Restriction site tagged (RST) microarrays: a novel technique to study the species composition of complex microbial systems

Abstract

We have developed a new type of microarray, restriction site tagged (RST), for example NotI, microarrays. In this approach only sequences surrounding specific restriction sites (i.e. NotI linking clones) were used for generating microarrays. DNA was labeled using a new procedure, NotI representation, where only sequences surrounding NotI sites were labeled. Due to these modifications, the sensitivity of RST microarrays increases several hundred-fold compared to that of ordinary genomic microarrays. In a pilot experiment we have produced NotI microarrays from Gram-positive and Gram-negative bacteria and have shown that even closely related Escherichia coli strains can be easily discriminated using this technique. For example, two E.coli strains, K12 and R2, differ by less than 0.1% in their 16S rRNA sequences and thus the 16S rRNA sequence would not easily discriminate between these strains. However, these strains showed distinctly different hybridization patterns with NotI microarrays. The same technique can be adapted to other restriction enzymes as well. This type of microarray opens the possibility not only for studies of the normal flora of the gut but also for any problem where quantitative and qualitative analysis of microbial (or large viral) genomes is needed.

INTRODUCTION

Identification and quantification of microbial species in their various habitats is very important for understanding and dealing with different aspects of human and animal health and disease. For example, identification of pathogenic bacteria in food, soil or in the air can prevent epidemics. Not very much is known about the normal human microflora. The human intestinal tract harbors a densely populated, active and complex bacterial ecosystem. The number of microbial cells in the colon is estimated to be 10–100 times larger than the number of eukaryotic cells in the entire human body and weighs >1 kg. Microscopic investigations demonstrated that many different species of microorganisms live in our digestive system, but at least 85% are unknown, mainly because they cannot easily be cultured in vitro (1–3). At the same time many studies have shown that the composition of the gut microflora plays a very significant role in human health (1,4,5).

There are some methods available to analyze complex microbial mixtures, e.g. by enzyme analysis, which requires growth of colonies outside the body, or analysis of the fatty acids composition in stools, both of which give crude and indirect indications of the composition of the normal flora (2,6,7). The application of culture-independent techniques based on molecular biology methods can overcome some shortcomings of conventional cultivation methods. In recent years an approach based on PCR amplification of 16S rRNA genes has become both popular and very useful (7–11). One modification of the approach utilized fingerprinting of all the species in the gut using, for instance, denaturing gradient gel electrophoresis with PCR amplified fragments of 16S rRNA genes. In another application, PCR amplified fragments of 16S rRNA genes were directly cloned and sequenced. These studies provided important information, however, intrinsic disadvantages of the approach limit its application. The problem is that 16S rRNA genes are highly conserved and therefore the same sequenced fragment can sometimes represent different species. It is also difficult to adapt for quantification. Moreover, in fingerprinting experiments similar fragments may represent different species and yet different fragments may also represent the same species.

Microarray technology using immobilized DNA has opened up new possibilities in the molecular biology of eukaryotes and prokaryotes (12–17). This approach has also been applied to the studies of the bacterial composition of the microflora and identification of specific microbial species. However, microarrays based on 16S rRNA genes suffer from the same problems as sequencing/fingerprinting methods and species-specific microarrays based on PCR amplification of specific DNA/gene fragments can only be used for identification of a limited number of microorganisms.

Recently we developed a new approach to genome mapping and sequencing based on ‘slalom libraries’ (18). Experiments demonstrated that even short 19 bp sequences suffice to complete a physical map and sequence scan of a small genome. These short signature sequences pinpointed differences in genome organization between organisms and allowed the establishment of a minimal set of overlapping clones. The slalom approach made it possible to locate virtually every gene in a genome, for more detailed studies. Based on these findings, we suggested a novel approach to analyze microbial mixtures: a NotI passporting or NotI tagging procedure (19). Short 19 bp tags flanking a NotI site are generated from the sample DNA. A collection of such tags constitutes a NotI passport. Such NotI passports allow efficient discrimination between closely related bacterial species and even strains. The approach can utilize any other rare cutting restriction enzyme and was called the passporting procedure or generation of restriction site tagged sequences (RSTS). Comparing the three available fully sequenced Escherichia coli genomes, among 1312 tags available for three different restriction enzymes (NotI, PmeI and SbfI) only 219 tags were found not to be unique. Hundreds of thousands of tags can be produced by a small group in a short time, allowing careful analysis of thousands of bacteria from many species/strains (20). Thus this approach uniquely describes bacterial genomes and allows analysis of complex microbial mixtures with high accuracy on a quantitative and qualitative basis.

