- Email: [email protected]
Locating transposable element polymorphisms in bacterial genomes
Journal of Microbiological Methods 53 (2003) 355 – 363 www.elsevier.com/locate/jmicmeth
Locating transposable element polymorphisms in bacterial genomes Hasan Yesilkaya a,*, Anne Thomson a, Christine Doig b, Brian Watt b, Jeremy W. Dale c, Ken J. Forbes a b
a Department of Medical Microbiology, University of Aberdeen, Medical School Building, Foresterhill, Aberdeen AB25 2ZD, UK Scottish Mycobacteria Reference Laboratory, New Royal Infirmary of Edinburgh, Lothian Universities Hospitals NHS Trust, Little France, Old, Dalkeith Road, Edinburgh EH16 4SU, UK c Molecular Microbiology Group, School of Biomedical and Life Sciences, University of Surrey, Guildford, Surrey GU2 7XH, UK
Received 30 August 2002; received in revised form 6 December 2002; accepted 6 December 2002
Abstract Although whole-genome sequencing is greatly extending our knowledge of the genetic capacity of those bacterial species, it is only directly informative for the particular strain sequenced. Many bacterial species exhibit more or less genetic polymorphism within their populations and characterising this variety is an extremely important way of elucidating the biology of these species. Often genomic polymorphisms are associated with multicopy elements, particularly transposable elements. We describe a novel method that efficiently characterises the sequences of such polymorphisms. We have optimised heminested inverse PCR (hINVPCR) to assess the diversity of insertional polymorphisms of a transposable element (IS6110) in clinical isolates of Mycobacterium tuberculosis. To increase the yield of information, genomic DNA was digested with different endonucleases (Bsp1286I, HaeII or PvuI), and primers based on both the 5Vand 3Vends of IS6110 were used to amplify and determine the genomic sequence upstream (or downstream) of the transposable element. We found that both the choice of restriction enzyme and the use of primers at both ends of the transposable element significantly increased the diversity of the insertion sites identified. Band stabbing was incorporated into the method as an alternative to cloning in order to screen large number of isolates at a sequence level in a rapid and labour-efficient fashion. We describe some of the purposes to which such data can be put. D 2002 Elsevier Science B.V. All rights reserved. Keywords: Bacteria; Genome polymorphism; In silico analysis; IS6110
1. Introduction The availability of whole-genome sequences of different bacterial species is ever increasing (http:// * Corresponding author. Tel: +44-1224-552864; fax: +44-1224685604. E-mail address: [email protected] (H. Yesilkaya).
www.ncbi.nlm.nih.gov/PMGifs/genomes/micr.html) and is greatly extending our understanding of microbial biology. Such resources are, however, restricted to single, hopefully representative, strains of the species, or at best a few such strains. The genetic diversity across the whole population of a species will not be revealed in such studies. Given the central role played by multicopy and transposable elements in
0167-7012/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved. doi:10.1016/S0167-7012(02)00256-7
356
H. Yesilkaya et al. / Journal of Microbiological Methods 53 (2003) 355–363
genomic rearrangements and in horizontal gene transfer (Mahillon et al., 1999), the determination of the sequences flanking these, be they in prokaryotic or eukaryotic species, provides a powerful tool for mining such biodiversity. Insertion sequences (IS) are 700 – 2500 bp long DNA elements that can translocate in the genome; they can cause spontaneous mutations, the physical disruption of genes, and are able to alter gene expression. In addition, IS can promote chromosomal rearrangements through inversions, translocations, duplications and deletions (Chalmers and Blot, 1999). IS in bacteria have been utilised in studies of microbial evolution (Fang et al., 1998), epidemiology (Bik et al., 1996), microbial virulence (Finlay and Falkow, 1997), and employed as tools in molecular biology (Cox et al., 1999). All such studies require an understanding of the dynamics of the transposition event (Wall et al., 1999) and how the target sites are selected; however, little is known about such activity in natural populations. IS6110 is found in Mycobacterium tuberculosis clinical isolates at between 0 and 25 copies per strain and it is thought to have little or no specificity at the nucleotide level for insertion in the M. tuberculosis genome (McAdam et al., 2000), although there are preferred regions (Fang and Forbes, 1997; Gillespie et al., 2000). IS6110 restriction fragment length polymorphism (RFLP) is the principal method for strain differentiation in M. tuberculosis. Using this as our model, we have developed a method based on heminested inverse PCR (hINVPCR) that allows the characterisation of IS-induced genomic polymorphism in large bacterial collections, without the need for cloning.
2. Materials and methods 2.1. Bacterial strains The M. tuberculosis clinical isolates tested in this study were received by the Scottish Mycobacteria Reference Laboratory (Edinburgh, Scotland, UK) between 1991 and 1997 (n = 52). DNA was extracted by growing the isolates in Middlebrook 7H9 broth in 50 ml centrifuge tubes at 37 jC for about 4 weeks according to a standard protocol (van Embden et al., 1993). 2.2. IS6110 restriction fragment length polymorphism (RFLP) analysis The isolates had a mean IS6110 copy number of 9.7 (range 2– 18) as assessed by RFLP (van Embden et al., 1993). Relatedness of IS6110 fingerprint patterns was calculated using the Dice coefficient of similarity (BioNumerics, version 2.0; Applied Maths, Kortrijk, Belgium) and isolates with patterns of < 70% relatedness were selected for hINVPCR in order to maximise the recovery of novel insertion sites. 2.3. Primers The IS6110 primer sequences were designed using OLIGO (version 5.0, National Bioscience, Plymouth, MN), and synthesised by Interactiva (Ulm, Germany). The relative positions of the primers are given in Fig. 1, and the sequence of the primers from 5V- to 3V-end as follows: PB: GTCTACTTGGTGTTGGCTGC (nt 170 – 189); PA1: GGGTCATGTCAGGTGGTTCA (nt 47 – 66 in the complementary strand); NP: GTCA-
Fig. 1. Location of primers and restriction endonuclease recognition sites in IS6110 (open line). Nucleotide positions of the restriction sites in IS6110 are indicated.
H. Yesilkaya et al. / Journal of Microbiological Methods 53 (2003) 355–363
357
Fig. 2. Flow diagram of hINVPCR. Procedure (for PvuI) showing the relative positions of IS6110 (open box) and the flanking DNA (thin line).
