Abstract
Borrelia burgdorferi, a causative agent of Lyme disease, encodes a protein BBB07 on the genomic plasmid cp26. BBB07 was identified as a candidate integrin ligand based on the presence of an RGD tripeptide motif, which is present in a number of mammalian ligands for β1 and β3 integrins . Previous work demonstrated that BBB07 in recombinant form binds to β1 integrins and induces inflammatory responses in synovial cells in culture. Several transposon mutants in bbb07 were attenuated in an in vivo screen of the transposon library in mice. We therefore tested individual transposon mutant clones in single-strain infections in mice and found that they were attenuated in terms of ID50 but did not have significantly reduced tissue burdens in mice. Based on data presented here we conclude that BBB07 is not essential for, but does contribute to, B. burgdorferi infectivity in mice.
Keywords: Borrelia burgdorferi, BBB07, transposon, infectivity, mouse
Introduction
Lyme disease, caused by Borrelia burgdorferi , is the most common arthropod-borne disease in North America. Despite antibiotic treatment, some patients develop chronic arthritis following infection. BBB07 is a Borrelia protein that acts as an integrin ligand and stimulates production of pro-inflammatory mediators in human cells in vitro [1, 2]. Several bbb07 mutants in a previously constructed transposon library [3] showed decreased mouse infectivity in laboratory mice in a large screen. To determine whether BBB07 might directly affect infection in the mouse model, or may be less competitive in the large-scale screening of multiple strains, we tested the parental B. burgdorferi strain B31 5A18-NP1 (wild-type, WT) in comparison to two clonal derivatives of a transposon (Tn) mutant T07TC184 [3], in which the Tn insertion disrupts bbb07, to test the hypotheses that BBB07 affects B. burgdorferi infection in a mouse model. The dose at which 50 % of the mice were infected was determined for each strain, and bacterial burdens were quantified in a number of mouse tissues.
Aim
To test clonal transposon mutants in the B. burgdorferi gene bbb07 to determine whether the encoded protein has a role in murine infection.
Methodology
B. burgdorferi strains
B. burgdorferi strain 5A18-NP1 (WT) and derivative Tn mutants [3] in gene bbb07 were grown in standard BSK II medium with the appropriate selection (40 µg ml−1 for gentamicin and 200 µg ml−1 for kanamycin). The mutant strains used were designated T09TC519 (insertion ratio 0.04, i.e. the distance in base pairs between the 5′end of the gene and the insertion site, divided by the gene length in base pairs) and T07TC184 (insertion ratio 0.34). We chose these two strains because the Tn insertions were in the 5′ portion of the gene, whereas the RGD tripeptide starts at base pair 820 of 1095 (amino acid 273 of 365; ratio 0.75). Pilot comparisons of the Tn strain stocks with the parental strain revealed no differences in mouse infection by ID50 determination, but further inspection revealed that the stocks were not clonal, as reported in some cases in the original work [3]. We therefore plated suspensions of each strain stock as per standard protocol [4] to isolate single colonies. Colonies were picked into liquid medium, and aliquots of each culture were frozen as strain stocks and for analyses of the DNA contents. Each DNA preparation was screened for the presence of the Tn by PCR using primers 5′JC (GCG AAG TTG TTG ACC GA) and 3′JC (TAT TAA TCG GGA GTT TGA AC). The reaction steps were 95 °C for 15 min; 92 °C for 30 s; 52 °C for 1 min; 72 °C for 3 min; cycle to 2 (repeat 39×); then a final extension at 72 °C for 7 min. The presence of the Tn results in a shift in the size of the PCR product from ~1491 to ~3341 bp. No pure clonal derivatives were derived from T09TC519, but two that retained the full genomic plasmid contents of the parental strain, clones 5 and 7, were derived from T07TC184. T07TC184-5 (designated T184-5) and T07TC184-7 (designated T184-7) were then tested in the mouse infection model. For each experiment, PCR was used to ensure that the Tn was inserted into the bbB07 gene, and that all genomic plasmids present in the parental strain were retained, using a multiplex PCR protocol [5]. All experiments reported here were performed with cultures that (1) contained all genomic plasmids present in the WT parental strain, and (2) showed the absence (WT) or presence of the Tn insertion (clones T184-5 and T184-7).