Additionally, we have also developed a new type of microarray: RST (restriction site tagged) microarrays (21). These studies have shown that NotI microarrays could be successfully used to detect deleted/amplified or methylated NotI sites in the human genome. The sensitivity of these microarrays is sufficient to detect even hemizygous deletions.

We now suggest a similar approach for the analysis of colon microflora composition.

MATERIALS AND METHODS

General methods

Growth of the bacteria and the microbiological and molecular biology procedures were performed according to standard methods.

Genomic DNA extraction from bacterial cells was performed according to the protocol at http://www.uct.ac.za/depts/mmi/bbhelp/bac1.html. The protocol was based on the method of Marmur (22) as modified by J.L. Johnson (Virginia Tech, Blacksburg, VA). Bacterial genomic DNA from mouse colon was prepared using a DNeasy Tissue Kit (Qiagen GmbH, Hilden, Germany). DNA from feces was isolated with a QIAamp DNA Stool Mini Kit (Qiagen).

The construction of NotI linking libraries was described previously (23), with the modification that in this work linking libraries were constructed in pBluescript KS(+) vector (Stratagene, La Jolla, CA), which had been pre-digested with NotI and dephosphorylated with alkaline phosphatase (Roche Molecular Biochemicals, Indianapolis, IN). Plasmid DNA was purified using a REAL prep kit (Qiagen).

Sequencing gels were run on ABI 310 and ABI 377 automated sequencers (Applied Biosystems, Foster City, CA) according to the manufacturer’s protocol.

Amplification and sequencing of 16S rRNA genes

Almost complete 16S rRNA genes were amplified using degenerate PCR primers targeting the termini of the gene. The primers were constructed as follows: 5′-agagtttgatiitggctcag-3′ and 5′-cggitaccttgttacgac-3′, where i denotes inosine and the 3′-ends target positions 27 and 1494 of the 16S rRNA gene of E.coli (24), respectively. The temperature profile was 96°C for 15 s, 40°C for 30 s and 72°C for 90 s. A total of 30 cycles was performed, followed by an extension step at 72°C for 10 min. The primers used for sequencing have been described elsewhere (8).

Preparation of NotI representations

Two oligonucleotides, NotX (5′-AAAAGAATGTCAGTGTGTCACGTATGGACGAATTCGC-3′) and NotY (3′-AAACTTACAGTGTGTGTCACGTATGGCTGCTTAAGCGC CGG-5′), were used to create the NotI linker. Annealing was carried out in a final volume of 100 µl, containing 20 µl 100 µM NotX, 20 µl 100 µM NotY, 10 µl 10× M buffer (Roche Molecular Biochemicals) and 50 µl H₂O. The reaction mixture was boiled for 8 min and allowed to cool slowly at room temperature. An aliquot of 2 µg of bacterial DNA was digested with 20 U BamHI and 20 U BglII (Roche Molecular Biochemicals) at 37°C for 5 h and then heat-inactivated for 20 min at 85°C. Then, 0.4 µg of the digested DNA was circularized overnight with T4 DNA ligase (Roche Molecular Biochemicals) in the appropriate buffer in 0.1 ml reaction mixtures. The DNA was then concentrated with ethanol, partially filled in (23) and digested with 10 U NotI at 37°C for 3 h. Following digestion, NotI was heat-inactivated and the DNA was ligated overnight in the presence of a 50-fold molar excess of NotI linker at room temperature. Subsequently, DNA was purified using a JETquick PCR Purification Spin Kit (Genomed GmbH, Bad Oeynhausen, Germany) and dissolved in 50 µl H₂O.

PCR was performed in 40 µl of a solution containing 67 mM Tris–HCl (pH 9.1), 16.6 mM (NH₄)₂SO₄, 1.0 mM MgCl₂, 0.1% Tween 20, 200 µM dNTPs, 100 ng tester amplicon DNA, 400 nM primer NotX and 5 U Taq DNA polymerase. The PCR cycling conditions were 72°C for 5 min, followed by 25 cycles at 95°C for 1 min and 72°C for 2.5 min, with a final extension at 72°C for 5 min. The amplification products had an average size of 0.5–1.5 kb. These PCR amplified fragments were called ‘NotI representations’ (NRs).

