample heterogeneity (Navin et al., 2011). Using this approach, cancer cells of different clonal origins, each containing a separate mutational profile, can be distinguished. However, single-cell level analysis carries an increased risk of contamination and analyte identification throughout the analysis is an important control step. Short tandem repeat (STR) analysis has been proposed as a means to overcome these limitations (Korzebor et al., 2013).

However, these methods are cumbersome and are not seamlessly integrated with functional analysis. Yet, this procedure can be applied to any routine NGS-based workflow. Combining single-cell methods and NGS would provide an effective means to understand single-cell biology and obviate the effects of sample heterogeneity. Here we report a single-cell identification method and seamless cancer gene profiling using semiconductor-based massively parallel sequencing.

Materials and methods

Cell culture and DNA extraction

A549 cells (adenocarcinomic human alveolar basal epithelial cells) were routinely maintained in RPMI 1640 medium with Glutamax-I supplemented with 10% fetal calf serum, penicillin (100 IU/ml), and streptomycin (100 ng/ml) (Life Technologies) with 5% CO2 in humidified air at 37 °C. Cell viability as estimated by trypan blue exclusion was > 95% prior to each experiment. For standard processing of a bulk cell population, DNA extraction and purification were performed using a PureLink™ genomic DNA kit (Life Technologies).

Single-cell capture

Single-cell capture was performed using laser capture microdissection (LCM) using an Arcturus® XT system (Life Technologies) (Pietersen et al., 2009) according to the manufacturer's instructions. A549 cells were cultured and adhered to a proton exchange membrane. A CapSure® LCM cap was placed over the target area. Laser pulsing through this cap caused a thermoplastic film to form a thin protrusion that bridged the membrane around a single A549 cell. The membrane around the A549 single cell was cut using a UV laser, and the cap was lifted to remove the target cell attached to it (Supplementary Fig. 1). A single captured cell and a blank sample, as a negative control, were subjected to whole genome amplification (WGA) using single-cell WGA kits (New England Bio Laboratories) (Zheng et al., 2011). The total amount of amplified DNA was 3.4 μg, as expected. After WGA, DNA from a single cell was purified using the PureLink™ PCR purification kit.

Library preparation

AmpliSeq technology is an ultra-high multiplex PCR method that utilizes up to 6144 PCR primer pairs in one tube (Yousem et al., 2013). Two primer pools were used for AmpliSeq target enrichment. For cell identification, the AmpliSeq™ SNP HID panel (Life Technologies) was used which interrogated 136 SNPs of high discriminatory power with a genotype concordance probability of 1031–35 (Pakstis et al., 2010, Sanchez et al., 2006). Although a 340 SNP panel was available for this technology, this panel provided sufficient discriminatory power and was cost effective.

For cancer gene profiling, we used AmpliSeq Cancer hotspot panel version 2 (Life Technologies), which included 207 primer pairs per tube to detect 50 cancer gene hotspots. DNA was extracted from a bulk population of A549 cells (≈ 106 cells), and 10 ng of DNA (≈ 3000 genome copies) was used as a PCR template. Amplicons were generated in a single PCR reaction tube with an endpoint thermal cycler. A total of 50 ng (single cell Library prep replicate #1) and 10 ng (single cell Library prep replicate #2) of WGA-amplified DNA from a single cell were subjected to PCR using the same conditions as above. The amplicons were partially digested and phosphorylated according to the manufacturer's instructions. Amplicons were ligated to adapters included in an Ion Xpress™ Barcode Adapters 1-16 kit (Life Technologies), nick-translated, and then subjected to another round of PCR to complete the linkage between adapters and amplicons. A BioAnalyzer High Sensitivity DNA kit (Agilent Technologies) was used to visualize the size range and determine the library concentration.

