Improved Method for Bacterial Cell Capture after Flow Cytometry Cell Sorting

Abstract

Fixed cells with different nucleic acid contents and scatter properties (low nucleic acid [LNA], high nucleic acid 1 [HNA1], and HNA2) were sorted by flow cytometry (FCM). For each sort, 10,000 cells were efficiently captured on poly-l-lysine-coated microplates, resulting in efficient and reproducible PCR amplification.

Combining flow cytometry (FCM) with cell sorting offers the opportunity to physically isolate specific subpopulations with distinct scattering properties or nucleic acid or lipid contents (3). The combination of cell sorting by FCM with molecular tools like whole-genome amplification (WGA) or PCR amplification allows for the molecular taxonomic identification of specific bacterial populations and potentially their spatial distribution, temporal dynamics, and physiology (13, 15-17). However, it can be challenging to apply molecular methods to sorted cells. Cell fixation, sea salts, and natural substances present in seawater can all inhibit DNA polymerase and other enzymes used in molecular methods (1, 14, 19), and the number of cells sorted from subpopulations may be too low to construct clone libraries.

The main goal of the present study was to efficiently concentrate FCM-sorted cells on poly-l-lysine-coated microtiter plates, which allow for the elimination of a number of PCR inhibitors and decreased contamination or sample loss. Further, we developed a reliable protocol, allowing for the sorting of a few fixed cells, to construct clone libraries of the high nucleic acid 1 (HNA1), HNA2, and low-nucleic-acid (LNA) subgroups and compared them to the entire community library.

Surface seawater samples were collected with 12-liter Niskin bottles at a depth of 5 m in the northwest Mediterranean Sea at the Microbial Observatory Laboratoire Arago (MOLA) station, located 20 nautical miles off Banyuls/mer (France) in June 2008. Subsamples (5 ml) of environmental samples either fixed with formalin (2% final concentration) or a mixture of 0.5% formaldehyde-0.1% glutaraldehyde (final concentrations) for 1 h at 4°C were stained with SYBR green II (Invitrogen-Molecular Probes) as described by Lebaron et al. (12) and discriminated by FCM with a FACSAria (Becton Dickinson) equipped with two lasers: a laser with 488-nm excitation (13-mW; Sapphire solid-state laser; Coherent, Inc.) and a laser with 633-nm excitation (11 mW; JDS Uniphase HeNe air-cooled laser). Measurements of SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate), green fluorescence (530/30 nm), and red fluorescence (695/40 nm) were done using 488-nm laser excitation. The sort precision mode used was the 4-way purity mode (0/32/0), and the sorting efficiency was checked by reanalysis of the sorted cells (data not shown). The sheath fluid used was 30 kDa (TFF cartridge; Millipore) of filtered seawater sterilized 6 h at 80°C. Bleach cleaning of all parts of the machine was performed to eliminate external sources of prokaryotic contamination.

Ten thousand bacteria from each of the HNA1, HNA2, or LNA subpopulations (Fig. 1) were sorted into either untreated or poly-l-lysine-treated 96-well microplates (PCR-96-C; Axygen) or directly onto 0.22-μm sterile multiscreen GV polyvinylidene difluoride (PVDF) 96-well devices (Millipore). Poly-l-lysine coating of 96-well microplates was performed by incubating 5 μl of poly-l-lysine solution (0.1-mg/ml P4832; Sigma) in each well for 1 h at 4°C. Wells were washed three times with ultrapure water (Sigma), dried at 40°C, and exposed to UV light for sterilization three times at 1,200 kJ for 30 s. Poly-l-lysine-coated microplates were stored at 4°C for up to 3 months (8). Sorted bacteria were centrifuged for 15 min at 10,400 × g at 4°C. Supernatants were collected in cytometry sampling tubes to be checked for uncaptured cells after SYBR green I staining to increase the fluorescence of free cells. Cell capture was significantly enhanced by the poly-l-lysine treatment, as determined by flow cytometry. Fewer uncaptured cells were detected in the poly-l-lysine-treated microplates (2% ± 1% [mean ± standard deviation] of total cells) than in the untreated ones (7.7% ± 2.7% of total cells) (one-way Student t test, P < 0.0001, n = 32). This significant improvement in recovery with the poly-l-lysine was confirmed by three additional independent experiments that showed 2.76% ± 2.71%, 2.9% ± 1.60%, and 3.19% ± 1.53% of uncaptured cells (analysis of variance [ANOVA] test, number of samples studied in the first experiment [n1] = 220, n2 = 68, n3 = 21; P > 0.05). The cell loss could be explained by dead or unfit cells (2, 5).

