FEMS Microbiology Letters 243 (2005) 303–310 www.fems-microbiology.org
Discrimination of Francisella tularensis subspecies using surface enhanced laser desorption ionization mass spectrometry and multivariate data analysis Margaretha Lundquist a, Mikael B. Caspersen b, Per Wikstro¨m a, Mats Forsman a
a,*
Swedish Defence Research Agency, Department of NBC-analysis, 901 82, Umea˚, Sweden b Ciphergen Biosystems, Copenhagen, Denmark
Received 13 October 2004; received in revised form 13 December 2004; accepted 16 December 2004 First published online 24 December 2004 Edited by M. Schembri
Abstract Francisella tularensis causes the zoonotic disease tularemia, and is considered a potential bioterrorist agent due to its extremely low infection dose and potential for airborne transmission. Presently, F. tularensis is divided into four subspecies; tularensis, holarctica, mediasiatica and novicida. Phenotypic discrimination of the closely related subspecies with traditional methods is difficult and tedious. Furthermore, the results may be vague and they often need to be complemented with virulence tests in animals. Here, we have used surface enhanced laser desorption ionization time-of-flight mass spectrometry (SELDI-TOF-MS) to discriminate between the four subspecies of F. tularensis. The method is based on the differential binding of protein subsets to chemically modified surfaces. Bacterial thermolysates were added to anionic, cationic, and copper ion-loaded immobilized metal affinity SELDI chip surfaces. After binding, washing, and SELDI-TOF-MS different protein profiles were obtained. The spectra generated from the different surfaces were then used to characterize each bacterial strain. The results showed that the method was reproducible, with an average intensity variation of 21%, and that the mass precision was good (300–450 ppm). Moreover, in subsequent cluster analysis and principal component analysis (PCA) data for the analyzed Francisella strains grouped according to the recognized subspecies. Partial least squares-discriminant analysis (PLS-DA) of the protein profiles also identified proteins that differed between the strains. Thus, the protein profiling approach based on SELDI-TOF-MS holds great promise for rapid high-resolution phenotypic identification of bacteria. 2004 Federation of European Microbiological Societies. Published by Elsevier B.V. All rights reserved. Keywords: Francisella tularensis; Mass spectrometry; Identification; Subspecies; Multivariate data analysis
1. Introduction Francisella tularensis (F. t.), the causative agent of tularemia, is a small, aerobic, gram-negative bacterium * Corresponding author. Address. Swedish Defence Research Agency, Department of NBC-analysis, 901 82 Umea˚, Sweden. Tel.: +46 90 106669; fax: +46 90 106800. E-mail address:
[email protected] (M. Forsman).
that is a highly virulent pathogen for humans and a wide range of mammals including rodents, hares and rabbits. The disease is zoonotic, transmitted to humans through ticks, mosquitoes, or handling infected animals [1]. Inhalation of as few as ten organisms may be sufficient to cause disease [2,3]. Because of its infectious nature, ease of dissemination and the high mortality rates associated with it, F. t. has recently received increasing attention as a Category A potential agent of bioterrorism
0378-1097/$22.00 2004 Federation of European Microbiological Societies. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.femsle.2004.12.020
304
M. Lundquist et al. / FEMS Microbiology Letters 243 (2005) 303–310
[4,5]. Presently, F. t. is divided into four subspecies [6]. F. t. subspecies tularensis (also referred to as type A) is most commonly isolated in North America and is the most virulent subspecies and F. t. subspecies holartica (type B), which has a lower virulence for humans, is predominantly found in Europe. A third subspecies, mediasiatica, has only been isolated from the Central Asian republics and shows moderate virulence. The fourth subspecies, novicida, is rare and only causes disease in immunocompromised individuals [7]. Notably, the various subspecies of F. t. cannot be distinguished serologically, apart from the least virulent F. t. subspecies, novicida [8]. Instead, laborious and time-consuming biochemical tests have been required, to date, due to the low-fermentation and fastidious nature of F. t. [6]. The substantial genetic similarity among Francisella subspecies also makes genotypic discrimination of the subspecies difficult. However, in recent years, several