Microarray preparation, hybridization and scanning

Microarrays were constructed essentially as described (25). In brief, DNA from NotI linking clones was PCR amplified with Zuniv (5′-ccagggttttcccagtcacgac-3′) and M13rev (5′-acacaggaaacagctatgaccatg-3′) in 30 µl as follows: initial heating at 95°C for 4 min followed by 25 cycles of PCR at 95°C for 30 s, 65°C for 30 s and 72°C for 3 min, with a final extension of 10 min at 72°C. These PCR fragments were spotted onto 3-aminopropyl trimethoxysilane-coated glass microscope slides. Most NotI clones contained inserts of 0.5–3 kb. The PCR amplified DNA was dissolved in 1× saline sodium citrate (SSC) with 30% dimethylsulfoxide and arrayed using a GMS 417 Arrayer (Genetic MicroSystems, Woburn, MA) with a spot density of 350 µm. The arrays were subsequently air dried, submerged in 70% ethanol for 30 min at room temperature, air dried again and stored in the dark at –20°C.

The NR probes (5 µl) were labeled in a PCR reaction (20 cycles of 95°C for 30 s and 72°C for 3 min) with the NotX primer. Digoxigenin or biotin was incorporated using the PCR DIG Labelling Mix (Roche Molecular Biochemicals) or Biotin Reaction Mix (Micromax, NEN Life Science Products, Boston, MA), respectively. PCR products were purified using SigmaSpin columns (Sigma-Aldrich Co., St Louis, MO) and the efficiency of labeling was determined by membrane-based chemiluminescence analysis (Micromax, NEN). Arrays were washed for 5 min at room temperature in low stringency buffer (0.06× SSC, 0.01% SDS) and developed using the TSA system (Micromax, NEN). Control results (not shown) indicated that, under the conditions used in this study, the strength of the signal had an almost linear dependence on the amount of DNA attached to the glass slide. Arrays were scanned using the GMS 418 Scanner (Genetic MicroSystems) and were analyzed and represented by ImaGene 3.02 software (BioDiscovery Inc., Marina del Rey, CA).

An alternative method for preparing NRs with low quality DNA was also used. In this method, genomic DNA was simultaneously digested with NotI and another enzyme or combination of enzymes without CpG pairs in their recognition sites (e.g. Sau3A or BamHI + BglII). After enzyme inactivation, two specific adaptors

(Sau00N:

5′-GATCCTCAAACGCGT-3′amine

3′-GAGTTTGCGCACAGCACTGACCCTTTTGGGACC-5′;

NBSgt99:

5′-GGCCTCCAGAAAACATCCACGGGCTCTAGGATAGATCGC-3′

3′-AGGTCTTTTGTAGG-5′)

were ligated and NRs were prepared using PCR in the presence of Zuniv and Zgt (5′-GGCGATCTATCCTAGAGCCCGT-3′) primers. The PCR cycling conditions were 95°C for 2 min, followed by 25 cycles of 95°C for 45 s, 65°C for 30 s and 72°C for 90 s.

In general, these NRs showed the same results in hybridization experiments, but the background was usually higher. This effect was probably caused by illegitimate ligation of the NotI linker to DNA fragments with BamHI or BglII sticky ends.

Sequence analysis

The analysis of sequences was performed at the Karolinska Institute Sequence Analysis Center (kisac@cgb.ki.se), using local versions of programs and public databases. Homology searches were done with the EMBL database (release 71) including all bacterial entries. Nucleotide similarity searches were performed with BLAST 2.2 (26) using the non-gapped alignment method (blast parameter g = F). The high scoring segment pairs report cut-off (blast parameter b) was restricted to 50 pairs (b = 50) and the statistical significance threshold (blast parameter e) was set to e = 1.E–10.

RESULTS AND DISCUSSION

General scheme of the experiment

The major problem of normal flora screening is the complexity of microbial genomes in the mixtures. In these experiments we propose to solve this problem by using specifically selected fragments of the genomes (e.g. NotI representations, NotI linking clones, etc.) located in various regions of the bacterial DNA that are not highly conserved. Thus we do not aim for analysis/sequencing of complete genomes or a study of all genes. We assign special signatures to particular microorganism/genes and analyze these signatures in different samples of colonic flora.