Semiconductor sequencing and data analysis

Individual and combined libraries were attached to Ion Sphere™ particles (ISPs) by emulsion PCR, and biotinylated ISPs were recovered from the emulsion using Dynabeads MyOne™ Streptavidin C1 beads (Life Technologies). Sequencing was performed using a semiconductor-based bench top sequencer (Ion PGM™, Life Technologies) (Rothberg et al., 2011). Four bar-coded samples were sequenced using an Ion PGM™ 200 Sequencing kit and an Ion 318™ Chip according to the manufacturer's instructions. Torrent Suite v3.2 software was used to parse bar-coded reads, to align reads to the reference genome, and to generate run metrics and total read counts and quality. Genetic variants were identified using Variant Caller v3.2 software.

Taqman® assay

A replication study was conducted using TaqMan® SNP genotyping assays with a step One Plus™ thermal cycler (Life technologies). To validate SNP HID sequencing results, allele-specific real-time PCR was used. Primers were used to identify any DNA sequence that contained a polymorphism. Allele discrimination could be determined when a fluorescent probe was hybridized in a complementary target region that should have been amplified.

Results

WGAs

We used a semiconductor-based sequencing system in combination with a cancer hotspot panel for mutational profiling of a single cell. Single-cell capture was performed using LCM (Taylor et al., 2004), followed by WGA. The procedures used and the time required are shown in Fig. 1. The total time required for a single experiment was about 21 h. Successful amplification of the samples was confirmed by agarose gel electrophoresis (Fig. 2). Negative controls were included with each amplification batch. No amplification was observed for negative cell controls. This protocol utilized a highly multiplexed PCR amplification method (AmpliSeq™, Life Technologies) to enrich target sequence pools, a human identification pool, and a Cancer hotspot panel. Amplification from a bulk population of cells and a whole-genome amplified from a single cell from the same bulk population were compared.

Fig. 1

Workflow used for this study.

A) Procedures used and the time required for an experiment. The total time required for a single experiment was approximately 21 h. B) Summary of single-cell identification and simultaneous functional sequence analysis with a semiconductor-based sequencer. Amplifications using a population of cells and a whole-genome amplified single cell from the same bulk population were compared.

Fig. 2

Results for single-cell capture and WGA.

A) Image of a single cell. This cell was captured by LCM using an Arcturus® XT system (Life Technologies). B) Captured single cells and a blank sample included as a negative control were lysed and WGA was performed using single cell WGA kits (New England Bio Laboratories). Successful amplification of the samples was checked by agarose gel electrophoresis.

Sequencing analysis

Sequence coverage was assessed from the distribution of reads across target amplicons as shown in Table 1. After subtracting multiple-template reads and poor quality sequence reads, approximately 4.7 × 106 reads were obtained. An A549 bulk population of cells mapped approximately 1.5 × 106 sequence reads, while the A549 single cell Library prep replicate #1 derived sample mapped approximately 1.2 × 106 reads. The distribution of reads across both HID and Cancer panel amplicons was consistent across samples. The average coverage between samples ranged from 2591 to 4430, and was sufficient to evaluate normal samples.

Table 1

Sequence data at SNP HID panel and Cancer hotspot panel v2.

Basic reads information
	Mapped reads (Cancer panel + HID panel)	Reads on targetb (Cancer panel + HID panel)
A549 single cell Library prep replicate #1	1,129,189	90.06%
A549 single cell Library prep replicate #2
Population cells	1,562,883	96.22%

Uniformity of coverage = percentage of bases covered at ≥ 20% of the mean coverage.

On-target reads = percentage of reads that mapped to target regions out of total mapped reads per run.

For the bulk population of cell,the percentages of sequence covered at coverage of more than 100 × were 99.04% for the HID panel and 98.83% for the Cancer panel, while the single cell percentages of sequence covered at coverage of more than 100 × were 55.93% for the HID panel and 65.96% for the Cancer panel. These differences were likely due to partial amplification failure or randomly distributed non-amplified regions across samples from single-cells during the WGA procedures or due to random allele drop out. Increased incidences of amplification failure and allele drop out have been previously reported (Garvin et al., 1998).