high-resolution genotypic methods have been developed for identifying the different subspecies [9–11]. The identification of bacteria by mass spectrometry (MS) is attractive in this context for several reasons, including its high speed, resolution and mass precision. Several investigators have had success using matrix-assisted laser desorption ionization (MALDI)-time-offlight (TOF) MS on a range of bacteria using a variety of sample preparation methods for their identification [12]. However, issues concerning the reproducibility of the acquired spectra have impeded wider application of the method. In addition, MALDI spectra of bacterial lysates generally have low numbers of peaks, in the order of 20–40, in the 2–20 kDa mass range [13]. More elaborate and time-consuming sample preparation methods employing liquid chromatography, concentration and desalting prior to MALDI-TOF have been shown to increase the number of peaks in the 2–20 kDa mass range to 150–400 [14]. SELDI is distinguished from MALDI in its use of an active sample probe – the proteinchip array – which has an adsorptive surface that allows bacterial lysates to be subjected directly, without prior treatments, to on-chip sample preparation steps, such as selective washing and desalting. This procedure minimizes sample losses, while speeding up and simplifying sample preparation, compared to the standard methods normally employed prior to MALDI. Furthermore, the active capture of the proteins by the proteinchip array ensures non-discriminatory binding of target proteins, which in turn improves the reproducibility and allows both peak mass-to-charge (m/z) ratios and intensity to be used in sample characterization. In this study we show that the SELDI-TOF technique holds great promise for high-resolution identification of closely related bacterial strains. By binding different subsets of proteins on different chemical surfaces in parallel and analyzing the resulting spectral protein profiles it
was possible to show that the data for the analyzed Francisella strains grouped according to the recognized subspecies.
2. Materials and methods 2.1. Bacterial strains and sample preparation The F. t. strains were obtained from the Francisella Strain Collection (FSC) at the Swedish Defence Research Agency. All strains have been characterized by specific agglutination, specific PCR amplification of the F. t. specific lpnA gene [15] and biochemical analyses. The strains are listed in Table 1. The F. t. cultures were initiated directly from frozen seed stocks and grown for 48 h on modified Thayer-Martin agar plates at 37 C and 5% ambient CO2 [16], harvested by scraping, and resuspended in saline (0.85% NaCl) at a concentration of 109 CFU mL 1. Bacterial thermolysates were produced by incubating the resulting suspensions at 65 C for 2 h. Next, 100 lL of the bacterial thermolysates were mixed with 200 lL of 9.5 M urea, 2% Chaps in 10 mM Tris, pH 9.0, by vortexing at 1400 rpm for 1 h at 4 C. The lysates were stored at 70 C until use. 2.2. Proteinchip array preparation Proteinchip arrays (Ciphergen Biosystems, CA, USA) with three different surface chemistries were employed: CM10 (cation exchange), Q10 (anion exchange), and IMAC30 (immobilized metal affinity chromatography). The arrays were mounted in a bioprocessor and all washing and incubation steps were performed using a MicroMix 5 microplate shaker (EURO DPC, UK) with a 20-5 program setting, at room temperature and washing steps in all cases of 5 min duration. All the pipetting steps were performed by a Biomek 2000 (Beckman Coulter, CA, US). The IMAC30 surface was Cu-loaded using 50 lL of 100 mM CuSO4 followed by two 50 lL of 1 mM Hepes, pH 7.0, washes. The binding buffers used for the arrays complemented the binding characteristics of the surfaces, thus 100 mM sodium acetate, pH 4.0, 0.1% Triton was used for CM10, 100 mM Tris/HCl, pH 9.0, 0.1% Triton for Q10, and PBS for IMAC30. Arrays were equilibrated twice for 5 min with 50 lL buffer. Samples were applied after mixing 90 lL of binding buffer with 10 lL of denatured sample and incubated for 60 min. The arrays were washed three times with 150 lL of the surface-matched binding buffer and twice with 150 lL of 1 mM Hepes, pH 7.0. The liquid was removed by inverting the bioprocessor and the arrays were left to dry. Two aliquots (0.6 lL each) of freshly made matrix (50% saturated sinapinic acid in 50% acetonitrile, 0.5% trifluroacetic acid) were then applied, allowing the matrix solution to evaporate between applications.