358
H. Yesilkaya et al. / Journal of Microbiological Methods 53 (2003) 355–363
GCACGATTCGGAGT (128 – 155 in the complementary strand); INS2: TTTGTCACCGACGCCTACGC (nt 855 –874 in the complementary strand); INV1: CTGGGTCGACTGGTTCAAC (nt 1202 – 1220); INV2: TACGCTCAACGCCAGAGAC (nt 1281 – 1299); INV3: CCTATGACAATGCACTAGCCG (nt 1093– 1113 in the complementary strand); P8: CTGAACCGTGAGGGCATCGAG (nt 624 – 644 in the complementary strand); P9: GGTTTGCGGTGGGGTGTCGAG (nt 426– 443). 2.4. Template preparation for PCR The strategy used was adapted from Patel et al. (1996) with some modifications (Fig. 2). For each sample, 100 – 200 ng genomic DNA in a 20-Al volume was digested with 2.5 U of Bsp1286I, HaeII or PvuI (New England Biolabs, Hitchin, UK) then heat-inactivated. Several criteria were considered in the choice of these enzymes. Firstly, the use of more than one restriction enzyme, each of which has a different recognition sequence, gives a better sampling across the genome than a single restriction enzyme alone could do, simply by virtue of the different, randomly distributed recognition sites in the genome. Secondly, the enzymes should cleave the DNA with a staggered cut to produce the cohesive-ended DNA fragments that ligate more efficiently than blunt-ended ones (Ochman et al., 1998). Thirdly, the selected enzymes should not cut the IS6110 sequence too close to the ends of the IS
leaving sufficient IS6110 sequence in which to place the primers required for the hINVPCR, but not so distant as to excessively increase the length of the self-ligation products (see below). This protocol allowed us to use the digested – ligated mixtures of genomic DNA as the template for PCR amplifications from both ends of the IS, thus reducing the cost and labour of the experiment, and increasing the probability of detecting more insertion sites in the genome. Our fourth consideration in the selection of the restriction endonucleases was the distribution of the restriction enzyme sites in the M. tuberculosis genome. In silico digests of the H37Rv genome indicated an even distribution of the selected recognition sites along the genome (Fig. 3). However, the lengths of the fragment must also be considered since there is a negative correlation between efficiency of self-ligation and the length of the DNA fragment, particularly for fragments above 3 kb (Ochman et al., 1998). The restriction enzymes must digest the genome into as many fragments as possible that are in the appropriate size range for hINVPCR. The restricted fragments were self-ligated at a concentration of 0.5 Ag ml 1 DNA in 20 Al of ligation buffer (stock: 50 mM Tris – HCl (pH 7.6), 10 mM dithioerythritol, bovine serum albumin 500 Ag ml 1) in the presence of 20 U ml 1 T4 ligase (Roche Diagnostics, Mannheim, Germany) by incubating at 16 jC for 16 h in a thermocycler with a hot lid. Heminested inverse PCR (hINVPCR) amplification used the protocol described previously (Patel et al.,
Fig. 3. Distribution of Bsp1286I, HaeII, and PvuI fragment lengths along the H37Rv genome by fragment size class (x Bsp1286I, 5 HaeII, E PvuI).
H. Yesilkaya et al. / Journal of Microbiological Methods 53 (2003) 355–363
1996). Second-round PCR product (8 Al) was electrophoresed (at least for 6 h at 1.3 V cm 1) in 2% Metaphor agarose (FMC BioProducts, Rockland, ME) and the resolved fragments visualised after ethidium bromide staining under UV light. Multiple bands were observed and the banding patterns were different for each isolate/restriction enzyme/primer combination (Fig. 4). Each band was stabbed with a hypodermic needle (Bjourson and Cooper, 1992) and the DNA adhering to the needle washed into fresh (secondround) PCR reagent and re-amplified using the same PCR conditions. Those re-amplified fragments that had a minimum of background amplification were purified to remove contaminating primers and dNTPs (Centricon Centrifugal Devices, Millipore, MA) and sequenced as described previously (Fang et al., 1998). The primers used for sequencing upstream (5V) of IS6110 were PB (for Bsp1286I, HaeII) or P9 (for PvuI), and for sequencing downstream (3V) of IS6110
359
were P8 (for Bsp1286I), INS2 (for HaeII) or INV3 (for PvuI). 2.5. Sequence analysis The sequences were mapped to the whole genome of the sequenced strains of M. tuberculosis H37Rv (http://www.sanger.ac.uk) and CDC1551 (http://www.tigr.org ) using BLA ST (http:// www.ncbi.nlm.nih.gov) and ARTEMIS (Rutherford et al., 2000).
3. Results Experimentally we observed that the hINVPCR amplicon sizes ranged from 100 to 1000 bp (Fig. 4), therefore the proportion of the genome that is accessible to hINVPCR can be estimated from in silico
Fig. 4. Representative gel of hINVPCR amplicons for two strains, 34 and 44, obtained with three different enzymes (Bsp1286I (B), HaeII (H) and PvuI (P)), and primer sets, 5Vand 3Vends, respectively. Size standards (lanes kb) were 1 kb DNA size markers (Gibco/BRL), and fragment sizes are indicated in kilobases.
360
H. Yesilkaya et al. / Journal of Microbiological Methods 53 (2003) 355–363
Table 1 Overall study results Restriction endonuclease used to digest genomic DNA Bsp1286I
HaeII
PvuI
In silico digest of H37Rv Percent genome in 3507726 bp 3339986 bp 1095798 bp fragments between (79.5%) (75.7%) (24.8%) 100 – 1000 nt long Number of 100 – 1000 < 10 043a 9353 2289 nucleotide long fragments hINVPCR bands Total number From 5Vend From 3Vend Insertion sites identified at nucleotide level a
269 138 131 105
548 258 294 268
429 254 171 235
See text.