Comparative analysis between A549 single and bulk cells

We made a comparative analysis between two A549 single-cell replicates (Library prep replicate #1 and Library prep replicate #2) and between an A549 single cell and an A549 population of cells. Correlations for read depths between two A549 single-cell replicates and between an A549 single cell and an A549 population of cells are shown in Fig. 3.

Fig. 3

Correlations for read depths between two A549 single-cell replicates and between an A549 single cell and an A549 population of cells.

Comparative analyses were conducted for two A549 single-cell replicates and for an A549 single cell and an A549 population of cells. A) Correlation for read depths between single cell Library prep replicates #1 and #2 (#1 was from 50 ng of DNA templates and # 2 was from 10 ng of DNA templates). These results indicated a high correlation between these replicates. B) Correlation for read depths between A549 single cell Library prep replicate #1 and an A549 population of cells.

There was a high correlation between read depths for single cell Library prep replicates #1 and #2 (R2 = 0.91191). Single cell Library prep replicate #1 data were from 50 ng of DNA templates and single cell Library prep replicate # 2 data were from 10 ng of DNA templates. However, the correlation between the read depths of A549 single cell Library prep replicate #1 and an A549 population of cells was poor (R2 = 0.02306). This may also have been due to partial amplification failure or random non-amplified regions across samples from single cells during the WGA procedures or due to random allele drop out.

HID SNP typing showed high concordance rates between single cell Library prep replicate #1 and single cell Library prep replicate #2 and between an A549 single cell and an A549 population of cells. In particular, as for between single cell Library prep replicates #1 and #2, typing results were nearly the same. All 136 SNPs in the SNP HID panel were typed with the A549 population of cells, although some SNPs in the single-cell data set could not be detected. On autosomal chromosomes, 103 SNPs were typed, of which 86 SNPs were perfectly matched, 2 SNPs were partially matched, and 15 SNPs with autosomal chromosome locations had < 7 reads or had no coverage (Table 2). None of 33 SNP cells were detected on the Y chromosome with single-cell data. To validate the SNP HID sequencing results, allele-specific real-time PCR was performed using a Step One Plus™ thermal cycler with 4 primer pairs for selected non-perfectly matched SNPs (Fig. 4). This showed perfect matching between NGS typing and allele-specific real-time PCR typing results.

Table 2

Comparison analysis at SNP HID panel on the autosomal chromosomes.

