Development of Genomic Array Footprinting for Identification of Conditionally Essential Genes in Streptococcus pneumoniae

The S. pneumoniae strains used in this study were the serotype type 2 isolate D39 (3) and its unencapsulated derivate R6 (8). Lactococcus lactis strain NZ9000 (10) was used to generate pGHT7:ISS1. Escherichia coli DH5αF was used as a host for the pR412 plasmid and the generation of pR412T7. All strains were routinely cultured in tryptone yeast broth or M17 broth containing glucose (GM17) as described previously (10). General DNA techniques were performed as described previously (10). Sequencing of the DNA adjacent to the insertion sites was performed using a single-primer PCR essentially as described previously (9). An in-frame deletion of czcD (SP1857) was constructed using pORI280 as described before (10) with primer pairs czcD-KO-1/czcD-KO-2 and czcD-KO-3/czcD-KO-4 (Table 1), which removed 852 bp of the open reading frame (ORF); a detailed description of its construction is provided elsewhere (T. G. Kloosterman, M. M. Pol, J. J. E. Bijlsma, and O. P. Kuipers, submitted for publication).

A pGh9:ISS1 derivative with two outward-facing T7 promoters in the ISS1 element was constructed as follows. The ISS1 element from pGh9:ISS1 (21) was removed by digestion with EcoRI and HindIII and replaced by a HindIII-, EcoRI-digested PCR product of the ISS1 element generated on pGh8:ISS1 (21) with primers ISS1_T7_up (Table 1) and ISS1_T7_down (Table 1). This resulted in pGh9T7:ISS1 which contains the ISS1 element (with the T7 promoters on both sides) in an orientation opposite to that in the pGh9:ISS1 plasmid.

The backbone of transposon donor plasmid pR412 (23), including the inverted left repeat and inverted right repeat of the transposon, was amplified with Pwo DNA polymerase (Roche) with a single phosphorylated primer PBMrIRPi (Table 1); PCR cycling conditions were as follows: 94°C for 4 min; 30 cycles of 94°C for 1 min, 55°C for 1 min, and 68°C for 5 min; and 68°C for 5 min. The spectinomycin-resistance cassette was PCR amplified with primers containing a T7 tag, i.e., PBMrStartT7 (Table 1) and primer PBMrEndT7 (Table 1); PCR cycling conditions were as follows: 94°C for 4 min; 30 cycles of 94°C for 1 min, 55°C for 1 min, and 68°C for 1.5 min; and 68°C for 4 min. The two PCR products were ligated to obtain the transposon donor plasmid pR412T7.

Random transposon libraries were routinely generated in S. pneumoniae as described before, using the in vitro mariner transposon (1) and the in vivo pGh:ISS1 system (21). Recombinant HimarC9 transposase was purified from the E. coli strain BL21(DE3) (Novagen) containing plasmid pET29C9 (13) essentially as described before (14). However, to prevent DNase contamination of the transposase stock, the DNase I treatment of the purification step was omitted from the protocol. For in vitro transposition, 1 μg of purified chromosomal DNA was incubated with 0.5 μg pR412T7, which contains the marinerT7 transposon, and 0.5 μg purified recombinant HimarC9 transposase in a 20-μl reaction mixture. After repair of the transposition reaction with T4 DNA polymerase and E. coli DNA ligase (7), the DNA was used for transformation of strain R6 (1, 13). After overnight growth on selective plates containing 120 mg liter⁻¹ spectinomycin, colonies were scraped, pooled, and stored at −80°C. For the construction of pGh9T7:ISS1 mutant libraries (21), the R6 and D39 strains were transformed with the pGh9T7:ISS1 plasmid and grown at 30°C on selective plates containing 0.25 mg liter⁻¹ erythromycin. Several colonies of S. pneumoniae containing the pGh9T7:ISS1 plasmid were pooled, grown overnight in GM17 broth with erythromycin at 30°C, and stored at −80°C. From this stock, bacteria were inoculated for overnight growth in GM17 broth at 30°C with the appropriate antibiotic, and the next day the cultures were diluted 1 to 50 in GM17 broth without antibiotics and after three hours of growth shifted to 37°C. After another three hours of growth at the nonpermissive temperature, 10⁻³ and 10⁻⁴ dilutions were plated on selective medium. Approximately 20,000 to 30,000 CFU were scraped from the plate, pooled, and stored at −80°C for further use. The same marinerT7 and pGh9T7:ISS1 transposon libraries were used for all experiments to determine the reproducibility and specificity of each method. New libraries were generated for the comparison of the marinerT7 and pGh9T7:ISS1 transposon libraries and the GAF zinc screen.