digests of the H37Rv DNA sequence as the proportion of such sized fragments of the total produced by that enzyme (Fig. 3). For Bsp1286I, 79.5% (3507726 bp) of the genome sequence fell within the amplifiable size range (Table 1). Similarly, HaeII would allow 75.7% (3339986 bp) and PvuI 24.8% (1095798 bp) of the H37Rv genome to be amplified by hINVPCR. In the case of Bsp1286I, a lower yield of hINVPCR bands than this would be expected since this enzyme has multiple recognition sequences and therefore several alternative sticky-end sequences; there would only be ligation for those fragments which had the IS6110 recognition sequence (GAGCTVC at the 5Vend and GTGCCVC at the 3Vend) at both ends of the fragment, and so only PCR amplification using IS6110 primers could occur from a subset of the IS6110-containing Bsp1286I fragments. In silico analysis predicted that HaeII would produce the most hINVPCR fragments, PvuI somewhat less and Bsp1286I much less. The 52 clinical isolates screened here produced 1246 amplified hINVPCR bands (Table 1): HaeII (548 bands, 44%), PvuI (429 bands, 34%) and Bsp1286I (269 bands; 22%). Predicted and observed yields were therefore in agreement with each other. Similar numbers of hINVPCR fragments were generated from the 5Vand 3Vends of IS6110 for all of the enzymes. A somewhat preferential amplification
from the 5Vend with PvuI digests generated rather more (60%) bands (Table 1), but this was probably artifactual since subsequent tests on more isolates (82) reduced this excess to 56%. All of the 1246 amplified hINVPCR fragments were band stab re-amplified and these purified products visualised by gel electrophoresis. Thirtytwo percent (402) of these re-amplified products had background smears, possibly attributable to the presence of single stranded or mismatched amplicons (Patel et al., 1996), or comigrating bands (data not shown). Although a very simple and effective method, we found the success of band stabbing to be very dependent on the good electrophoretic separation of the amplicons and this was optimised here by the use of high resolution agarose (2% Metaphor agarose gel) and by extended electrophoresis times. Of the remaining 844 (68%) products that were sequenced, 608 (49% of the total hINVPCR bands) yielded IS6110 flanking sequences which could be mapped to the H37Rv or CDC1551 genomes. Eighty-five percent (516) of these sequences were of sufficient quality (generally a sequence run of >50 bp with sequence identity >80%) to locate the IS6110 insertion to a unique nucleotide position in the genome. The remaining 15% (92) of sequences, although of poorer quality, were adequate to locate the insertion site to within tens of base pairs. The sequence efficiency, ratio between the number of sequenced fragments and the number of sequence data available for analysis, was higher for the bands amplified from the 5Vend of IS6110 (93%) compared to the 3Vend of IS6110 (56%), and it is likely that this was methodological in origin, perhaps due to greater misamplification of fragments from the 3Vend of the IS. In total, 312 insertion sites representing 189 unique sites were identified (Table 2), with all of the IS6110 insertion sites being identified in nearly half of the isolates (n = 24) (characterised from either 5V- or 3Vend of the element). In one quarter of the isolates (n = 14), the number of unique insertion sites identified exceeded the copy number of IS6110 identified by RFLP (Table 2). This was most probably due to the low resolution of the RFLP method in separating comigrating bands, less plausibly to rare transposition events during culture (de Boer et al., 2000). Of the
H. Yesilkaya et al. / Journal of Microbiological Methods 53 (2003) 355–363 Table 2 Strain-specific efficiency of hINVPCR for identification of IS6110 insertion sites Isolate
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50
Number of IS6110 5 10 8 12 9 6 8 12 8 6 9 8 9 7 5 11 8 9 7 7 8 10 12 11 9 7 6 13 11 12 12 9 10 11 17 14 7 9 15 8 2 18 7 14 16 9 18 12 15 12
Number identified by hINVPCR 7a 10 9a 8 9 5 8 11 8 8a 9 8 10a 5 3 6 9a 8 7 7 8 7 5 6 0 8a 8a 12 10 7 8 7 6 11 13 11 13a 12a 12 9a 3a 13 10a 13 12 12a 11 10 9 9
Number identified at 5Vend 3 6 6 5 6 4 4 6 6 6 6 6 7 4 2 3 6 7 6 5 5 3 4 4 0 5 5 11 7 4 5 4 5 7 7 6 9 6 8 9 2 8 8 8 9 6 7 6 7 4
Number identified at 3Vend 2 7 4 6 4 3 5 5 3 2 4 5 6 2 2 4 5 1 4 3 5 4 1 1 0 4 4 3 6 4 4 4 2 7 6 6 5 6 7 1 2 7 3 7 6 7 6 7 2 5
361
Table 2 (continued ) Isolate
Number of IS6110
Number identified by hINVPCR
Number identified at 5Vend
Number identified at 3Vend
51 52 Total
6 2 506
7a 0 312
4 0 287
6 0 215
a
See text.
189 independent genomic insertion sites that were identified, only 38% (71) were detected with two or more of the enzymes. Analysis of 608 sequence files available for analysis indicated that, the majority of the data were detected with only one enzyme: Bsp1286I (12%, 23 independent sites), HaeII (29%, 55 independent sites) or PvuI (21%, 40 independent sites). As predicted then, the screening of both ends of the IS and using a battery of different restriction endonucleases resulted in a more effective sampling of the genome.