M. Lundquist et al. / FEMS Microbiology Letters 243 (2005) 303–310
305
Table 1 F. tularensis strains used in this study Strain information F. tularensis subspecies tularensis Human, Alaska Mite, 1988, Slovakia Human lymph node, 1920, Utah Human ulcer, 1941, Ohio F. tularensis subspecies holarctica (non-Japanese) Passage in small mammals of FSC074 Hare, 1974, Na˚s, Sweden Live Vaccine Strain, Russia Human, 1994, Norway Human ulcer, 1998, Sweden Tick, 1949, Moscow area, Russia F. tularensis subspecies holarctica (Japanese) Human, 1950, Japan Tick, 1957, Japan F. tularensis subspecies mediasiatica Midday gerbil, 1965, Central Asia Hare, 1965, Central Asia F. tularensis subspecies novicida Water, 1950, Utah Human blood, 1995, Texas
2.3. SELDI-TOF-MS and data analysis SELDI was performed using a PBS IIC instrument (Ciphergen Biosystems, CA) and standard settings. All SELDI data collection and handling were performed using Ciphergen Proteinchip Software 3.1 (Ciphergen Biosystems CA). The mass scale was calibrated externally using a five-point calibration and third-order curve fitting with b-endorphin (human), 3465.0 Da; insulin (bovine), 5733.58 Da; ubiquitin (bovine), 8564.8 Da; cytochrome C (bovine), 12230.92; and b-lactoglobulin A (bovine), 18363.34 Da. Mass spectra were collected over the 5000–20 000 Da mass range. The background was subtracted using the default software settings. For the data analysis, all peaks with a signal-to-noise ratio higher than five within a peak window of 0.2% of the mass were considered. Only components (proteins) present in more than two of the studied bacterial strains were included. The resulting data were organized in a matrix, as shown in Fig. 1. Cluster analysis was performed using Pearson correlation coefficients and the unweighted pair group method with arithmetic mean (UPGMA) of the software package Bionumerics 3.5 (Applied Maths, BVBA, Sint-Martens-Latem, Belgium). Bootstrap analysis was performed to calculate branching significance of the resulting UPGMA tree. Further, Principal Component Analysis (PCA) was performed to describe the relationship between the 16 different F. t. strains, plus eight replicates of F. t. novicida FSC 040. The spectral dataset could be described as stretches of data variables, and thus represented as points in a multi-dimensional space in a PCA score plot. PCA is a projection method allow-
FSC No.
Alternative strain designation
047 198 230 237
093-1 SE-219/38 ATCC 6223 SCHU S4
069 074 155 158 200 257
SVA T7K SVA T7 LVS, ATCC 29684 CCUG 33391
022 075
Ebina Jama
147 149
543 120
040 159
ATCC15482 fx2
503/840
Strains 022
B
CM10 Component 1-37
C
Q10 Component 1-46
D
IMAC Component 1-65
A
040 047 069 074 075 147 149 155 158 159 198 200 230 237 257 A
Strain number, see Table 1
B
CM10, cation exchange surface
C
Q10, anion exchange surface
D
IMAC, immobilized metal affinity surface
Fig. 1. Schematic diagram of the data matrix. The mass and the intensity of the identified protein peaks are listed in the supplementary data.