The main idea of the approach is shown in Figure 1. NotI linking clones (or oligonucleotides flanking NotI sites) are generated from different bacterial species and attached to the glass. NRs are produced from the experimental samples and only sequences surrounding NotI sites are labeled. For this purpose we used the first steps of the NotI-CODE procedure (cloning of deleted sequences; 21). The NotI-CODE procedure represents a modification of the CODE procedure (27,28). With this approach only a small fraction of genomic DNA is labeled (0.1–0.5% depending on the experimental conditions), increasing the sensitivity of hybridization detection several hundred-fold compared to conventional labeling of the total DNA/RNA.

Figure 1.

Flow chart diagram explaining generation of NotI representations and hybridization to microarrays containing NotI linking clones. Only circles containing NotI sites are shown and they represent 0.1–0.5% of all circles depending on the bacterial species and restriction enzymes used in the first step. In strain K1 all three NotI sites will be labeled and in strain K2 only one (N1). The N2 site is deleted and N3 produces too large a DNA fragment to be efficiently amplified.

Importantly, the scheme is resistant to incomplete digestion and ligations, i.e. if incomplete reactions take place no artificial probe will be generated. For example, incomplete digestion with BamHI + BglII or inefficient self-ligation would result in large DNA fragments that could not be PCR amplified. DNA incompletely digested with NotI or inefficiently ligated with NotI linkers could not serve as a substrate for PCR amplification. This feature is one of the main advantages of the procedure that was discussed previously (21,23).

Discrimination of closely related bacterial species/strains using NotI microarrays

Our preliminary experiments had several aims. The first was to check if NotI microarrays could be used for microbial species identification. We prepared NotI linking libraries from four bacterial species (E.coli R2, Enterobacter cloaceae R4, Klebsiella pneumoniae B4958 and Serratia liquefaciens) and sequenced 192 random clones. Ninety-six of these clones were used to prepare microarrays. NR probes were prepared using DNA from the same bacteria and hybridized to the microarrays. Specific hybridization patterns were obtained and therefore this approach can be used for microbial identification.

The second question was to establish the discriminative power of the procedure. In addition to the previous four species, NR probes were prepared from two E.coli strains: K12 and DS17. All six bacteria are highly related. The E.coli strains are almost identical (99%) in 16S rRNA gene sequence and the other three microbial species are hardly distinquishable from E.coli K12 (E.cloaceae R4, 97% identity, K.pneumoniae B4958, 96% and S.liquefaciens, 94%). In practice, only 16S RNA gene sequences that display similarity <95% are usually considered different (B.Pettersson, personal communication). These six NR probes were hybridized to the microarrays containing selected NotI linking clones from E.coli R2, E.cloaceae R4, K.pneumoniae B4958 and S.liquefaciens and all six strains and species had distinctive hybridization patterns (Fig. 2). Thus the specificity of this approach is higher than 16S rRNA gene sequencing. This result is easy to explain because NRs represent randomly selected genomic fragments while ribosomal genes are highly conserved. NotI and other recognition sites for restriction enzymes are distributed very specifically when strains and species are compared (see Table 1) (19).

Figure 2.

Microarray hybridization of NRs produced from different bacterial species to the selected NotI linking clones. Selected NotI linking clones from E.coli R2, E.cloaceae R4, K.pneumoniae B4958 and S.liquefaciens, shown at the bottom, were used for the microarrays. NR probes were prepared from six highly related species and the E.coli strains shown on the right. The hybridization pattern is specific for each of these six bacteria even though individual NotI linking clones demonstrated cross-hybridization between different bacteria.

Table 1. The number of recognition sites for rare cutting restriction enzymes in selected bacterial species.

No.	Genome	Size (Mb)	No. of recognition sites
			AscI	FspAI	NotI	PmeI	SbfI	SwaI
1	Aeropyrum pernix K1	1.6	23	11	88	29	62	15
2	Bacillus subtilis	4.2	39	123	81	89	51	176
3	Borrelia burgdorferi	1.5	0	15	1	37	8	548
4	Campylobacter jejuni	1.6	0	16	0	42	13	526
5	Chlamydophila pneumoniae AR39	1.2	2	31	2	10	21	60
6	Deinococcus radiodurans R1	3.3	197	83	15	4	28	1
7	Escherichia coli K12	4.6	164	397	23	87	68	117
8	Escherichia coli O157:H7	5.5	213	416	36	92	108	126
9	Helicobacter pylori 26695	1.7	13	112	7	35	4	67
10	Helicobacter pylori J99	1.6	12	103	14	43	4	76
11	Lactococcus lactis subsp. lactis	2.4	1	26	3	47	17	235
12	Neisseria meningitidis MC58	2.3	129	70	74	27	28	96
13	Pasteurella multocida PM70	2.3	13	100	5	61	12	195
14	Rickettsia prowazekii	1.1	0	16	1	20	10	229
15	Staphylococcus aureus Mu50	2.9	7	75	0	83	12	602
16	Streptococcus pneumoniae R6	2.0	2	28	1	25	30	51
17	Synechocystis PCC6803	3.6	6	44	44	104	40	167
18	Thermotoga maritima	1.9	5	26	10	5	14	9
19	Treponema pallidum	1.1	39	335	20	11	67	12
20	Vibrio cholerae	4.0	86	353	73	117	37	104