4. Discussion The discovery of ever more transposable elements in newly sequenced species will continue to maintain interest in the roles that they play in evolution. To date, it has really only been practicable to screen for transposable element induced genomic polymorphisms in highly constrained situations: in isogenic laboratory strains (mutagenesis experiments), or in natural populations after sampling very few strains, or by using low resolution methods (RFLP). We have shown here that they can be readily mapped in quantity using a simple but highly effective strategy. High resolution methods for the characterisation of the insertion sites of transposable elements in the genomes of bacteria are often complex. The screening of clone libraries is comprehensive, but the many complex steps limit the number of isolates that are likely to be tested (Mendiola et al., 1992). DNA microarrays potentially offer comprehensivity of detection of insertion sites, but they require foreknowledge of the genome sequence, the availability of an array for that species, the resolution with which insertions in the genome can be located is only as good as the density of the probes along the genome, and most critically, novel sequences flanking insertions will not be identified (Kivi et al.,
362
H. Yesilkaya et al. / Journal of Microbiological Methods 53 (2003) 355–363
2002). On the other hand, hINVPCR with DNA sequencing allows the efficient screening of large numbers of insertion sites from large numbers of isolates and requires minimal amounts of genomic DNA. In silico analysis has been shown to be a highly predictive, rapid and powerful methodology with which to determine those restriction enzymes likely to be most suitable for hINVPCR and especially those producing a maximum proportion of fragments in the correct size range. In particular, the use of a battery of enzymes allowed the identification of many more insertion sites than any one of them alone would have done. Yield was further increased by screening both ends of the element, and this could be most efficiently done by sampling the ligation mixtures with different primer pairs. The number of insertion sites identified here is noteworthy from two perspectives. Firstly, although a large number of sites were identified, it is probable that there are, perhaps many, others present. This is best seen in the large population of sites that were only identified by one of the three enzymes; the use of even more enzymes could increase the yield further, but the greater effect would have to be balanced by the output. Secondly, and conversely, the strains studied here were deliberately selected to be as diverse from each other as possible; yet in spite of this, many IS sites were found on several occasions from different strain. This suggests that there are conserved insertions in the population, reflective presumably of insertion events earlier in the evolution of this species. Knowledge of the genomic locations of transposable elements in strains selected to represent the diversity of a species’ population opens up many avenues for further study. From the perspective of transposable element, it offers greater understanding of the factors that influence their transposition and we are now analysing our collection from this angle, looking particularly to see whether this element, IS6110, transposes in nature as would be predicted from its affinity to the IS3 family of elements (Chalmers and Blot, 1999). From the stand point of the host, it opens up new opportunities to understand their adaptation and evolution. For example, knowing the locations of many different insertion sites can shed light on genomic mutations and polymorphisms in wild populations and we are analysing our M. tuberculosis
dataset from this perspective. Transposable element induced polymorphisms are extensively used to type strains of a species, and while RFLP methods are simple, the low information yield does not allow much interpretation of the results beyond simple epidemiology. The ability to recognise insertions at specific genomic locations will allow strain relationships to be determined more speedily (using PCR) and to greater effect in allowing evolutionary relationships to be examined. Insertion site typing (Steinlein and Crawford, 2001) uses a combination of multiplex PCR and probing to identify the presence of particular insertions in strains and knowledge of the insertion sites found across a population will greatly facilitate such methods.
References Bik, E.M., Gouw, R.D., Mooi, F.R., 1996. DNA fingerprinting of Vibrio cholerae strains with a novel insertion sequence element: a tool to identify epidemic strains. J. Clin. Microbiol. 34, 1453 – 1461. Bjourson, A.J., Cooper, J.E., 1992. Band stab PCR: a simple technique for the purification of individual PCR products. Nucleic Acids Res. 20, 4675. Chalmers, R., Blot, M., 1999. Insertion sequences and transposons. In: Charlebois, R.L. (Ed.), Organisation of the Prokaryotic Genome. ASM Press, Washington, DC, pp. 151 – 170. Cox, J.S., Chen, B., McNeil, M., Jacobs, W.R., 1999. Complex lipid determines tissue specific replication of M. tuberculosis in mice. Nature (Lond.) 402, 79 – 83. de Boer, A.S., Borgdorff, M.W., de Haas, P.E.W., Nagelkerke, N.J.D., van Embden, J.D.A., van Soolingen, D., 2000. Analysis of rate of change of IS6110 RFLP patterns of Mycobacterium tuberculosis based on serial patient isolates. J. Infect. Dis. 180, 1238 – 1244. Fang, Z., Forbes, K.J., 1997. A Mycobacterium tuberculosis IS6110 preferential locus (ipl) for insertion into the genome. J. Clin. Microbiol. 35, 479 – 481. Fang, Z., Morrison, N., Watt, B., Doig, C., Forbes, K.J., 1998. IS6110 transposition and evolutionary scenario of the direct repeat locus in a group of closely related Mycobacterium tuberculosis strains. J. Bacteriol. 180, 2102 – 2109. Finlay, B.B., Falkow, S., 1997. Common themes in microbial pathogenicity revisited. Microbiol. Mol. Biol. Rev. 61, 136 – 169. Gillespie, S.H., Dickens, A., McHugh, T.D., 2000. False molecular clusters due to nonrandom association of IS6110 with Mycobacterium tuberculosis. J. Clin. Microbiol. 38, 2081 – 2086. Kivi, M., Liu, X., Raychaudhuri, S., Altman, R.B., Small, P., 2002. Determining the genomic locations of repetitive DNA sequences with a whole-genome microarray: IS6110 in Mycobacterium tuberculosis. J. Clin. Microbiol. 40, 2192 – 2198.
H. Yesilkaya et al. / Journal of Microbiological Methods 53 (2003) 355–363 Mahillon, J., Leonard, C., Chandler, M., 1999. IS elements as constituents of bacterial genomes. Res. Microbiol. 150, 675 – 687. McAdam, R.A., Quan, S., Guilhot, C., 2000. Mycobacterial transposans and their applications. In: Hatfull, G.H., Jacobs, W.R. (Eds.), Molecular Genetics of Mycobacteria. ASM Press, Washington, DC, pp. 69 – 84. Mendiola, M.V., Martin, C., Otal, I., Gicquel, B., 1992. Analysis of the regions responsible for IS6110 RFLP in a single Mycobacterium tuberculosis strain. Res. Microbiol. 143, 767 – 772. Ochman, H., Gerber, A.S., Hartl, D.L., 1998. Genetic applications of an inverse polymerase chain reaction. Genetics 120, 621 – 623. Patel, S., Wall, S., Saunders, N.A., 1996. Heminested inverse PCR for IS6110 fingerprinting of Mycobacterium tuberculosis strains. J. Clin. Microbiol. 34, 1686 – 1690. Rutherford, K., Parkhill, J., Crook, J., Horsnell, T., Rice, P., Rajan-
363
dream, M.A., Barrell, B., 2000. Artemis: sequence visualisation and annotation. Bioinformatics 16, 944 – 945. Steinlein, L.M., Crawford, J.T., 2001. Reverse dot blot assay (insertion site typing) for precise detection of sites of IS6110 insertion in the Mycobacterium tuberculosis genome. J. Clin. Microbiol. 39, 871 – 878. van Embden, J.D.A., Cave, M.D., Crawford, J.T., Dale, J.W., Eisenach, K.D., Gicquel, B., Hermans, P., Martin, C., McAdam, R., Shinnick, T.M., Small, P., 1993. Strain identification of Mycobacterium tuberculosis by DNA fingerprinting: recommendations for a standardised methodology. J. Clin. Microbiol. 31, 406 – 409. Wall, S., Ghanekar, K., McFadden, J., Dale, J.W., 1999. Contextsensitive transposition of IS6110 in mycobacteria. Microbiology 145, 3169 – 3176.