FIG. 1.

Flow cytometric signatures and cell abundances of the LNA and HNA populations in surface waters (5 m) at the MOLA station on 10 June 2008.

The trapped cells were incubated for 30 min at 4°C in a 5× PCR buffer (SuperTaq 10× buffer; HT Biotechnology Ltd.), 5× Tris-EDTA (TE; product no. 86377, Sigma), and 0.1 μg/μl bovine serum albumin (BSA) (product no. R3961; Promega). Blocking using BSA allows downstream molecular detection by PCR within the wells used for capture, reducing sample manipulations, decreasing potential external contamination, and enhancing the recovery of genetic material. All samples were subjected to three freeze-thaw cycles using liquid nitrogen and boiling water. For each subpopulation library, three replicate PCRs were performed. All PCR controls consisting of an empty poly-l-lysine-treated well and a poly-l-lysine-treated well filled with 10,000 sorted fluorescent beads (1,002 μm; Polysciences Europe) gave negative amplification results when the same conditions were used for each. The 16S rRNA genes were amplified as described by West et al. by using a modified enzyme mix: 1 U SuperTaq (HT Biotechnology Ltd.) and 0.1× TaqAdvantage (Clontech) (18). The amplified 16S rRNA genes of total, HNA1, HNA2, and LNA populations were formerly purified with the QiaQuick PCR purification kit (Qiagen), and their respective concentrations were determined using the PicoGreen assay (Invitrogen, Molecular Probe) by following the manufacturer's instructions. In the 0.5% formaldehyde-0.1% glutaraldehyde-fixed samples and formalin-fixed samples, the amplification success rates were similar for the 10,000 cells captured with the poly-l-lysine treatment (73% and 79%, respectively) and the PVDF filters (86% and 72%, respectively) (Table 1). Most of the amplifications were negative when the cells were filtered onto 0.2-μm polycarbonate filters (two successful amplifications among 17 analyzed samples). The same results were obtained with the different fixatives with a slight improvement of PCR amplification for the 0.5% formaldehyde-0.1% glutaraldehyde mixture. Only the amplicon yield seemed to be affected by the concentration and to some extent by the fixation procedures (Table 2). These data indicate that the poly-l-lysine approach and the 0.5% formaldehyde-0.1% glutaraldehyde fixation offer a better sensitivity, a higher yield, and a good reproducibility of PCR amplification compared to those offered by the filter technique (9).

TABLE 1.

Comparison of the effects of fixatives and concentration procedures on the 16S rRNA gene amplification reproducibility^a

Procedure	0.5% formaldehyde-0.1% glutaraldehyde-fixed samples			Formalin-fixed samples
	No. of successfully amplified samples	No. of samples analyzed	% of successfully amplified samples	No. of successfully amplified samples	No. of samples analyzed	% of successfully amplified samples
Poly-l-lysine treatment	32	44	73	33	42	79
Classical filtering	2	17	12	ND	ND	ND
PVFD device filtering	18	21	86	13	18	72

^aND, not determined.

TABLE 2.