At the same time cross-hybridization between some NotI clones from different species was detected. It was especially strong between clones 28–36 from S.liquefaciens and NR probes from E.coli (Fig. 2). This result is easy to explain keeping two factors in mind. On the one hand, none of the four bacterial species used for the production of NotI linking clones have been sequenced and thus randomly selected clones were used for microarrays. On the other hand, our previous study (19) demonstrated that some NotI flanking sequences could be very similar in related species. However, it is almost always possible to find unique NotI clones that will hybridize only to a particular bacterial strain (see below).

The third question was whether the quality of DNA isolated from colon or feces is sufficient for the method. Experiments with samples obtained from colon or feces of germ-free mice colonized with specific bacteria demonstrated that the quality of DNA was sufficient to obtain a practically identical pattern of hybridization (Fig. 3).

Figure 3.

Comparison of microarray hybridization images produced with NR generated from in vivo (colon biopsies) and in vitro (bacterial cultures) samples. Selected NotI linking clones were used for the microarrays. Clones 52–66 were produced from K.pneumoniae B4958 and clones 2–9 from E.coli R2 genomic DNA.

It is important to mention that we have analyzed NotI flanking sequences from 20 bacterial species and did not detect identical sequences in human or rodent DNA.

RST microarrays for comparative microbiological studies

The present work is a pilot study and additional large-scale studies are important to evaluate the robustness of the method when it is scaled up. Table 1 shows that some sequences flanking restriction sites are highly specific for the particular species/strains and can be used for further analysis. Screening of 70 completely sequenced bacterial genomes for RSTS for NotI, PmeI and SbfI showed that >96% of them were species-specific (19). In a more detailed study (Table 2), the presence of strain-specific NotI flanking sequences was directly demonstrated for all sequenced bacterial strains except Chlamydophila pneumonia strains AR39 and J138. However, these very small and closely related genomes can still be differentiated using, for example, FspAI tags. Importantly, even if these or two other strains have identical short RSTS, their microarray hybridization pattern could still be different if flanking 4 or 6 bp cutters, e.g. BamHI and BglII, are polymorphic in these two strains. In this case differently sized fragments will be labeled with different efficiency (see the right fragment in Fig. 1). For example, the BamHI fragment containing SbfI site no. 7 in C.pneumoniae AR39 is 12.3 kb and in C.pneumoniae J138 this fragment is only 5.4 kb. Moreover, specific oligonucleotides can be designed for the 12.3 – 5.4 = 6.9 kb DNA fragment that will be labeled only in C.pneumoniae AR39.

Table 2. Specificity of restriction sites for the strains.

Species	Strain	Cutting sites in the genome				Unique for the species				Unique for the strain
		PmeI	SbfI	NotI	Total	PmeI	SbfI	NotI	Total	PmeI	SbfI	NotI	Total
Escherichia coli	K12 (4.6 Mb)	87	68	23	178	74	61	20	155	25	26	6	57
	O157H7 (5.5 Mb)	92	108	36	236	77	90	34	201	28	55	20	103
Helicobacter pylori	26695 (1.7 Mb)	35	4	7	46	35	4	7	46	27	2	3	32
	J99 (1.6 Mb)	43	4	14	61	43	4	14	61	35	2	10	47
Chlamydophila pneumoniae	AR39 (1.2 Mb)	10	21	2	33	10	21	2	33	0	0	0	0
	J138 (1.2 Mb)	10	21	2	33	10	21	2	33	0	0	0	0
Agrobacterium tumefaciens	Cereon (5.7 Mb)	9	258	609	876	9	251	609	869	0	0	1	1
	Dupont (5.7 Mb)	9	258	608	875	9	251	608	868	0	0	0	0
Mycobacterium tuberculosis	CDC1551 (4.4 Mb)	2	163	1214	1379	2	154	1152	1308	0	0	12	12
	MTBH37RV (4.4 Mb)	2	164	1216	1382	2	158	1153	1313	0	4	15	19
Neisseria meningitidis	MC58 (2.3 Mb)	27	28	74	129	25	28	70	123	17	4	31	52
	Z2491 (2.2 Mb)	26	30	76	132	24	29	72	125	19	3	38	60
Streptococcus pneumoniae	R6 (2.0 Mb)	25	30	1	56	25	29	1	55	11	5	0	16
	TIGR4 (2.2 Mb)	26	31	1	58	22	30	1	53	8	6	0	14
Staphylococcus aureus	Mu50 (2.9 Mb)	83	12	0	95	82	12	0	94	2	0	0	2
	M135 (2.8 Mb)	81	13	0	94	81	13	0	94	2	0	0	2