	Chromosome	Position	Target ID	A549 population cells	A549 single cell Library prep replicate #1	A549 single cell Library prep replicate #2	Allele matching between population and single cell #1
Reads	Reads	Reads
1	chr1	4367323	rs1490413	783	506	887	m
2	chr1	14155402	rs7520386	8221	4566	4018	m
3	chr1	160786670	rs560681	2079	0	9	n
4	chr1	238439308	rs10495407	3626	1512	1923	m
5	chr1	239881926	rs891700	4255	91	147	m
6	chr1	242806797	rs1413212	3737	1078	2094	m
7	chr2	114974	rs876724	12185	5851	4528	m
8	chr2	182413259	rs12997453	3130	2359	4558	m
9	chr3	961782	rs1357617	4374	0	13	n
10	chr3	59488340	rs9866013	2409	951	1115	m
11	chr3	113804979	rs1872575	7710	276	311	p
12	chr3	190806108	rs1355366	8155	3452	2696	m
13	chr3	193207380	rs6444724	2755	1719	2107	m
14	chr4	76425896	rs13134862	1081	0	0	n
15	chr4	169663615	rs6811238	9403	405	0	m
16	chr4	157489906	rs1554472	511	0	325	n
17	chr4	190318080	rs1979255	7192	4561	2920	m
18	chr5	2879395	rs717302	16202	1432	934	m
19	chr5	17374898	rs159606	10053	4047	4056	m
20	chr5	136633338	rs13182883	3322	6	3	m
21	chr5	159487953	rs7704770	1111	434	917	m
22	chr5	174778678	rs251934	6320	2997	4315	m
23	chr5	178690725	rs338882	7017	15978	16675	m
24	chr6	1135939	rs1029047	372	105	350	m
25	chr6	12059954	rs13218440	6175	1066	1118	m
26	chr6	55155704	rs2811231	255	1019	1574	m
27	chr6	120560694	rs1478829	984	0	1	n
28	chr6	123894978	rs1358856	2631	0	0	n
29	chr6	148761456	rs2272998	4807	0	0	n
30	chr6	152697706	rs214955	3266	211	329	m
31	chr6	165045334	rs727811	8586	15	31	m
32	chr7	4310365	rs6955448	5403	2536	4468	m
33	chr7	4457003	rs917118	5497	792	716	m
34	chr7	13894276	rs1019029	6249	1272	1605	m
35	chr7	137029838	rs321198	7283	251	277	p
36	chr7	155990813	rs737681	13712	3881	2151	m
37	chr8	28411072	rs10092491	7523	67	121	m
38	chr8	136839229	rs4288409	4312	277	168	m
39	chr8	139399116	rs2056277	11214	4874	5782	m
40	chr8	144656754	rs4606077	1964	738	680	m
41	chr9	14747133	rs2270529	9489	1405	1911	m
42	chr9	27985938	rs7041158	3365	5217	6537	m
43	chr9	126881448	rs1463729	3678	21817	17240	m
44	chr9	137417308	rs10776839	7888	1506	1608	m
45	chr10	3374178	rs735155	2297	1198	1769	m
46	chr10	17193346	rs3780962	7251	3741	2789	m
47	chr10	97172595	rs1410059	5509	32	57	m
48	chr10	118506899	rs740598	9602	193	198	m
49	chr10	132698419	rs964681	10487	15260	16681	m
50	chr11	5098714	rs10768550	4277	69	101	m
51	chr11	5099393	rs10500617	8876	0	3	n
52	chr11	5709028	rs1498553	5400	10838	24725	m
53	chr11	11096221	rs901398	8284	12116	10728	m
54	chr11	105912984	rs6591147	3389	0	0	n
55	chr11	122195989	rs590162	1306	0	0	n
56	chr12	888320	rs2107612	868	16	49	m
57	chr12	6909442	rs2255301	6306	1009	933	m
58	chr12	6945914	rs2269355	6034	19889	20339	m
59	chr12	106328254	rs2111980	1820	14	12	m
60	chr12	130761696	rs10773760	20475	75	37	m
61	chr13	22374700	rs1886510	554	35	49	m
62	chr13	84456735	rs9546538	2879	9826	17784	m
63	chr13	100038233	rs1058083	2334	532	1110	m
64	chr13	106938411	rs354439	2635	2	4	m
65	chr14	25850832	rs1454361	9917	1251	915	m
66	chr14	98845531	rs873196	6715	2750	3275	m
67	chr14	104769149	rs4530059	6079	22987	23406	m
68	chr15	39313402	rs1821380	4909	1162	1481	m
69	chr16	5606197	rs729172	10102	2275	2360	m
70	chr16	5868700	rs2342747	2610	6048	5248	m
71	chr16	78017051	rs430046	6083	8464	11680	m
72	chr16	80106361	rs1382387	3983	443	444	m
73	chr17	41286822	rs2175957	7090	3093	4591	m
74	chr17	41341984	rs8070085	6172	9215	11726	m
75	chr17	41691526	rs1004357	5712	22482	23964	m
76	chr17	80526139	rs2291395	8351	19	9	m
77	chr17	80531643	rs4789798	6312	43	80	m
78	chr17	80715702	rs689512	6038	129	129	m
79	chr17	80739859	rs3744163	8121	480	652	m
80	chr17	80765788	rs2292972	7626	8748	12108	m
81	chr18	1127986	rs1493232	2756	18	34	m
82	chr18	9749879	rs9951171	6520	872	683	m
83	chr18	22739001	rs7229946	5395	2718	3322	m
84	chr18	29311034	rs985492	4040	14112	16841	m
85	chr18	47371014	rs521861	3959	0	0	n
86	chr18	55225777	rs1736442	5762	16	16	m
87	chr18	75432386	rs1024116	1300	1762	2807	m
88	chr19	28463337	rs719366	17160	5626	5487	m
89	chr19	39559807	rs576261	5800	8440	8923	m
90	chr20	16241416	rs12480506	11582	3938	4690	m
91	chr20	23017082	rs2567608	11131	18667	22692	m
92	chr20	39487110	rs1005533	4770	2345	3535	m
93	chr20	51296162	rs1523537	3113	2924	5102	m
94	chr21	16685598	rs722098	746	98	264	m
95	chr21	28023370	rs464663	5969	1750	1357	m
96	chr21	33582722	rs2833736	11000	6320	3864	m
97	chr21	42415929	rs914165	4174	6788	5841	m
98	chr22	19920359	rs9606186	8589	26670	25031	m
99	chr22	23802171	rs2073383	3889	659	798	m
100	chr22	27816784	rs733164	4370	1531	1524	m
101	chr22	33559508	rs987640	3691	1511	2156	m
102	chr22	47836412	rs2040411	11123	1985	1680	m
103	chr22	48362290	rs1028528	5252	5580	4370	m