S. pneumoniae microarrays were produced essentially as described previously (12, 40) and contained amplicons of on average 600 bp representing 2,087 ORFs of S. pneumoniae TIGR4, each of which are present twice on the array (36) as an internal control to monitor variations in hybridization efficiency per slide. Purified RNA obtained after in vitro transcription was used to generate fluorescent DNA probes by direct or indirect labeling using standard methods. Hybridization, washing, and acquisition of array images were performed as described previously (11). Spots were screened visually to identify those of low quality. The amount of low-quality spots varied per slide but was never more than 10%; they were removed from the data set prior to analysis. Slide data were processed and normalized using MicroPreP (41). A net mean intensity filter based on hybridization signals obtained with amplicons representing open reading frames unique for S. pneumoniae strain TIGR4 (36) was applied in all experiments. For one-array experiments (Fig. 2 and 3), only genes that had reliable measurements (i.e., passed filters) for both duplicate spots present on the array were analyzed. For the other experiments, microarray data were analyzed when at least three measurements were available, using a CyberT implementation of the Student's t test (20). False discovery rates were calculated as described previously (40). The correlation between normalized signal intensities per gene in both channels was determined by linear regression analysis.

For the identification of transposon insertion sites, microarray tracking of transposon mutants (MATT) was performed as described previously (32) with some modifications. A linear PCR with AccuPrime (Invitrogen) polymerase was performed on chromosomal DNA containing pGh9T7:ISS1 insertions as a template and the MATT1 (Table 1) primer located on the transposon, for 30 cycles at 56°C. The single-stranded fragments were purified with a GFX column (GE Healthcare), and 100 ng was used as a template for a semirandom PCR in a 20-μl reaction mixture containing 2 μM each of an anchored random primer (CEKG2A, CEKG2D, or CEKG2E; Table 1) and MATT1 and 1 U Taq polymerase (Invitrogen) with the following cycling reactions: 1 cycle at 94°C for 4 min; 12 cycles at 94°C for 30 s, 42°C for 30 s (−0.5°C each cycle), and 72°C for 2 min; and 30 cycles at 94°C for 30 s, 65°C for 30 s, and 72°C for 2 min. The reaction products were diluted fivefold, and 1 μl was used as a template for a subsequent reaction using a nested transposon-specific primer MATT2 (Table 1) and the anchor primer CEKG4 (Table 1) in a 100-μl reaction mixture using the following cycling conditions: 1 cycle at 94°C for 4 min, and 30 cycles at 94°C for 30 s, and 56°C for 30 s, and 72°C for 2 min. The PCR products were purified using a Roche PCR purification kit and used as input for a T7 RNA polymerase reaction using a T7 MEGAscript kit (Ambion).