Locating transposable element polymorphisms in bacterial genomes Hasan Yesilkaya a,*, Anne Thomson a, Christine Doig b, Brian Watt b, Jeremy W. Dale c, Ken J. Forbes a b
a Department of Medical Microbiology, University of Aberdeen, Medical School Building, Foresterhill, Aberdeen AB25 2ZD, UK Scottish Mycobacteria Reference Laboratory, New Royal Infirmary of Edinburgh, Lothian Universities Hospitals NHS Trust, Little France, Old, Dalkeith Road, Edinburgh EH16 4SU, UK c Molecular Microbiology Group, School of Biomedical and Life Sciences, University of Surrey, Guildford, Surrey GU2 7XH, UK
Received 30 August 2002; received in revised form 6 December 2002; accepted 6 December 2002
Abstract Although whole-genome sequencing is greatly extending our knowledge of the genetic capacity of those bacterial species, it is only directly informative for the particular strain sequenced. Many bacterial species exhibit more or less genetic polymorphism within their populations and characterising this variety is an extremely important way of elucidating the biology of these species. Often genomic polymorphisms are associated with multicopy elements, particularly transposable elements. We describe a novel method that efficiently characterises the sequences of such polymorphisms. We have optimised heminested inverse PCR (hINVPCR) to assess the diversity of insertional polymorphisms of a transposable element (IS6110) in clinical isolates of Mycobacterium tuberculosis. To increase the yield of information, genomic DNA was digested with different endonucleases (Bsp1286I, HaeII or PvuI), and primers based on both the 5Vand 3Vends of IS6110 were used to amplify and determine the genomic sequence upstream (or downstream) of the transposable element. We found that both the choice of restriction enzyme and the use of primers at both ends of the transposable element significantly increased the diversity of the insertion sites identified. Band stabbing was incorporated into the method as an alternative to cloning in order to screen large number of isolates at a sequence level in a rapid and labour-efficient fashion. We describe some of the purposes to which such data can be put. D 2002 Elsevier Science B.V. All rights reserved. Keywords: Bacteria; Genome polymorphism; In silico analysis; IS6110
1. Introduction The availability of whole-genome sequences of different bacterial species is ever increasing (http:// * Corresponding author. Tel: +44-1224-552864; fax: +44-1224685604. E-mail address: [email protected] (H. Yesilkaya).
www.ncbi.nlm.nih.gov/PMGifs/genomes/micr.html) and is greatly extending our understanding of microbial biology. Such resources are, however, restricted to single, hopefully representative, strains of the species, or at best a few such strains. The genetic diversity across the whole population of a species will not be revealed in such studies. Given the central role played by multicopy and transposable elements in
0167-7012/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved. doi:10.1016/S0167-7012(02)00256-7
356
H. Yesilkaya et al. / Journal of Microbiological Methods 53 (2003) 355–363
genomic rearrangements and in horizontal gene transfer (Mahillon et al., 1999), the determination of the sequences flanking these, be they in prokaryotic or eukaryotic species, provides a powerful tool for mining such biodiversity. Insertion sequences (IS) are 700 – 2500 bp long DNA elements that can translocate in the genome; they can cause spontaneous mutations, the physical disruption of genes, and are able to alter gene expression. In addition, IS can promote chromosomal rearrangements through inversions, translocations, duplications and deletions (Chalmers and Blot, 1999). IS in bacteria have been utilised in studies of microbial evolution (Fang et al., 1998), epidemiology (Bik et al., 1996), microbial virulence (Finlay and Falkow, 1997), and employed as tools in molecular biology (Cox et al., 1999). All such studies require an understanding of the dynamics of the transposition event (Wall et al., 1999) and how the target sites are selected; however, little is known about such activity in natural populations. IS6110 is found in Mycobacterium tuberculosis clinical isolates at between 0 and 25 copies per strain and it is thought to have little or no specificity at the nucleotide level for insertion in the M. tuberculosis genome (McAdam et al., 2000), although there are preferred regions (Fang and Forbes, 1997; Gillespie et al., 2000). IS6110 restriction fragment length polymorphism (RFLP) is the principal method for strain differentiation in M. tuberculosis. Using this as our model, we have developed a method based on heminested inverse PCR (hINVPCR) that allows the characterisation of IS-induced genomic polymorphism in large bacterial collections, without the need for cloning.
2. Materials and methods 2.1. Bacterial strains The M. tuberculosis clinical isolates tested in this study were received by the Scottish Mycobacteria Reference Laboratory (Edinburgh, Scotland, UK) between 1991 and 1997 (n = 52). DNA was extracted by growing the isolates in Middlebrook 7H9 broth in 50 ml centrifuge tubes at 37 jC for about 4 weeks according to a standard protocol (van Embden et al., 1993). 2.2. IS6110 restriction fragment length polymorphism (RFLP) analysis The isolates had a mean IS6110 copy number of 9.7 (range 2– 18) as assessed by RFLP (van Embden et al., 1993). Relatedness of IS6110 fingerprint patterns was calculated using the Dice coefficient of similarity (BioNumerics, version 2.0; Applied Maths, Kortrijk, Belgium) and isolates with patterns of < 70% relatedness were selected for hINVPCR in order to maximise the recovery of novel insertion sites. 2.3. Primers The IS6110 primer sequences were designed using OLIGO (version 5.0, National Bioscience, Plymouth, MN), and synthesised by Interactiva (Ulm, Germany). The relative positions of the primers are given in Fig. 1, and the sequence of the primers from 5V- to 3V-end as follows: PB: GTCTACTTGGTGTTGGCTGC (nt 170 – 189); PA1: GGGTCATGTCAGGTGGTTCA (nt 47 – 66 in the complementary strand); NP: GTCA-
Fig. 1. Location of primers and restriction endonuclease recognition sites in IS6110 (open line). Nucleotide positions of the restriction sites in IS6110 are indicated.
H. Yesilkaya et al. / Journal of Microbiological Methods 53 (2003) 355–363
357
Fig. 2. Flow diagram of hINVPCR. Procedure (for PvuI) showing the relative positions of IS6110 (open box) and the flanking DNA (thin line).