Mouse infections
All animal work was performed according to a protocol approved by the Institutional Animal Care and Use Committee of the Medical College of Wisconsin. In the conventional infection model, C3H/HeN mice were inoculated subcutaneously with 101–109 B. burgdorferi WT, T184-5, or T184-7, or with phosphate-buffered saline (PBS) supplemented with 0.2 % heat-inactivated C3H/HeN serum as the vehicle control. Mice were euthanized at 1 week or 4 weeks following inoculation, and tissue samples or organs were cultured in BSK II supplemented with 50 µg ml−1 rifampin, 20 µg ml−1 phosphomycin and 2.5 µg ml−1 amphotericin B [6–8] to assess the presence of live Borrelia . Cultures were followed for up to 8 weeks post-inoculation before being declared negative. Subsequent independent infections using independent B. burgdorferi cultures refined the dose range to more precisely establish the ID50, so the number of mice (n) per dose varies from 5 to 15. ID50 determinations were estimated by a Probit regression model fitted to the culture data. In the model, we included the three groups and log10-transformed dose as covariates and response rates as outcome. The ID50 levels were compared by using Chi-square statistics for testing whether the effects for T07TC184-5 and T07TC184-7 on the dose–response relationship are equal to the corresponding effect of WT. The P values were multiple comparisons corrected by the Bonferroni method. To gain insight into possible differences in dissemination kinetics or colonization efficiencies, comparisons of culture positivity rates in different tissues after 1 and 4 weeks of infection were analysed using the Benjamini–Hochberg false discovery rate adjustment for P values (called the q-value). Bacterial burdens in a variety of tissues were measured by quantitative PCR (qPCR) using primers targeting recA ( B. burgdorferi ) and mouse (nidogen) DNA [6–11]. The quantitative PCR (qPCR) data were analysed for statistical significance using Kruskal–Wallis followed by Dunn’s multiple comparisons test in GraphPad Prism. Immediately before inoculation, and through the 4 week course, mouse ankle (tibiotarsal joint) measurements were taken at weekly intervals using digital callipers (Pittsburgh digital thickness gauge #66319). The anterior–posterior diameter was measured by extending the mouse rear leg manually and taking three measurements for each mouse ankle (left and right). The six values were averaged, and data were analysed by two-way analysis of variance (ANOVA) with the Geisser–Greenhouse correction.
Short-term mouse infections
In a separate set of experiments, we used a short-term infection model [9] to assess the possibility that the adhesive function of BBB07 previously reported [2] might affect interactions of the different B. burgdorferi strains with the vasculature in a variety of tissues. In this model, 108 bacteria were delivered via tail vein injection and allowed to circulate for 1 h. Unbound bacteria were then flushed out by perfusion with saline, and bacteria associated with different tissues/organs were enumerated using qPCR as described above. Each experiment used five mice per group, and was repeated once using independent cultures of B. burgdorferi for a total number (n) of 10 mice/group. Data were analysed for statistical significance using Kruskal–Wallis followed by Dunn’s multiple comparisons test in GraphPad Prism.
Results
Table 1 shows aggregate data obtained in at least three independent experiments for each B. burgdorferi strain. The ID50 for each strain, which was calculated on the basis of culture data, assesses the presence of live bacteria, at 4 weeks post-infection, a time at which B. burgdorferi is fully disseminated. The ID50 of the parental WT strain (5A18-NP1) was 291 bacterial cells/mouse, that of Tn mutant T184-5 was 1162 (P vs WT=0.046), and that of Tn mutant T184-7 was 4015 (P vs WT=0.0026). Thus the increases in ID50 of the Tn mutants in bbb07 were on the order of 4- to 14-fold as compared to the WT. There were no significant differences between the Tn mutants and the WT in the prevalence of culture positivity in any of the individual tissues assessed, indicating that tissue colonization was not affected in the Tn mutants as assessed by culture data.