ing multi-dimensional point swarms of data to be visualized in a two or three dimensional plot. An intuitive assessment of the relationships between the strains based on their protein profiles can be obtained from visual examination of the resulting score plots. PCA was performed on the unit variance scaled data matrix using SIMCA-P+ 10 software (Umetrics AB, Umea˚, Sweden). Finally, Partial Least Squares Discriminant Analysis
306
M. Lundquist et al. / FEMS Microbiology Letters 243 (2005) 303–310
3. Results
on different batches of media. Triplicates of bacterial thermolysates were prepared on each occasion and subjected to protein profiling analysis. An average CV of 21% and a mass precision of 450 ppm were obtained. The reproducibility employing the more fastidious holarctica subspecies (strain 155) was also investigated and found to be in the same range.
3.1. Protein profiling
3.2. Cluster analysis
Initially, conditions were optimized for the CM10 (cation exchange) and Q10 (anion exchange) surfaces, in order to maximize protein binding, by analyzing spectra obtained using various permutations of pH (pH 4 and 6, for CM10; pH 7.5 and 9 for Q10), detergents (Triton X-100 and Tween20), and several different denaturing buffers (data not shown). The best results were obtained with the buffers described in Section 2, which were therefore chosen for use in the subsequent experiments. The manufacturerÕs recommendations were followed for loading Cu and binding protein to the IMAC30 surface. The effects of varying the amount of protein lysate added to the proteinchip array were investigated by collecting spectra from cation exchange surfaces where the protein concentrations added corresponded to initial numbers of bacteria ranging approximately from 2 · 105 to 2 · 107 CFU (or from 0.6 to 60 lL of lysate). Within this range no major differences in the spectra were observed (data not shown), so an intermediate loading volume of 3.3 lL bacterial lysate was selected for the subsequent experiments. Using the protocols described above for the different chemical surfaces, protein profiles for each of the different strains were generated. The spectra obtained from each of the surfaces contained in the order of 25–50 peaks within the mass range of 5–20 kDa, the range chosen for the data evaluation, and revealed several differences in protein patterns between the strains (supplementary data). To ascertain that the differences found reflected real differences between the strains, and had not occurred by chance, the reproducibility of the method was investigated as follows. First, the within-day variation attributable to the SELDI equipment was assessed by running replicates of the same thermolysate on different proteinchip arrays with the same surface chemistry. This yielded an average coefficient of variance for the signal intensity of 11% and a mass precision of 300 ppm. Next, the day-to-day reproducibility of the SELDI equipment was assessed by repeating the whole experiment a week later using the same thermolysate. This resulted in an average CV of 16% and a mass precision of 450 ppm. Finally, the reproducibility of the sample preparation method was investigated. The FSC 040 strain was grown from the same seed stock at three different points in time, separated by more than a week,
Sixteen F. t. strains, originating from different parts of the world and isolated from different sources (Table 1), were each subjected to SELDI analysis on three different chemical surfaces. The data were analyzed by UPGMA cluster analysis (Fig. 1 and supplementary data). Data generated from any one of the Q10, CM10 or IMAC surfaces alone were insufficient to discriminate consistently between the individual strains (data not shown). However, using the protein profiles from two (Q10 and CM10) or all three surfaces allowed the subspecies to which each strain belonged to be resolved (Fig. 2(a) and (b)). It was also apparent that resolution beyond subspecies was not possible for any case. Furthermore, the tree branch lengths generated from the protein profiles were longer if the analysis was based on Q10 and CM10, rather than all three of the surfaces (Fig. 2(a) and (b)). Thus, using the additional profiles generated from the metal affinity chip surface did not increase the resolution of the subspecies.
(PLS-DA) was performed to identify proteins that differed most between the five groups of strains. Here FSC 022 and 075, F. t. strains from Japan, were considered to be members of a distinct group.