m = match; p = partial match; n = no depth.

Fig. 4

Allelic description plots as replication study using TaqMan® SNP genotyping assays.

To validate SNP HID sequencing results, allele-specific real-time PCR was performed. Four representative plots showing performance of four assays in analysis of A549 samples and reference samples. VIC signal (x-axis) is associated with the probe for allele A (graph (1), (3)) and allele C (graph (2), (4)), while FAM (y-axis) labels the allele G (graph (1), (3)) and allele T (graph (2), (4)) probes. Aqua blue × symbols indicate A549 bulk cells and a single cell with NGS reads data. Circles symbols and black × symbols indicate 20 Coriell gDNA samples as reference.

Cancer gene analysis

A Cancer gene panel was used for a functional analysis (Table 3). We again found high concordance rates between A549 single cell Library prep replicates #1 and #2 and between an A549 single cell and an A549 population of cells. A total of 11 variants were typed for both samples, of which 1 was partially matched and 5 SNPs were not detected because of low or no depth in the single cell Library prep replicates #1 and #2 data set. A total of 16 variants were detected in A549 single cell Library prep replicates #1 and #2 cell and 13 variants were detected in an A549 population of cells; 11 variant cells were perfectly consistent. Five SNPs were called as variants and some discrepancies were observed. No frameshifts or deletions were observed at 2790 hotspots.

Table 3

Comparative analysis for the Cancer hotspot panel of 50 cancer-related genes.

Chromosome	Position	Gene Sym	Hotspot ID	A549 population cells							A549 single cell Library prep replicates #1
Zygosity	Ref	Variant	Var freq	Coverage	Ref cov	Var cov	Zygosity	Ref	Variant	Var freq	Coverage	Ref cov	Var cov
Match pairs list
chr4	1807894	FGFR3	–	Hom	G	A	99.7	2003	6	1997	Hom	G	A	99.25	2135	14	2119
chr4	55141055	PDGFRA	–	Hom	A	G	100	1605	0	1605	Hom	A	G	99.69	12002	21	11965
chr5	149433597	CSF1R	–	Hom	G	A	97.6	1503	36	1467	Hom	G	A	96.18	12894	458	12402
chr5	149433596	CSF1R	–	Hom	T	G	97.88	1464	1	1433	Hom	T	G	97.2	12285	33	11941
chr7	55249063	EGFR	–	Hom	G	A	99.88	2456	2	2453	Hom	G	A	100	12	0	12
chr10	43615633	RET	–	Het	C	G	66.46	3208	1075	2132	Het	C	G	64.45	422	149	272
chr10	43613843	RET	–	Hom	G	T	99.85	6073	0	6064	Hom	G	T	99.56	1609	0	1602
chr12	25398285	KRAS	COSM517;	Hom	C	T	99.62	4487	17	4470	Hom	C	T	100	24	0	24
chr13	28610183	FLT3	–	Hom	A	G	99.9	4910	4	4905	Hom	A	G	99.88	3342	4	3338
chr17	7579472	TP53		Het	G	C	91.03	2520	225	2294	Het	G	C	88.23	3865	446	3410
chr19	1207021	STK11	COSM12925;	Hom	C	T	99.9	2909	3	2906	Hom	C	T	99.35	10927	47	10856
Not mutch pairs list
chr3	178917005	PIK3CA	–	Hom	A	G	99.57	1153	5	1148	Not detected
chr4	55602749	KIT		Not detected							Het	T	C	46.36	4864	2581	2255
chr4	55979623	KDR	COSM32339	Het	C	G	48.72	2422	1243	1180	Het	C	T	73.68	19	5	14
chr11	108155120	ATM		Not detected							Het	G	T	50	12	6	6
chr11	108204661	ATM		Not detected							Het	T	C	70.54	258	76	182
chr11	108204660	ATM		Not detected							Hom	T	C	91.89	259	21	238