Table 1.
Culture of mouse tissues at 4 weeks post-inoculation. Data are from three independent experiments. Mice injected with the vehicle control were not culture-positive
|
No. of cultures positive for B. burgdorferi growth/total |
||||||||
|---|---|---|---|---|---|---|---|---|
|
B. burgdorferi strain, dose |
Blood |
Bladder |
Heart |
Tibiotarsus |
Inoculation site |
Ear |
No. of mice positive/total |
% positive |
|
WT 5A18-NP1 |
|
|||||||
|
101 |
0/5 |
0/15 |
0/15 |
0/15 |
0/15 |
0/15 |
0/15 |
0 |
|
102 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0 |
|
103 |
1/5 |
7/15 |
7/15 |
6/15 |
7/15 |
10/15 |
11/15 |
73 |
|
104 |
2/5 |
5/5 |
5/5 |
5/5 |
5/5 |
5/5 |
5/5 |
100 |
|
105 |
1/5 |
15/15 |
15/15 |
15/15 |
15/15 |
15/15 |
15/15 |
100 |
|
106 |
1/15 |
15/15 |
14/15 |
12/15 |
15/15 |
6/15 |
15/15 |
100 |
|
107 |
nd |
10/10 |
10/10 |
10/10 |
10/10 |
10/10 |
10/10 |
100 |
|
109 |
nd |
10/10 |
10/10 |
10/10 |
10/10 |
10/10 |
10/10 |
100 |
|
Tn 184-5 |
|
|||||||
|
101 |
0/5 |
0/10 |
0/10 |
0/10 |
0/10 |
0/10 |
0/10 |
0 |
|
102 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0 |
|
103 |
2/5 |
2/10 |
2/10 |
2/10 |
2/10 |
2/10 |
2/10 |
20 |
|
104 |
2/5 |
5/5 |
5/5 |
5/5 |
5/5 |
5/5 |
5/5 |
100 |
|
105 |
1/5 |
10/10 |
10/10 |
10/10 |
10/10 |
10/10 |
10/10 |
100 |
|
106 |
0/15 |
15/15 |
14/15 |
15/15 |
15/15 |
7/15 |
15/15 |
100 |
|
107 |
nd |
5/5 |
5/5 |
5/5 |
5/5 |
5/5 |
5/5 |
100 |
|
109 |
nd |
5/5 |
5/5 |
5/5 |
5/5 |
5/5 |
5/5 |
100 |
|
Tn 184-7 |
|
|||||||
|
101 |
0/5 |
0/10 |
0/10 |
0/10 |
0/10 |
0/10 |
0/10 |
0 |
|
102 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0 |
|
103 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0 |
|
104 |
0/5 |
3/5 |
3/5 |
3/5 |
3/5 |
3/5 |
3/5 |
60 |
|
105 |
0/5 |
10/10 |
10/10 |
10/10 |
10/10 |
10/10 |
10/10 |
100 |
|
106 |
1/15 |
15/15 |
15/15 |
15/15 |
15/15 |
5/15 |
15/15 |
100 |
|
107 |
nd |
5/5 |
5/5 |
5/5 |
5/5 |
5/5 |
5/5 |
100 |
|
109 |
nd |
5/5 |
5/5 |
5/5 |
5/5 |
5/5 |
5/5 |
100 |
nd, not determined.
Bacterial burdens in different tissue sites of mice infected with the WT and the Tn mutants were determined by inoculating the mice with 106 B. burgdorferi cells of each strain to ensure that all mice were infected. At 4 weeks post-inoculation, tissues were harvested, and bacterial burdens were compared by qPCR (Fig. 1). The only significant difference seen was that both of the Tn mutant clones had higher burdens in the tibiotarsal joint (tibio) than the WT.