358
H. Yesilkaya et al. / Journal of Microbiological Methods 53 (2003) 355–363
GCACGATTCGGAGT (128 – 155 in the complementary strand); INS2: TTTGTCACCGACGCCTACGC (nt 855 –874 in the complementary strand); INV1: CTGGGTCGACTGGTTCAAC (nt 1202 – 1220); INV2: TACGCTCAACGCCAGAGAC (nt 1281 – 1299); INV3: CCTATGACAATGCACTAGCCG (nt 1093– 1113 in the complementary strand); P8: CTGAACCGTGAGGGCATCGAG (nt 624 – 644 in the complementary strand); P9: GGTTTGCGGTGGGGTGTCGAG (nt 426– 443). 2.4. Template preparation for PCR The strategy used was adapted from Patel et al. (1996) with some modifications (Fig. 2). For each sample, 100 – 200 ng genomic DNA in a 20-Al volume was digested with 2.5 U of Bsp1286I, HaeII or PvuI (New England Biolabs, Hitchin, UK) then heat-inactivated. Several criteria were considered in the choice of these enzymes. Firstly, the use of more than one restriction enzyme, each of which has a different recognition sequence, gives a better sampling across the genome than a single restriction enzyme alone could do, simply by virtue of the different, randomly distributed recognition sites in the genome. Secondly, the enzymes should cleave the DNA with a staggered cut to produce the cohesive-ended DNA fragments that ligate more efficiently than blunt-ended ones (Ochman et al., 1998). Thirdly, the selected enzymes should not cut the IS6110 sequence too close to the ends of the IS
leaving sufficient IS6110 sequence in which to place the primers required for the hINVPCR, but not so distant as to excessively increase the length of the self-ligation products (see below). This protocol allowed us to use the digested – ligated mixtures of genomic DNA as the template for PCR amplifications from both ends of the IS, thus reducing the cost and labour of the experiment, and increasing the probability of detecting more insertion sites in the genome. Our fourth consideration in the selection of the restriction endonucleases was the distribution of the restriction enzyme sites in the M. tuberculosis genome. In silico digests of the H37Rv genome indicated an even distribution of the selected recognition sites along the genome (Fig. 3). However, the lengths of the fragment must also be considered since there is a negative correlation between efficiency of self-ligation and the length of the DNA fragment, particularly for fragments above 3 kb (Ochman et al., 1998). The restriction enzymes must digest the genome into as many fragments as possible that are in the appropriate size range for hINVPCR. The restricted fragments were self-ligated at a concentration of 0.5 Ag ml 1 DNA in 20 Al of ligation buffer (stock: 50 mM Tris – HCl (pH 7.6), 10 mM dithioerythritol, bovine serum albumin 500 Ag ml 1) in the presence of 20 U ml 1 T4 ligase (Roche Diagnostics, Mannheim, Germany) by incubating at 16 jC for 16 h in a thermocycler with a hot lid. Heminested inverse PCR (hINVPCR) amplification used the protocol described previously (Patel et al.,
Fig. 3. Distribution of Bsp1286I, HaeII, and PvuI fragment lengths along the H37Rv genome by fragment size class (x Bsp1286I, 5 HaeII, E PvuI).
H. Yesilkaya et al. / Journal of Microbiological Methods 53 (2003) 355–363
1996). Second-round PCR product (8 Al) was electrophoresed (at least for 6 h at 1.3 V cm 1) in 2% Metaphor agarose (FMC BioProducts, Rockland, ME) and the resolved fragments visualised after ethidium bromide staining under UV light. Multiple bands were observed and the banding patterns were different for each isolate/restriction enzyme/primer combination (Fig. 4). Each band was stabbed with a hypodermic needle (Bjourson and Cooper, 1992) and the DNA adhering to the needle washed into fresh (secondround) PCR reagent and re-amplified using the same PCR conditions. Those re-amplified fragments that had a minimum of background amplification were purified to remove contaminating primers and dNTPs (Centricon Centrifugal Devices, Millipore, MA) and sequenced as described previously (Fang et al., 1998). The primers used for sequencing upstream (5V) of IS6110 were PB (for Bsp1286I, HaeII) or P9 (for PvuI), and for sequencing downstream (3V) of IS6110
359
were P8 (for Bsp1286I), INS2 (for HaeII) or INV3 (for PvuI). 2.5. Sequence analysis The sequences were mapped to the whole genome of the sequenced strains of M. tuberculosis H37Rv (http://www.sanger.ac.uk) and CDC1551 (http://www.tigr.org ) using BLA ST (http:// www.ncbi.nlm.nih.gov) and ARTEMIS (Rutherford et al., 2000).
3. Results Experimentally we observed that the hINVPCR amplicon sizes ranged from 100 to 1000 bp (Fig. 4), therefore the proportion of the genome that is accessible to hINVPCR can be estimated from in silico
Fig. 4. Representative gel of hINVPCR amplicons for two strains, 34 and 44, obtained with three different enzymes (Bsp1286I (B), HaeII (H) and PvuI (P)), and primer sets, 5Vand 3Vends, respectively. Size standards (lanes kb) were 1 kb DNA size markers (Gibco/BRL), and fragment sizes are indicated in kilobases.
360
H. Yesilkaya et al. / Journal of Microbiological Methods 53 (2003) 355–363
Table 1 Overall study results Restriction endonuclease used to digest genomic DNA Bsp1286I
HaeII
PvuI
In silico digest of H37Rv Percent genome in 3507726 bp 3339986 bp 1095798 bp fragments between (79.5%) (75.7%) (24.8%) 100 – 1000 nt long Number of 100 – 1000 < 10 043a 9353 2289 nucleotide long fragments hINVPCR bands Total number From 5Vend From 3Vend Insertion sites identified at nucleotide level a
269 138 131 105
548 258 294 268
429 254 171 235
See text.