Genomic DNA was diluted in 20 μl H₂O to 0.5 μg/μl and fragmented by sonication with the microtip of a Branson digital sonifier (eight times for 0.5 s, output 20%). To prepare double-stranded, blunt-end DNA, fragmented DNA was successively treated with S1 nuclease, T4 DNA polymerase, and E. coli DNA ligase. To facilitate the ligation of a TOPO-isomerase-activated linker (TOPO-Walker kit; Invitrogen), the blunt-end, double-stranded DNA was incubated with calf intestinal phosphatase for dephosphorylation and with Taq DNA polymerase (Thermoperfect; Integro) to create 3′ A-tailed DNA ends. Ligated DNA was used as a template in two consecutive PCRs with Taq DNA polymerase: the first reaction contained primers LinkAmp1 (Table 1) and PBMrTn5 (Table 1), and the second contained nested PCR primers LinkAmp2 (Table 1) and PBMrTn3 (Table 1). PCR cycling conditions for both reactions were as follows: 94°C for 4 min; 20 cycles of 94°C for 30 s, 63°C for 1 min, and 72°C for 1.5 min; and 72°C for 3 min. PCR products were loaded on a 2% agarose gel, and DNA fragments with the preferred size were excised and extracted with a Qiaquick gel purification kit (QIAGEN). In vitro transcription reaction and subsequent DNase I digestion were performed with a T7 MEGAscript kit (Ambion).

The TraSH method was performed essentially as described previously (33) with a few modifications. Genomic DNA was digested with ApoI or TacI and precipitated and ligated overnight in a 15-μl volume with an approximately 1,000-fold molar excess of the appropriate adapter (Adapter1 for ApoI- and Adapter2 for TacI-digested DNA; Table 1). The ligated products were purified, and 0.3 to 1 μl was used as a template in the following PCR containing primer TmR1 (Table 1): 1 cycle at 95°C for 1 min; and 5 cycles at 95°C for 30 s, 72°C for 30 s, and 72°C for 1 min. Subsequently the appropriate adapter primer (Adapter3 for ApoI- and Adapter4 for TacI-digested DNA; Table 1) was added, and the reaction was continued with the following cycling conditions: 5 cycles at 95°C for 30 s, 72°C for 30 s, and 72°C for 1 min, and 30 cycles at 94°C for 30 s, 65°C for 30 s (−0.3°C each cycle), and 72°C for 70 s. PCR products were purified and used as a template for a T7 RNA polymerase reaction using a MEGAscript kit (Ambion).

The procedure was performed essentially as described previously (6). Genomic DNA (>10 μg) was digested overnight with TacI, DdeI, or AluI, and the purified DNA was used for an in vitro transcription reaction with a MEGAscript T7 RNA polymerase kit (Ambion) in a total volume of 20 μl.

Aliquots of a marinerT7 transposon library generated in S. pneumoniae R6 containing approximately 20,000 independent mutants were diluted 1 to 20 in GM17 broth or in GM17 broth containing 0.5 mM ZnSO₄. Both cultures were grown for 10 generations as determined by the optical density at 600 nm, after which the bacteria were spun down (9,000 × g, 4 min) and used for the extraction of genomic DNA.

The microarray data have been deposited in the NCBI Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/ under GEO Series accession numbers GSM158348, GSM158350, GSM158558, GSM158562, GSM158563, GSM158564, GSM158565, GSM158566, GSM158567, GSM158568, GSM158569, and GSM158570.

Important considerations for the development of GAF were that the method should be sensitive enough to detect all mutants in a library and should be reproducible. Reproducibility, specificity, and sensitivity are in large part determined by the amplification and detection step (Fig. 1). Therefore, we tested the performance of various methods for transposon insertion site detection (6, 32, 33) on both small (<100 CFU) and large (>20,000 CFU) mutant libraries.