3.3. Principal component analysis (PCA) and discriminatory analysis (PLS-DA) The cumulative protein profiles derived from two chip surfaces (Q10 and CM10) for each strain were also depicted in a PCA score plot (Fig. 3). An advantage of using PCA is that it is better for detecting differences between the protein profiles than cluster analysis, which is more useful for detecting similarities between datasets. The first three principal components were calculated, describing 42%, 17%, and 13%, respectively (72% in total), of the variance. The score plot confirms that the strains grouped according to their respective subspecies (Fig. 3). The first principal component (t1) mainly describes the differences between the novicida group and the rest of the strains, whereas the second principal component (t2) describes differences between the holarctica and tularensis groups. The third principal component (t3) describes the differences in protein profiles between the mediasiatica group and the rest of the strains. In addition, the PCA analysis showed more clearly, in comparison to the cluster analysis, the difference between the two Japanese isolates from the remaining holarctica subspecies. Moreover, the variation between the three triplicates of lysates of strain 040 prepared at three
M. Lundquist et al. / FEMS Microbiology Letters 243 (2005) 303–310
307
Fig. 2. Cluster (UPGMA) analysis based upon intensities of peaks identified in the protein profiles obtained from the different Francisella strains using (a) two different protein chip surfaces (CM10, cation exchange; and Q10, anion exchange) and (b) three different protein chip surfaces (CM10, Q10 and IMAC30, immobilized metal affinity chromatography).The analyses included 83 and 138 protein peaks, respectively. Numbers refer to the different strains as explained in Table 1. Boot strap values using 100 iterations are shown. The scale above indicates percent similarity.
different time points was low since their data-points cluster tightly, indicating that the total variation of the method is low in relation to the differences between the different subspecies. Thus, the inter-sample variation was much lower than the differences between the subspecies groups. PLS-DA identified mass peaks (proteins) that explained the difference in t1, i.e. distinguished novicida from the rest of strains. The three most influential peaks correspond to proteins with estimated molecular weights
of 8.8, 13.6, and 10.9 kDa. This means that these proteins were all expressed to a larger extent in novicida compared to the other strains. Analogously, proteins with masses of 17.2, 16.8, and 7.5 kDa were more abundantly expressed in holarctica compared to tularensis strains, and vice versa for 13.4, 14.8, and 7.8 kDa proteins. Finally, proteins of 8.4, 5.9, and 16.2 kDa were found to be more highly expressed in mediasiatica, in comparison to the other strains.
308
M. Lundquist et al. / FEMS Microbiology Letters 243 (2005) 303–310
Fig. 3. PCA score plot based on intensities of peaks identified in the protein profiles obtained from the different Francisella strains using two different protein chip surfaces (CM10, cation exchange; and Q10, anion exchange). Key: triangles, F. tularensis subspecies holarctica; diamonds, F. tularensis subspecies mediasiatica; squares, F. tularensis subspecies holarctica isolated in Japan; stars, F. tularensis subspecies tularensis; circles, F. tularensis subspecies novicida. The cluster of circles for F. tularensis subspecies novicida indicates the nine spectra obtained from the three triplicate thermolysates of the 040 strain prepared at different time points (as described in Section 3.1) and thus show the inter-sample variance of the method.