Fig. 1.
Bacterial numbers at 4 weeks post-subcutaneous inoculation of mice. Shown are the medians and ranges of qPCR results for 10 mice per group for the wild-type parental strains and two clonal transposon mutants in gene bbb07. The asterisk in the tibio (tibiotarsal joint) graph indicates that mice injected with the wild-type bacteria had significantly lower burdens than the mice injected with either of the mutant clones.
Measurements of the ankles (tibiotarsal joints) of mice were taken at weekly intervals at 0, 1, 2, 3 and 4 weeks of infection (Fig. 2). No significant differences were observed between the uninfected and WT-infected mice, so no further investigations of arthritis development were performed. It has previously been reported that B. burgdorferi B31-derived strains are less arthritiogenic than N40-derived strains [12], and our results are consistent with this observation.
Fig. 2.
Ankle joint measurements over time. The ankle (tibiotarsal joint) measurements of mice inoculated with the WT 5A18-NP1 or the vehicle control were taken at weekly intervals throughout infection up to 4 weeks, at which point the mice were euthanized. The average of six ankle measurements per time point for each mouse is plotted; the medians (horizontal lines) and ranges (error bars) for each group are shown in millimetres (mm). No significant differences between the groups of mice were observed at any time point.
To investigate the possibility that BBB07 may contribute to the dissemination of B. burgdorferi in mice, we also harvested tissues from infected mice at 1 week post-inoculation. The culture-based results are shown in Table 2. For the WT and both Tn mutants, the only significant differences in the percentage of tissues colonized between the 1 and 4 week time points were in the bladder and tibiotarsal joint (q=0.04 for all), so the mutants did not behave differently from the WT. Peripheral ear skin culture positivity was only significantly different between weeks 1 and 4 for T184-5 (q=0.04). These data indicate that if the Tn mutants have any defect in dissemination, it is negligible.
Table 2.
Culture of mouse tissues at 1 week post-inoculation. Mice injected with the vehicle control were not culture-positive.
|
No. of cultures positive for B. burgdorferi growth/total |
||||||||
|---|---|---|---|---|---|---|---|---|
|
B. burgdorferi strain, dose |
Blood |
Bladder |
Heart |
Tibiotarsus |
Inoculation site |
Ear |
No. of mice positive/total |
% positive |
|
WT 5A18-NP1 |
||||||||
|
101 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0 |
|
102 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0 |
|
103 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0 |
|
104 |
3/5 |
1/5 |
3/5 |
0/5 |
4/5 |
4/5 |
5/5 |
100 |
|
105 |
5/5 |
2/5 |
5/5 |
1/5 |
5/5 |
3/5 |
5/5 |
100 |
|
106 |
5/5 |
1/5 |
5/5 |
1/5 |
5/5 |
3/5 |
5/5 |
100 |
|
Tn 184-5 |
||||||||
|
101 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0 |
|
102 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0 |
|
103 |
1/5 |
0/5 |
0/5 |
0/5 |
1/5 |
0/5 |
1/5 |
20 |
|
104 |
1/5 |
1/5 |
1/5 |
0/5 |
1/5 |
1/5 |
1/5 |
20 |
|
105 |
3/5 |
1/5 |
3/5 |
0/5 |
3/5 |
0/5 |
3/5 |
60 |
|
106 |
5/5 |
0/5 |
4/5 |
1/5 |
5/5 |
2/5 |
5/5 |
100 |
|
Tn 184-7 |
||||||||
|
101 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0 |
|
102 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0 |
|
103 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0/5 |
0 |
|
104 |
5/5 |
0/5 |
5/5 |
0/5 |
4/5 |
1/5 |
5/5 |
100 |
|
105 |
5/5 |
0/5 |
5/5 |
0/5 |
5/5 |
2/5 |
5/5 |
100 |
|
106 |
4/5 |
1/5 |
5/5 |
0/5 |
5/5 |
4/5 |
5/5 |
100 |
We also investigated the possibility that BBB07 may contribute to tropism to the vasculature beds in different tissues using the short-term intravenous infection model [9]. In this model, T184-5 showed significantly higher burdens in the lung, liver and spleen, than either the WT or the T184-7 mutant. No other statistically significant differences were observed (Fig. 3). These data are considerably different from those obtained when other adhesins are tested in the same model [9, 13], in that decreases in bacterial burdens in tissues known to be colonized by B. burgdorferi , such as the heart, tibiotarsus, ear and skin, were not observed for the Tn insertion in bbb07. The increases in T187-5 burdens are only seen in tissues known for filtration activities, and are unlikely to be due to specific interactions between the bacteria and the vasculature.