Although we used strains R6 and D39 throughout this study, we wanted our method to be applicable to (in principle) all S. pneumoniae strains, some of which are not as easily transformable. Therefore, we established two methods for the generation of random mutant libraries: in vitro mariner mutagenesis (1), which depends on a high-transformation efficiency of the accepting strain for the generation of large libraries and is not expected to have polar effects on downstream genes (31), and the in vivo pGh:ISS1 system (21), which consists of a plasmid with a temperature-sensitive origin of replication, the ISS1 insertion element, and an erythromycin resistance gene with a weak terminator (−7.5 kcal mol⁻¹); thus, insertion of pGh:ISS1 could cause some polar effects. Growth of bacteria that contain the plasmid for several generations at 30°C, the permissive temperature, followed by a temperature shift to 37°C, induces the insertion of the plasmid into the chromosome. Thus, a one-time introduction of this plasmid in a strain with low transformation efficiency is sufficient to generate large (>20,000 CFU) libraries, avoiding the need for high-transformation efficiencies in accepting strains. In each transposon, an outward-facing T7 RNA polymerase promoter close to both ends of the insertion element was introduced for amplification purposes. Both transposons readily generated large (>20,000 CFU) libraries in S. pneumoniae strains R6 and D39, as expected, and in strain TIGR4 with the pGh:ISS1 system (34) as well (results not shown). Mutant libraries generated with these transposons were used exclusively throughout this study, and the RNA generated with the aid of the T7 RNA polymerase promoter was used in our standard microarray procedures.

Microarray tracking of transposon mutants combines two rounds of PCR: first, a reaction with specific and nonspecific primers, followed by a nested PCR to further amplify the material (32). The specific primers hybridize to sequences on the transposon, and the nonspecific primers contain 10 degenerative nucleotides followed by a unique anchor sequence. Several anchor sequences were chosen that are present on average every 400 bp in the genome. MATT was found to be specific and reproducible when used for artificial GAF experiments on a few mutants (comparison of 12 mutants with the same set in which three mutants were removed; results not shown). However, when two parallel reactions on one large (>20,000 CFU) genome-wide library were compared, it was obvious that the method was not reproducible (Fig. 2A), as few genes displayed equal signal intensities in both channels (r², 0.036). When the products of the second round of PCR were visualized on an agarose gel, several distinct bands of various sizes in each reaction were visible, indicating that some DNA fragments were selected randomly during the PCR steps (results not shown).

Next, we tested a method that consists of the ligation of a TOPO-isomerase-activated adapter to sheared chromosomal DNA, followed by two PCR steps using nested primers that anneal specifically to the adapter and the transposon. Similar to MATT, this method was specific and reproducible when applied to artificial GAF experiments on a small number of mutants (i.e., comparison of 21 mutants with the same set from which 6 mutants were removed; results not shown) but was clearly nonreproducible when used on a genome-wide library (Fig. 2B). When the PCR products of separate reactions were examined, on an agarose gel, different distinct banding patterns were visible, again indicating that the PCR steps introduced a bias in the whole procedure. Repeated attempts to improve the PCR steps in both the MATT and the TOPO-adapter method by using Taq polymerase variants with increased specificity, touchdown PCR, and even more stringent annealing conditions were unsuccessful (results not shown).

Subsequently, we tested the TraSH method, which consists of the ligation of an adapter to chromosomal DNA digested with TaqI or ApoI, followed by a short linear PCR with a transposon-specific primer and a PCR with transposon-specific and adapter-specific primers (33). This method was fairly specific and reproducible both when used on a small number of mutants (i.e., comparison of 21 mutants with the same set of mutants from which 6 mutants were removed; results not shown) and on a large (>20,000 CFU) genome-wide library (Fig. 2C), as the majority of the genes had comparable signal intensities in both channels on two parallel reactions (r², 0.83). However, the optimal annealing temperature and template concentration had to be determined for each new sample, as standardized PCR conditions resulted in distinct banding patterns in the PCR products and lower correlations between signal intensities in each channel. As the continuous optimization of the PCR conditions made this method cumbersome, we did not pursue it further.