4. Discussion The genus Francisella is antigenically coherent and the subspecies are almost indistinguishable by serological methods, although monoclonal antibodies raised against lipopolysaccharides (LPS) seem to be able to discriminate between novicida and other subspecies of F. t. [8]. In this context, the protein profile method presented in this study gives comparatively high resolution since all four subspecies could be distinguished. In accordance with the antigenic differences in LPS mentioned above, F. t. subspecies novicida (040) was clearly discriminated from the other subspecies by the protein profiles obtained from all three surfaces used. More interesting was the finding that the two most clinically important subspecies, the highly virulent tularensis and the weakly virulent holarctica, could also be easily discriminated from each other. Phenotypic differentiation of these two main subspecies previously depended on glycerol fermentation and citrulline ureidase activity assays. However, these biochemical tests are rather unreliable [6]. Classically, these two subspecies were defined by performing virulence tests on rabbits. Thus, overall there was a clear differentiation of F. t. subspecies holarctica and tularensis and differentiation
between all four subspecies were evident, although very few strains of the latter two subspecies were compared. The F. t. subspecies mediasiatica strains, although clearly distinguishable from the F. t. subspecies tularensis group, cluster on the same lineage. This shows that they are related to subspecies tularensis. Interestingly, subspecies mediasiatica has lower virulence than subspecies tularensis [17], and SELDI analyses also facilitate the identification and subsequent characterization of proteins that differ between the two groups. The close relationship of subspecies mediasiatica with subspecies tularensis corroborates recent genotypic results obtained both by high-resolution MLVA analyses and DNA micro-array hybridizations [9–11,18]. It was also apparent that there are significant differences between the two F. t. subspecies novicida isolates (Fig. 3). There are very few isolates within subspecies novicida, and several isolates have been designated novicida-like [19]. Strain 040 is considered to be the type strain and strain 159 (fx2) was originally characterized as an atypical, non-cysteine-requiring novicida-like F. t. [20]. Biochemical and recently published genetic evidence suggests that Japanese holarctica isolates represent a fifth F. t. subspecies [11,17,18,21–23]. Our PCA also indicates that the two holarctica strains isolated in Japan may be members of a separate subgroup. However, the resolution of the protein profiling method and the number of strains is insufficient to clearly differentiate the Japanese strains from the other strains included in this study. The resolution of the method was limited to the subspecies level. This was best illustrated by the strains within subspecies holarctica. It is known from MLVA analysis that strains within subspecies holarctica show less diversity than strains within subspecies tularensis [11]. Hence, the bootstrap values for the branches separating the individual holarctica strains are generally lower than the corresponding values for individual strains within subspecies tularensis (Fig. 2a and b), except that holarctica strains 069 and 074 were clearly separated. The latter finding is probably misleading, since strain 069 is the same strain as strain 074, but recovered after passage through laboratory animals. Thus, branching within the subspecies should be interpreted cautiously. Cluster analysis based on protein profiles from two surfaces resulted in longer tree branch lengths compared with data generated from three surfaces, but the overall topology of the tree was the same, i.e. it did not increase the resolution beyond the subspecies level. The proportion of differentially expressed proteins increased when data from two surfaces were used rather than data from a single surface. However, this ratio did not increase further when cumulated data from three surfaces were used rather than two. The observed stability of the SELDI profiles across a 100-fold range of protein concentrations underlines the techniqueÕs potential for comparative protein expression
M. Lundquist et al. / FEMS Microbiology Letters 243 (2005) 303–310
profiling, which is attributable to the relatively low number of protein molecules in a standard bacterial sample that are able to bind to a SELDI surface compared to its capacity. Typically, there is an intermediate concentration range, usually spanning around two orders of magnitude, where the profile is stable, or changes only slightly. Moreover, we observed high reproducibility in the protein profiles. The main reasons for this are the active binding of the protein to the homogenous surface of the proteinchip array and the data collection method. In SELDI, uniform distribution of the protein is promoted by its active binding to the surface, while the data collection is based on automated scanning of large parts of the surface area, which yields a representative average signal. This approach allows the mass and intensity of a protein peak to be used, and not just its presence or absence, in data analysis. An advantage with SELDI techniques is that differentially expressed proteins that are detected can be further characterized, purified, and identified [24,25]. We identified several proteins that differed in abundance between the subspecies. In this context further characterization of differently expressed proteins between the two most clinically important subspecies, the highly virulent subspecies tularensis and the weakly virulent subspecies holarctica, are of especial interest. The results show that the protein profiling approach based on SELDI-TOF could be used for rapid high-resolution phenotypic identification of bacteria. In addition, the underlying information on differences in individual proteins yielded by the generated protein profiles is useful for pinpointing functional differences between the analyzed strains. Acknowledgement This study was supported by the Swedish MoD, Project No. A4854. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/ j.femsle.2004.12.020. References [1] Hopla, C.E. and Hopla, A.K. (1994) Tularemia In: Handbook of Zoonoses (Beran, G.W., Ed.), 2nd edn, pp. 113–126. CRC Press, Boca Raton. [2] Saslaw, S., Eigelsbach, H.T., Prior, J.A., Wilson, H.E. and Carhart, S. (1961) Tularemia vaccine study. I. Intracutaneous challenge. Arch. Int. Med. 107, 689–701. [3] Saslaw, S., Eigelsbach, H.T., Prior, J.A., Wilson, H.E. and Carhart, S. (1961) Tularemia vaccine study. II. Respiratory challenge. Arch. Int. Med. 107, 702–714.