Discussion

We have described a genomic single-cell identification method with simultaneous functional analysis using NGS. We used the A549 cell line to check for concordance rates between a single cell and ≈ 106 cells in a bulk population. Working with single cells requires careful monitoring, for which two approaches are primarily used: LCM and cell sorting.

Using these approaches, technical contamination should be ruled out. Sources of contamination can be unrelated genetic material that is inadvertently introduced into a sample. Simple and robust techniques to identify or confirm the genetic origin of a cellular material under investigation are a critical quality control step. With the application described here, we paired cell identification with cancer profiling.

HID SNP typing showed high concordance rates between an A549 single cell and an A549 population of cells. However, some SNPs on autosomal chromosomes and all SNP cells on the Y chromosome in a single-cell data set could not be detected. Depletion of the Y chromosome is often observed for transferred culture cells; thus, this may also have occurred with our preparations (Ono et al., 2001). There have been many reports regarding allele drop out and failed amplification rates after single cell WGA (Baslan et al., 2012, Spits et al., 2006, Handyside et al., 2004, Handyside et al., 2010, Konings et al., 2012).

Regarding the WGA methodology, some investigators have indicated that multiple displacement amplification (MDA), such as with QIAgen's REPLI-g technology, was more appropriate for microarray genotyping applications than PCR-based WGA, such as the NEB WGA kit used in this study (Treff et al., 2011). MDA-based WGA (Repli-G) may result in less allele dropout, which may suggest better results for the AmpliSeq protocol. We intend to compare amplification methodologies in future studies.

Although genomic instability or inefficient WGA may compromise analysis using single cells, we used 136 SNPs that were evenly distributed across the entire genome for discrimination purposes. Thus, despite the fact that some genome regions were missing in our single-cell data sets, the HID SNP set used here retained its discriminatory capability. To confirm the utility and robustness of our method, we intend to repeat our experiment using more single cell replicates and different cell-picking methods. The former should help to understand genomic instability or efficiency of WGA, the latter should help identify any background that results from using LCM. Although we plan to explore these issues in the future, in this report, we cannot deal with these issues because of the costs involved and the labor-intensive nature of the procedures used.

Regarding cancer gene analysis, 5 SNPs were called as variants and some discrepancies were found. Only 3 of 5 variants were detected for the ataxia telangiectasia mutated (ATM) gene. This was likely due to random non-amplified regions across samples of single cells during WGA.

Other possible applications for our method include forensics, transplantation medicine, regenerative medicine, and pre-natal testing using maternal blood (Fan et al., 2008). Forensic samples are often heterogeneous. In many cases, samples at crime scenes are mixtures from multiple subjects (e. g., offender, victim, or unrelated individual). Single-cell analysis should remove any ambiguity in data interpretation.

In conclusion, our method provides an easy to implement and effective method to investigate sample heterogeneity in various areas, such as tumor biology, forensics, regenerative medicine, and fetal DNA tracing in maternal blood samples.

The following are the supplementary data related to this article.

Supplementary Fig. 1

Workflow of living single-cell capture.

Completing interests

All authors work for Life Technologies Japan, Ltd.