To avoid the bias introduced by methods employing PCR, we directly used purified chromosomal fragments digested with TaqI (which on average cuts every 442 bp) as input for an in vitro transcription reaction using the T7 RNA polymerase promoter that we had introduced in both transposons (Fig. 1) (6, 17). Comparison of duplicate reactions on a small set of 90 mutants with this method gave rise to specific and reproducible results (data not shown). Next, we tested the feasibility of this method by performing an artificial GAF experiment. To a small library of 90 mutants, 9 defined mutants (transposon insertion site determined by sequencing) were added (90 + 9 library). Subsequently, transposon insertion sites in both libraries were determined by microarray analysis as described above. As expected, the signal intensities of most genes were found to be similar in both channels (i.e., ratios between 0.5 and 2), indicating that mutants in these genes were present in both populations. Furthermore, the signal intensities of 13 genes in the 90 + 9 library were increased significantly in one channel (Fig. 3A), leading to ratios of >9 (Table 2), indicating that transposon insertions in or near these genes were not present in the 90 mutant library. All 13 of these genes corresponded to ORFs present on the TaqI fragments of 8 defined mutants added to the library (Fig. 3A; Table 2). The corresponding amplicons of the genes on the TaqI restriction fragment in one defined mutant (SP2129 to SP2131) had high signal intensities in both channels (ratio of 1), indicating that a mutant with a transposon insertion in this location was already present in the 90 mutant library. Three additional genes were detected as having been added to the library (Fig. 3A; Table 2). The detection of SP1792 is probably due to the fact that the amplicon has significant homology with, among others, ORF SP1262 (data not shown), and thus the high Cy5 signal is probably due to cross hybridization with the fragments generated by the added mutant A2. SP0001 is adjacent to ORF SP2240 on the genome and is present on the TaqI fragment of added mutant A10, and the intermediate-level signals in the Cy5 channel could be due to the occurrence of some partially digested fragments. It is not clear why gene SP2169 is detected; however, the signals in both channels and the ratio are close to the arbitrarily chosen cutoff value of 2 and differ substantially from the signal intensities of the amplicons that represent the 9 added mutants.

Subsequently, this procedure was tested on one genome-wide marinerT7 library (>100,000 CFU) grown for approximately 20 generations in GM17. Three different restriction enzymes were used, TaqI, DdeI, and AluI, which have an increasing number of sites in the S. pneumoniae genome; on average, they cut every 442, 337, and 204 bp, respectively. The scatterplots derived from the analysis of each separate hybridization (Fig. 3B, C, and D) clearly show a strong correlation between the signal intensities in each channel for each gene (r² of >0.96 in all cases) (the normalized signal data and CyberT output data for these arrays are available at http://molgen.biol.rug.nl/publication/GAF_data/ ). Restriction with DdeI resulted in the detection of 1,301 genes, and the numbers for the TaqI and AluI digests were comparable (1,193 and 1,265, respectively). Therefore, the three duplicate hybridizations were combined for further analysis. The library was generated in R6, which contains 2,116 ORFs, 1,995 of which are homologous to ORFs in strain TIGR4, and thus are present on our array (8). As expected, all amplicons that represent ORFs present in strain TIGR4 but not in R6 (36) (TIGR4 specific) had low signal intensities. These signals were used to manually generate a background filter that was applied to the analysis of the data from all three restrictions. Of the 1,995 R6 ORFs on the array, 1,322 (66%) were detected in at least two restrictions and had a less than twofold ratio difference, demonstrating the reproducibility of this method. In principle, no mutants in genes that are essential for the growth of S. pneumoniae can be present in the library; thus, amplicons representing these genes should not give signals above background (i.e., should be classified as nondetected). Comparison of the 673 nondetected R6 genes with the reported essential genes of S. pneumoniae (1, 5, 15, 18, 25, 35, 37, 43) showed that 80 had been designated as essential and 97 were located adjacent to putative essential genes (see Table S1 in the supplemental material). These analyses showed that this method was robust and reproducible for incorporation in GAF and that all three restriction enzymes can be used. The latter is important, as not all restriction enzymes cut in all genes and the use of several restriction enzymes on the same DNA improves the chance of detecting all mutants that have disappeared from a population.