309
[4] Rotz, L.D., Khan, A.S., Lillibridge, S.R., Ostroff, S.M. and Hughes, J.M (2002) Public health assessment of potential biological terrorism agents. Emerg. Infect. Dis. 8, 225–230. [5] Dennis, D.T., Inglesby, T.V., Henderson, D.A., Bartlett, J.G., Ascher, M.S., Eitzen, E., Fine, A.D., Friedlander, A.M., Hauer, J., Layton, M., Lillibridge, S.R., McDade, J.E., Osterholm, M.T., OÕToole, T., Parker, G., Perl, T.M., Russell, P.K. and Tonat, K. (2001) Tularemia as a biological weapon: medical and public health management. JAMA 285, 2763–2773. [6] Sjo¨stedt, A. (2005) Francisella, The proteobacteria, Part B In: (Brenner, D.J., Kreig, N.R., Staley, J.T. and Garrity, G.M., Eds.) 2nd ed., BergeyÕs Manual of Systematic Bacteriology. Springer, New York, pp. 200–210. [7] Hollis, D.G., Weaver, R.E., Steigerwalt, A.G., Wenger, J.D., Moss, C.M. and Brenner, D.J. (1989) Francisella philomiragia comb. nov. (formerly Yersinia philomiragia ) and Francisella tularensis biogroup novicida (formerly Francisella novicida) associated with human disease. J. Clin. Microbiol. 27, 1601–1608. [8] Grunow, R., Splettstoesser, W., McDonald, S., Otterbein, C., OÕBrien, T., Morgan, C., Aldrich, J., Hofer, E., Finke, E.J. and Meyer, H. (2000) Detection of Francisella tularensis in biological specimens using a capture enzyme-linked immunosorbent assay, an immunochromatographic handheld assay, and a PCR. Clin. Diagn. Lab. Immunol. 7, 86–90. [9] Farlow, J., Smith, K.L., Wong, J., Abrams, M., Lytle, M. and Keim, P. (2001) Francisella tularensis strain typing using multiplelocus, variable-number tandem repeat analysis. J. Clin. Microbiol. 39, 3186–3192. [10] Johansson, A., Go¨ransson, I., Larsson, P. and Sjo¨stedt, A. (2001) Extensive allelic variation among Francisella tularensis strains in a short-sequence tandem repeat region. J. Clin. Microbiol. 39, 3140–3146. [11] Johansson, A., Farlow, J., Larsson, P., Dukerich, M., Bystro¨m, M., Fox, J., Chu, M.C., Forsman, M., Sjo¨stedt, A. and Keim, P. (2004) World-wide genetic relationships among Francisella tularensis isolates using multiple-locus variable-number tandem repeat analysis (MLVA). J. Bacteriol. 186, 5808–5818. [12] Lay, J.O. (2001) MALDI-TOF spectrometry of bacteria. Mass. Spectrom. Rev. 20, 172–194. [13] Ryzhov, V. and Fenselau, C. (2001) Characterization of the protein subset desorbed by MALDI from whole bacterial cells. Anal. Chem. 73, 746–750. [14] Dai, Y., Li, L., Roser, D.C. and Long, S.R. (1999) Detection and identification of low-mass peptides and proteins from solvent suspensions of Escherichia coli by high performance liquid chromatography fractionation and matrix assisted laser desorption/inoization mass spectrometry. Rapid Commun. Mass. Spectrom. 13, 73–78. [15] Sjo¨stedt, A, Kuoppa, K, Johansson, T and Sandstro¨m, G (1992) The 17 kDa lipoprotein and encoding gene of Francisella tularensis LVS are conserved in strains of Francisella tularensis. Microb. Pathog. 