digests of the H37Rv DNA sequence as the proportion of such sized fragments of the total produced by that enzyme (Fig. 3). For Bsp1286I, 79.5% (3507726 bp) of the genome sequence fell within the amplifiable size range (Table 1). Similarly, HaeII would allow 75.7% (3339986 bp) and PvuI 24.8% (1095798 bp) of the H37Rv genome to be amplified by hINVPCR. In the case of Bsp1286I, a lower yield of hINVPCR bands than this would be expected since this enzyme has multiple recognition sequences and therefore several alternative sticky-end sequences; there would only be ligation for those fragments which had the IS6110 recognition sequence (GAGCTVC at the 5Vend and GTGCCVC at the 3Vend) at both ends of the fragment, and so only PCR amplification using IS6110 primers could occur from a subset of the IS6110-containing Bsp1286I fragments. In silico analysis predicted that HaeII would produce the most hINVPCR fragments, PvuI somewhat less and Bsp1286I much less. The 52 clinical isolates screened here produced 1246 amplified hINVPCR bands (Table 1): HaeII (548 bands, 44%), PvuI (429 bands, 34%) and Bsp1286I (269 bands; 22%). Predicted and observed yields were therefore in agreement with each other. Similar numbers of hINVPCR fragments were generated from the 5Vand 3Vends of IS6110 for all of the enzymes. A somewhat preferential amplification
from the 5Vend with PvuI digests generated rather more (60%) bands (Table 1), but this was probably artifactual since subsequent tests on more isolates (82) reduced this excess to 56%. All of the 1246 amplified hINVPCR fragments were band stab re-amplified and these purified products visualised by gel electrophoresis. Thirtytwo percent (402) of these re-amplified products had background smears, possibly attributable to the presence of single stranded or mismatched amplicons (Patel et al., 1996), or comigrating bands (data not shown). Although a very simple and effective method, we found the success of band stabbing to be very dependent on the good electrophoretic separation of the amplicons and this was optimised here by the use of high resolution agarose (2% Metaphor agarose gel) and by extended electrophoresis times. Of the remaining 844 (68%) products that were sequenced, 608 (49% of the total hINVPCR bands) yielded IS6110 flanking sequences which could be mapped to the H37Rv or CDC1551 genomes. Eighty-five percent (516) of these sequences were of sufficient quality (generally a sequence run of >50 bp with sequence identity >80%) to locate the IS6110 insertion to a unique nucleotide position in the genome. The remaining 15% (92) of sequences, although of poorer quality, were adequate to locate the insertion site to within tens of base pairs. The sequence efficiency, ratio between the number of sequenced fragments and the number of sequence data available for analysis, was higher for the bands amplified from the 5Vend of IS6110 (93%) compared to the 3Vend of IS6110 (56%), and it is likely that this was methodological in origin, perhaps due to greater misamplification of fragments from the 3Vend of the IS. In total, 312 insertion sites representing 189 unique sites were identified (Table 2), with all of the IS6110 insertion sites being identified in nearly half of the isolates (n = 24) (characterised from either 5V- or 3Vend of the element). In one quarter of the isolates (n = 14), the number of unique insertion sites identified exceeded the copy number of IS6110 identified by RFLP (Table 2). This was most probably due to the low resolution of the RFLP method in separating comigrating bands, less plausibly to rare transposition events during culture (de Boer et al., 2000). Of the
H. Yesilkaya et al. / Journal of Microbiological Methods 53 (2003) 355–363 Table 2 Strain-specific efficiency of hINVPCR for identification of IS6110 insertion sites Isolate
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50
Number of IS6110 5 10 8 12 9 6 8 12 8 6 9 8 9 7 5 11 8 9 7 7 8 10 12 11 9 7 6 13 11 12 12 9 10 11 17 14 7 9 15 8 2 18 7 14 16 9 18 12 15 12
Number identified by hINVPCR 7a 10 9a 8 9 5 8 11 8 8a 9 8 10a 5 3 6 9a 8 7 7 8 7 5 6 0 8a 8a 12 10 7 8 7 6 11 13 11 13a 12a 12 9a 3a 13 10a 13 12 12a 11 10 9 9
Number identified at 5Vend 3 6 6 5 6 4 4 6 6 6 6 6 7 4 2 3 6 7 6 5 5 3 4 4 0 5 5 11 7 4 5 4 5 7 7 6 9 6 8 9 2 8 8 8 9 6 7 6 7 4
Number identified at 3Vend 2 7 4 6 4 3 5 5 3 2 4 5 6 2 2 4 5 1 4 3 5 4 1 1 0 4 4 3 6 4 4 4 2 7 6 6 5 6 7 1 2 7 3 7 6 7 6 7 2 5
361
Table 2 (continued ) Isolate
Number of IS6110
Number identified by hINVPCR
Number identified at 5Vend
Number identified at 3Vend
51 52 Total
6 2 506
7a 0 312
4 0 287
6 0 215
a
See text.
189 independent genomic insertion sites that were identified, only 38% (71) were detected with two or more of the enzymes. Analysis of 608 sequence files available for analysis indicated that, the majority of the data were detected with only one enzyme: Bsp1286I (12%, 23 independent sites), HaeII (29%, 55 independent sites) or PvuI (21%, 40 independent sites). As predicted then, the screening of both ends of the IS and using a battery of different restriction endonucleases resulted in a more effective sampling of the genome.