We suggest isolating, for each bacterial strain, a few NotI (and/or other) linking clones that will specifically identify that particular strain. They will represent a random selection of genomic sequences surrounding NotI sites that can be physically absent in other species or can be absent due to the polymorphism of NotI sites (Fig. 1). The polymorphism of NR may be increased by polymorphisms of 4/6 bp cutters. Therefore, by selecting specific clones that can hybridize to only one strain, very specific microarrays can be designed. Additionally, specific oligonucleotides can be designed close to NotI sites and this approach is very efficient for human genome studies (21; A.Protopopov et al., unpublished data).

Preliminary experiments also demonstrated that such NotI (RST) microarrays could be designed not only for qualitative but also for quantitative analysis. Hybridization to NotI microarrays of NRs obtained from DNA samples from four different closely related bacterial species mixed in different proportions clearly reveals this potential (Fig. 4). In this particular case, a decrease in S.liquefaciens DNA concentration is more difficult to evaluate due to the overlapping hybridization with E.coli DNA, but more careful design of the microarray could solve this problem. Results with NotI microarrays for human chromosome 3 were rather precise (21) and the fluctuations in abundance of different bacterial species in the gut are expected to be very significant compared to the hemizygous losses (i.e. two-fold copy number changes) in the human genome that were detected with NotI microarrays. That is why we think that this method will also be useful for quantitative estimates.

Figure 4.

Hybridization of NR probe prepared from bacterial mixtures to NotI microarrays. DNA from E.coli R2 (A), E.cloacae R4 (B), K.pneumonia B4958 (C) and S.liquefaciens ATCC14460 (D) were mixed in different proportions (shown on the right) and a NR probe was prepared using BamHI and BglII digestion. NotI linking clones selected from S.liquefaciens, K.pneumoniae B4958, E.cloaceae R4 and E.coli R2 and shown at the bottom were used for the microarrays.

Another problem is that a few bacterial species representing the vast majority of cells in a mixture can obscure detection and studies of other minor species. We did not address this question specifically in this study, but in a previous study with human NotI microarrays we used a modified version of CODE (21,28) to remove common sequences from normal and tumor DNA. The results demonstrated that even one cycle of NotI-CODE was sufficient to remove the majority of contaminating sequences. Thus bacterial DNA from the species present in excess can be selectively removed to allow analysis of minor species.

Contrary to the approach described above, where only carefully selected NotI clones were used to construct microarrays, another way may be even more fruitful.

In these experiments we shall produce a limited number of NotI (and other, for example PmeI) linking libraries from a few samples of colon flora. Several thousand of these linking clones will be microarrayed and used for hybridization with NR from different samples. Unknown microbial strains will be represented in these microarrays. Interesting clones will be identified and PCR primers could be used for further experiments (e.g. sequencing the whole bacterial genome).

For this purpose, the slalom approach (18) and a new DNA polymerase, φ29 (Amersham Biosciences), allowing isothermal amplification of large DNA molecules, including whole bacterial genomes (29,30), can be used. It means that RST microarrays give the possibility both for studying biodiversity and isolation of new microorganisms.

For preliminary classification of unknown bacterial sequences a bioinformatic method based on the frequency of short oligonucleotide motifs recently developed by us may be used (31).

As bacterial species are very different in nucleotide content we suggest using combinations of different types of representation probes/microarrays (NotI and PmeI) for different applications. A common database containing results of experiments with RST microarrays can be created and used for comparison of results from different groups. It is important to mention that a program can be created that will allow comparison of results with NotI (RST) microarrays and NotI (RST) passports and thus the advantages of these techniques will be combined and enhanced.