13, 243–249. [16] Sandstro¨m, G., Ta¨rnvik, A., Wolf-Watz, H. and Lo¨fgren, S. (1984) Antigen from Francisella tularensis: nonidentity between determinants participating in cell-mediated and humoral reactions. Infect. Immun. 45, 101–106. [17] Olsufjev, N.G. and Meshcheryakova, I.S. (1983) Subspecific taxonomy of Francisella-tularensis. Int. J. Syst. Bacteriol. 33, 872–874. [18] Broekhuijsen, M., Larsson, P., Johansson, A., Bystro¨m, M., Eriksson, U., Larsson, E., Prior, R.G., Sjo¨stedt, A., Titball, R.W. and Forsman, M. (2003) Genome-wide DNA microarray analysis of Francisella tularensis strains demonstrates extensive genetic conservation within the species but identifies regions that are unique to the highly virulent F. tularensis subsp. tularensis. J. Clin. Microbiol. 41, 2924–2931.
310
M. Lundquist et al. / FEMS Microbiology Letters 243 (2005) 303–310
[19] Whipp, M.J., Davis, J.M., Lum, G., de Boer, J., Zhou, Y., Bearden, S.W., Petersen, J.M., Chu, M.C. and Hogg, G. (2003) Characterization of a novicida-like subspecies of Francisella tularensis isolated in Australia. J. Med. Microbiol. 52, 839–842. [20] Clarridge, J.E., Raich, T.J., Sjo¨stedt, A., Sandstro¨m, G, Darouiche, R.O., Shawar, R.M., Georghiou, P.R., Osting, C. and Vo, L. (1996) Characterization of two unusual clinically significant Francisella strains. J. Clin. Microbiol. 34, 1995–2000. [21] Sandstro¨m, G., Sjo¨stedt, A., Forsman, M., Pavlovich, N.V. and Mishankin, B.N. (1992) Characterization and classification of strains of Francisella tularensis isolated in the central Asian focus of the Soviet Union and in Japan. J. Clin. Microbiol. 30, 172–175. [22] Johansson, A., Ibrahim, A., Go¨ransson, I., Eriksson, U., Gurycova, D., Clarridge, J.E. and Sjo¨stedt, A. (2000) Evaluation of PCR-based methods for discrimination of Francisella species and subspecies and development of a specific PCR that distinguishes
the two major subspecies of Francisella tularensis. J. Clin. Microbiol. 38, 4180–4185. [23] Thomas, R., Johansson, A., Neeson, B., Isherwood, K., Sjo¨stedt, A., Ellis, J. and Titball, R.W. (2003) Discrimination of human pathogenic subspecies of Francisella tularensis by using restriction fragment length polymorphism. J. Clin. Microbiol. 41, 50–57. [24] Thulasiraman, V., McCutchen-Maloney, S.L., Motin, V.L. and Garcia, E. (2000) Factors in Yersinia pestis using SELDI ProteinChip System. BioTechniques 30, 428–432. [25] Diamond, D.L., Zhang, Y., Gaiger, A., Smithgall, M., Vedvick, T.S. and Carter, D. (2003) Use of ProteinChip array surface enhanced laser desorption/ionization time-of-flight mass spectrometry (SELDI-TOF MS) to identify thymosin b-4, a differentially secreted protein from lymphoblstoid cell lines. J. Am. Soc. Mass. Spectrom. 14, 760–765.