4. Discussion The discovery of ever more transposable elements in newly sequenced species will continue to maintain interest in the roles that they play in evolution. To date, it has really only been practicable to screen for transposable element induced genomic polymorphisms in highly constrained situations: in isogenic laboratory strains (mutagenesis experiments), or in natural populations after sampling very few strains, or by using low resolution methods (RFLP). We have shown here that they can be readily mapped in quantity using a simple but highly effective strategy. High resolution methods for the characterisation of the insertion sites of transposable elements in the genomes of bacteria are often complex. The screening of clone libraries is comprehensive, but the many complex steps limit the number of isolates that are likely to be tested (Mendiola et al., 1992). DNA microarrays potentially offer comprehensivity of detection of insertion sites, but they require foreknowledge of the genome sequence, the availability of an array for that species, the resolution with which insertions in the genome can be located is only as good as the density of the probes along the genome, and most critically, novel sequences flanking insertions will not be identified (Kivi et al.,
362
H. Yesilkaya et al. / Journal of Microbiological Methods 53 (2003) 355–363
2002). On the other hand, hINVPCR with DNA sequencing allows the efficient screening of large numbers of insertion sites from large numbers of isolates and requires minimal amounts of genomic DNA. In silico analysis has been shown to be a highly predictive, rapid and powerful methodology with which to determine those restriction enzymes likely to be most suitable for hINVPCR and especially those producing a maximum proportion of fragments in the correct size range. In particular, the use of a battery of enzymes allowed the identification of many more insertion sites than any one of them alone would have done. Yield was further increased by screening both ends of the element, and this could be most efficiently done by sampling the ligation mixtures with different primer pairs. The number of insertion sites identified here is noteworthy from two perspectives. Firstly, although a large number of sites were identified, it is probable that there are, perhaps many, others present. This is best seen in the large population of sites that were only identified by one of the three enzymes; the use of even more enzymes could increase the yield further, but the greater effect would have to be balanced by the output. Secondly, and conversely, the strains studied here were deliberately selected to be as diverse from each other as possible; yet in spite of this, many IS sites were found on several occasions from different strain. This suggests that there are conserved insertions in the population, reflective presumably of insertion events earlier in the evolution of this species. Knowledge of the genomic locations of transposable elements in strains selected to represent the diversity of a species’ population opens up many avenues for further study. From the perspective of transposable element, it offers greater understanding of the factors that influence their transposition and we are now analysing our collection from this angle, looking particularly to see whether this element, IS6110, transposes in nature as would be predicted from its affinity to the IS3 family of elements (Chalmers and Blot, 1999). From the stand point of the host, it opens up new opportunities to understand their adaptation and evolution. For example, knowing the locations of many different insertion sites can shed light on genomic mutations and polymorphisms in wild populations and we are analysing our M. tuberculosis
dataset from this perspective. Transposable element induced polymorphisms are extensively used to type strains of a species, and while RFLP methods are simple, the low information yield does not allow much interpretation of the results beyond simple epidemiology. The ability to recognise insertions at specific genomic locations will allow strain relationships to be determined more speedily (using PCR) and to greater effect in allowing evolutionary relationships to be examined. Insertion site typing (Steinlein and Crawford, 2001) uses a combination of multiplex PCR and probing to identify the presence of particular insertions in strains and knowledge of the insertion sites found across a population will greatly facilitate such methods.
References Bik, E.M., Gouw, R.D., Mooi, F.R., 1996. DNA fingerprinting of Vibrio cholerae strains with a novel insertion sequence element: a tool to identify epidemic strains. J. Clin. Microbiol. 34, 1453 – 1461. Bjourson, A.J., Cooper, J.E., 1992. Band stab PCR: a simple technique for the purification of individual PCR products. Nucleic Acids Res. 20, 4675. Chalmers, R., Blot, M., 1999. Insertion sequences and transposons. In: Charlebois, R.L. (Ed.), Organisation of the Prokaryotic Genome. ASM Press, Washington, DC, pp. 151 – 170. Cox, J.S., Chen, B., McNeil, M., Jacobs, W.R., 1999. Complex lipid determines tissue specific replication of M. tuberculosis in mice. Nature (Lond.) 402, 79 – 83. de Boer, A.S., Borgdorff, M.W., de Haas, P.E.W., Nagelkerke, N.J.D., van Embden, J.D.A., van Soolingen, D., 2000. Analysis of rate of change of IS6110 RFLP patterns of Mycobacterium tuberculosis based on serial patient isolates. J. Infect. Dis. 180, 1238 – 1244. Fang, Z., Forbes, K.J., 1997. A Mycobacterium tuberculosis IS6110 preferential locus (ipl) for insertion into the genome. J. Clin. Microbiol. 35, 479 – 481. Fang, Z., Morrison, N., Watt, B., Doig, C., Forbes, K.J., 1998. IS6110 transposition and evolutionary scenario of the direct repeat locus in a group of closely related Mycobacterium tuberculosis strains. J. Bacteriol. 180, 2102 – 2109. Finlay, B.B., Falkow, S., 1997. Common themes in microbial pathogenicity revisited. Microbiol. Mol. Biol. Rev. 61, 136 – 169. Gillespie, S.H., Dickens, A., McHugh, T.D., 2000. False molecular clusters due to nonrandom association of IS6110 with Mycobacterium tuberculosis. J. Clin. Microbiol. 38, 2081 – 2086. Kivi, M., Liu, X., Raychaudhuri, S., Altman, R.B., Small, P., 2002. Determining the genomic locations of repetitive DNA sequences with a whole-genome microarray: IS6110 in Mycobacterium tuberculosis. J. Clin. Microbiol. 40, 2192 – 2198.
H. Yesilkaya et al. / Journal of Microbiological Methods 53 (2003) 355–363 Mahillon, J., Leonard, C., Chandler, M., 1999. IS elements as constituents of bacterial genomes. Res. Microbiol. 150, 675 – 687. McAdam, R.A., Quan, S., Guilhot, C., 2000. Mycobacterial transposans and their applications. In: Hatfull, G.H., Jacobs, W.R. (Eds.), Molecular Genetics of Mycobacteria. ASM Press, Washington, DC, pp. 69 – 84. Mendiola, M.V., Martin, C., Otal, I., Gicquel, B., 1992. Analysis of the regions responsible for IS6110 RFLP in a single Mycobacterium tuberculosis strain. Res. Microbiol. 143, 767 – 772. Ochman, H., Gerber, A.S., Hartl, D.L., 1998. Genetic applications of an inverse polymerase chain reaction. Genetics 120, 621 – 623. Patel, S., Wall, S., Saunders, N.A., 1996. Heminested inverse PCR for IS6110 fingerprinting of Mycobacterium tuberculosis strains. J. Clin. Microbiol. 34, 1686 – 1690. Rutherford, K., Parkhill, J., Crook, J., Horsnell, T., Rice, P., Rajan-
363
dream, M.A., Barrell, B., 2000. Artemis: sequence visualisation and annotation. Bioinformatics 16, 944 – 945. Steinlein, L.M., Crawford, J.T., 2001. Reverse dot blot assay (insertion site typing) for precise detection of sites of IS6110 insertion in the Mycobacterium tuberculosis genome. J. Clin. Microbiol. 39, 871 – 878. van Embden, J.D.A., Cave, M.D., Crawford, J.T., Dale, J.W., Eisenach, K.D., Gicquel, B., Hermans, P., Martin, C., McAdam, R., Shinnick, T.M., Small, P., 1993. Strain identification of Mycobacterium tuberculosis by DNA fingerprinting: recommendations for a standardised methodology. J. Clin. Microbiol. 31, 406 – 409. Wall, S., Ghanekar, K., McFadden, J., Dale, J.W., 1999. Contextsensitive transposition of IS6110 in mycobacteria. Microbiology 145, 3169 – 3176.