Fig. 3.
Bacterial numbers at 1 h post-intravenous inoculation of mice. Shown are the medians and ranges of qPCR results for 10 mice per group for the wild-type parental strains and two clonal transposon mutants in gene bbb07. The asterisks in the spleen, liver and lung indicate that clone Tn 184-5 had significantly higher burdens than either the wild-type or the other mutant clone.
Discussion
bbb07 is the third gene in a three-gene operon. The upstream genes bbb05 and bbb06, together with gene bbb04, appear to comprise a chitobiose uptake system [14] that is not required for B. burgdorferi survival in either ticks or mice [15], but the function of bbb07 has not been described. It is possible that other transport systems can compensate for the defect in chitobiose transport, but there are no closely related genes encoded elsewhere in the genome. It was not clear whether bbb07 may encode a secretion signal or whether the protein may be exported or available to interact with host molecules on the bacterial cell surface; early versions of signal peptide predictors suggested yes, but more recent versions suggest that the protein does not have a secretion signal peptide.
In agreement with the previous in vivo Tn mutant screen data showing decreased infectivity [3], our in vivo data showed a modest increase in the ID50. However, the increased ID50 was on the order of 10-fold or less, which is less severe that might have been expected from the previous work [3]. Although this is speculative, the differences may in part be due to differences in the protocols, e.g. testing individual clones as opposed to pools of Tn mutants, as some clones in a Tn screen may be at a competitive disadvantage in comparison to other mutants in the pools. Our in vivo single-strain infection model provides a robust and accurate representation of the role of BBB07 in the infection and disease processes.
While early predictions suggested that BBB07 could potentially be a surface-localized protein, current algorithms suggest that BBB07 is not predicted to contain a secretion signal, and is therefore unlikely to be available on the bacterial surface for interaction with host integrins when produced by B. burgdorferi . Our data suggest that BBB07 plays a minor role in B. burgdorferi infection in laboratory mice.
Funding information
This work was supported by R01 AI118799 from the National Institutes of Health, and by the Department of Medicine, Division of Infectious Diseases, at the Medical College of Wisconsin.
Author contributions
B. H., investigation, writing drafts; P. A., investigation, Z. L., investigation, R. D., investigation; Z. Z. and N. H., formal analyses; N. H., writing; L. G., resources; T. L., resources; S. N., resources; J. C., writing, conceptualization, methodology, formal analyses, supervision, project administration, funding.
Conflicts of interest
The authors declare that there are no conflicts of interest.
Footnotes
Abbreviations: DNA, deoxyribonucleic acid; ID50, 50% infectious dose; PCR, polymerase chain reaction; qPCR, quantitative polymerase chain reaction; RGD, Arg-Gly-Asp; Tn, transposon; WT, wild-type.