Effect of fixatives and concentration procedures on the 16S rRNA gene amplification yield

Fixation	Treatment	Type of bacterial population	Amplicon concn (ng/μl)a	% of relative amplification yield
0.5% formaldehyde-0.1% glutaraldehyde	Poly-l-lysine treatment	Total population	7.37 ± 0.72	99
		HNA1 population	7.44 ± 1.85	100
		HNA2 population	6.42 ± 3.26	86
		LNA population	3.6 ± 1.22	48
	PVDF filtered	Total population	1.99 ± 1.20	27
	PC filtered	Total population	3.42 ± 0.38	46
2% formalin	Poly-l-lysine treatment	Total population	5.12 ± 2.37	69
		HNA1 population	7.16 ± 3.66	96
		HNA2 population	5.83 ± 3.88	78
		LNA population	6.19 ± 2.52	83
	PVDF filtered	Total population	1.91 ± 0.05	26

^aMean ± SD of results from three different sorted samples.

To assess whether our protocol allowed for further molecular work on sorted cells, we conducted PCR amplification on 10,000 sorted cells from the HNA1, HNA2, and LNA groups and constructed clone libraries from each subpopulation. We then compared the HNA1, HNA2, and LNA bacterial diversity to that of a clone library prepared from a mixed population of 10⁹ unsorted bacterial cells sampled at the same time and constructed as described by Laghdass et al. (11).

The coverage values of all the 16S rRNA gene clone libraries were calculated according to Good's formula (6). The diversity indices and the rarefaction analysis were obtained using the PAST software (7).

At a high phylogenetic level (98.5% similarity), the HNA1, HNA2, and LNA communities had high coverage values, indicating that the sequencing effort covered the subpopulations well (Table 3). The Shannon index and rarefaction curves indicated that HNA1 and HNA2 community diversities were higher than the LNA diversity but similar to the diversity of the total community (Table 3 and Fig. 2). The lower coverage observed for the total community library can be explained by the higher number of unique clones (n = 47), reflected in a low coverage value (Table 3). Considering the rarefaction curve and the high clone coverage of the LNA library (Fig. 2), these results suggest a different phylogenetic composition rather than methodological biases (4).

TABLE 3.

Sequence diversity and library coverage estimations^a

Population	No. of clones	No. of phylotypes	No. of singletons	% of library coverageb	Diversity index valuec
					Dominance_D	Shannon_H	Simpson_1-D	Evenness_E
HNA1	92	29	9	90.2	0.0822	2.945	0.9178	0.6553
HNA2	91	32	12	86.8	0.05615	3.148	0.9438	0.7275
LNA	85	15	3	96.5	0.1039	2.462	0.8961	0.7818
Total	162	32	47	71	0.1211	2.735	0.8789	0.4814

^aAll the sequences were analyzed using Clusterer at a similarity threshold of 98.5%, for which no significant decrease in the number of operational taxonomic units (OTUs) was observed (UPGMA settings, 98.5% similarity) (10).

^bLibrary coverage was calculated according to Good's formula: C = 1 − (n/N), where n is the number of clones occurring only once in the library and N is the total number of clones.

^cThe different diversity indices were calculated by the PAST software. Dominance_D is the Simpson index of dominance (D), Shannon_H is the Shannon diversity index (H), Simpson_1-D is the Simpson index of diversity 1− D, and Evenness_E is the relative abundance.

FIG. 2.

Rarefaction curves and clone coverage of HNA1, HNA2, LNA, and total libraries computed by the PAST software. OTUs, operational taxonomic units.

All together, these results support the efficiency of our protocol in constructing clone libraries for comparing specific fractions of environmental communities and allow a precise comparison of the community diversity by molecular tools.

Our protocol will be very useful to better understand how the structure of HNA and LNA subpopulations change with time and space in relation to environmental changes. It should also allow a better understanding of the ecological meaning of these cytometric groups. This protocol should also be used to complement other approaches to further investigate the species diversity of bacterial communities.

Nucleotide sequence accession numbers.

A total of 266 sequences were generated from the three sorted subpopulations and submitted to GenBank under accession numbers HM037370 to HM037635.