Previous experiments were performed on marinerT7-generated libraries. To test whether the method could also be applied to pGh9T7:ISS1 libraries, and to assess the level of saturation and randomness achieved by each transposon, we compared a large (>20,000 CFU) marinerT7 library with a pGh9T7:ISS1 one (>20,000 CFU), each grown for approximately 20 generations in GM17. After isolation, the DNA was digested with DdeI and AluI and each detection reaction was performed in duplicate. Microarray data were analyzed as before by using the background filter generated with the TIGR4-specific amplicons (the normalized signal data and CyberT output data for these arrays are available at http://molgen.biol.rug.nl/publication/GAF_data ). This analysis resulted in the detection of 1,473 genes, and the nondetected genes were similar to those not detected in the analysis of the three marinerT7 digests. The ratio of the majority of the genes was between 0.5 and 2, indicating that they are mutated by both transposons. However, 139 (9%) genes had a ratio lower than 0.5 or higher than 2, indicating that these genes are only, or preferentially, mutagenized by one of the two transposons. These “hot spots” are distributed throughout the genome (Table 3). There were 60 genes preferentially mutated by pGh9T7:ISS1, whereas 79 were preferentially mutated by marinerT7, indicating that it has a slightly higher saturation rate than pGh9T7:ISS1. Anchored PCR using primers located on some of the ORFs that had a ratio close to the cutoff point of 0.5 indicated that these genes were indeed preferentially hit by either transposon and that these ratios were not due to slight differences in, for instance, T7 RNA polymerase efficiency (data not shown).

As all components for GAF were successfully developed, we tested its ability to identify conditionally essential genes from a library grown under in vitro stress conditions. Although Zn²⁺ is an essential ion, it is toxic to bacteria in high concentrations. Bacteria often contain specific proteins for Zn²⁺ secretion (26), but none have been identified in S. pneumoniae so far. The concentration of Zn²⁺ in serum is reported to be around 15.3 μM (1 mg liter⁻¹) and in lung tissue 229.4 μM (15 μg/g wet tissue) (42), and during inflammation concentrations increase in blood and other body sites (24, 38), indicating that S. pneumoniae is likely to encounter Zn²⁺ stress during infection. Therefore, we decided to screen for genes essential for the survival of Zn²⁺ stress. In addition, a zinc exporter had already been identified by other means by our group (T. G. Kloosterman et al., submitted). A large (>20,000 CFU) marinerT7 library in strain R6 was grown in quadruplicate in GM17 or GM17 containing 500 μM ZnSO₄, which is half the MIC of zinc for S. pneumoniae (data not shown), for approximately 10 generations. Subsequent detection of transposon insertion sites in both populations using a TaqI restriction showed that gene SP1857 and a cluster of two genes, SP0856-SP0857, had a significantly lower ratio (Table 4), suggesting that these are essential genes for growth in high-Zn²⁺ concentrations. SP0856-SP0857 are annotated as ilvE and as an oligopeptide-binding protein, respectively. SP1857, annotated as czcD, encodes a protein homologous to cation exporters (36) and was the exporter already identified as being involved in Zn²⁺ resistance (T. G. Kloosterman et al., submitted). A mutant generated in this gene was indeed unable to grow in GM17 containing 500 μM ZnSO₄ (Fig. 4A and B). Although the mutant was at least ten times more sensitive to ZnSO₄ than the parent strain, the ratio in the GAF experiment was 0.43. In silico TacI restriction analysis of the R6 nucleotide sequence of the homologue of SP1857 and the chromosomal region surrounding it results in a DNA fragment that contains SP1857, its promoter, and part of the adjacent gene, SP1858 (Fig. 4C). As polar effects are unlikely to occur, mutants containing a transposon inserted in front of the czcD ORF probably have no growth defect in GM17 plus ZnSO₄ and will not disappear from the population. The presence of such mutants in the library will result in a DNA fragment that also hybridizes to the SP1857 amplicon, which can explain why the ratio was close to 0.5. However, this ratio was found to be significant by statistical analysis, and the phenotype of the deletion mutant confirmed our GAF findings. Thus, we have successfully developed GAF, as it correctly identifies conditionally essential genes in S. pneumoniae.