References
- 1.Behera AK, Hildebrand E, Uematsu S, Akira S, Coburn J, et al. Identification of a TLR-independent pathway for Borrelia burgdorferi-induced expression of matrix metalloproteinases and inflammatory mediators through binding to integrin α3β1. J Immunol. 2006;177:657–664. doi: 10.4049/jimmunol.177.1.657. [DOI] [PubMed] [Google Scholar]
- 2.Behera AK, Durand E, Cugini C, Antonara S, Bourassa L, et al. Borrelia burgdorferi BBB07 interaction with integrin α3β1 stimulates production of pro-inflammatory mediators in primary human chondrocytes. Cell Microbiol. 2008;10:320–331. doi: 10.1111/j.1462-5822.2007.01043.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Lin T, Gao L, Zhang C, Odeh E, Jacobs MB, et al. Analysis of an ordered, comprehensive STM mutant library in infectious Borrelia burgdorferi: insights into the genes required for mouse infectivity. PLoS One. 2012;7:e47532. doi: 10.1371/journal.pone.0047532. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Samuels DS. Electrotransformation of the spirochete Borrelia burgdorferi . Methods Mol Biol. 1995;47:253–259. doi: 10.1385/0-89603-310-4:253. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Bunikis I, Kutschan-Bunikis S, Bonde M, Bergström S. Multiplex PCR as a tool for validating plasmid content of Borrelia burgdorferi . J Microbiol Methods. 2011;86:243–247. doi: 10.1016/j.mimet.2011.05.004. [DOI] [PubMed] [Google Scholar]
- 6.Ristow LC, Bonde M, Lin YP, Sato H, Curtis M, et al. Integrin binding by Borrelia burgdorferi P66 facilitates dissemination but is not required for infectivity. Cell Microbiol. 2015;17:1021–1036. doi: 10.1111/cmi.12418. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Ristow LC, Miller HE, Padmore LJ, Chettri R, Salzman N, et al. The β3-integrin ligand of Borrelia burgdorferi is critical for infection of mice but not ticks. Mol Microbiol. 2012;85:1105–1118. doi: 10.1111/j.1365-2958.2012.08160.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Curtis MW, Hahn BL, Zhang K, Li C, Robinson RT, et al. Characterization of stress and innate immunity resistance of wild-type and Δp66 Borrelia burgdorferi . Infect Immun. 2018;86 doi: 10.1128/IAI.00186-17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Caine JA, Coburn J. A short-term Borrelia burgdorferi infection model identifies tissue tropisms and bloodstream survival conferred by adhesion proteins. Infect Immun. 2015;83:3184–3194. doi: 10.1128/IAI.00349-15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Caine JA, Lin YP, Kessler JR, Sato H, Leong JM, et al. Borrelia burgdorferi outer surface protein C (OspC) binds complement component C4b and confers bloodstream survival. Cell Microbiol. 2017;19:e12786. doi: 10.1111/cmi.12786. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Morrison TB, Ma Y, Weis JH, Weis JJ. Rapid and sensitive quantification of Borrelia burgdorferi-infected mouse tissues by continuous fluorescent monitoring of PCR. J Clin Microbiol. 1999;37:987–992. doi: 10.1128/JCM.37.4.987-992.1999. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Chan K, Awan M, Barthold SW, Parveen N. Comparative molecular analyses of Borrelia burgdorferi sensu stricto strains B31 and N40D10/E9 and determination of their pathogenicity. BMC Microbiol. 2012;12:157. doi: 10.1186/1471-2180-12-157. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Lin Y-P, Chen Q, Ritchie JA, Dufour NP, Fischer JR, et al. Glycosaminoglycan binding by Borrelia burgdorferi adhesin BBK32 specifically and uniquely promotes joint colonization. Cell Microbiol. 2015;17:860–875. doi: 10.1111/cmi.12407. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Tilly K, Elias AF, Errett J, Fischer E, Iyer R, et al. Genetics and regulation of chitobiose utilization in Borrelia burgdorferi . J Bacteriol. 2001;183:5544–5553. doi: 10.1128/JB.183.19.5544-5553.2001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Tilly K, Grimm D, Bueschel DM, Krum JG, Rosa P. Infectious cycle analysis of a Borrelia burgdorferi mutant defective in transport of chitobiose, a tick cuticle component. Vector Borne Zoonotic Dis. 2004;4:159–168. doi: 10.1089/1530366041210738. [DOI] [PubMed] [Google Scholar]