点击显示 收起
Department of Medicine, Section of Infectious Diseases
Department of Medical Microbiology and Immunology, University of Wisconsin, Madison
Gene expression analysis after the host-pathogen interaction is revolutionizing our understanding of the host response to infection. Numerous studies have utilized microarray analysis to follow host cell transcriptome alterations in response to interactions with infectious pathogens. However, similar analyses of pathogen transcriptional adaptation at the infection site have been limited. Understanding the nature of this interaction from the pathogen perspective at different sites and stages of infection is central to strategies for development of new anti-infective therapies. Toward this end, we developed a protocol to analyze changes in gene expression for a eukaryotic pathogen, Candida albicans, during systemic infection in mice. The experimental approach takes advantage of the resistance of the cell wall of many fungal pathogens to cell lysis, relative to mammalian cells. After lysis of mammalian cells, the tissue mixture containing fungal cells is depleted of mammalian RNA by centrifugation, followed by enzymatic digestion. RNA-digesting enzymes are then inhibited before eukaryotic cell lysis and RNA isolation. The protocol provides a reproducible quantity of RNA based on pathogen cell number. The quality of the RNA allowed reliable downstream transcriptional analysis using reverse-transcription polymerase chain reaction and microarrays. The in vivo gene expression data confirmed involvement of several putative pathogenesis genes. More importantly, the results provided a wealth of biologically interesting hypotheses to direct future investigation.
Understanding the response of both the host and the pathogen during interactions at the infection site is critical for the development of new anti-infective strategies. Numerous studies have utilized microarray analysis to follow host cell transcriptome alterations in response to pathogen interactions [13]. Similar analyses of pathogen transcriptional adaptation at the infection site have been limited. Methods have been developed to isolate prokaryotic RNA from a mixture of host and pathogen RNA at the infection site [4, 5]. However, a similar approach to the isolation of eukaryotic pathogen RNA from a mammalian host has not been developed. Toward this end, we adapted technology to analyze the change in gene expression for a eukaryotic pathogen, Candida albicans, during systemic infection in mice.
Few investigations have monitored gene expression in C. albicans in response to interaction with host tissue [69]. Studies using an in vivo model have utilized a mucosal infection site [8]. The superficial nature of this infection allows the pathogen to be easily removed by lavage. Systemic infection with pathogens, however, involves dissemination to multiple internal organs, such that separation of the pathogen and host cells has not been possible. Furthermore, once the cells are disrupted for RNA isolation, it is difficult to separate the pathogen nucleic acid from that of the mammalian host. Other studies have utilized ex vivo exposure to a single type of host cells, such as macrophages or red blood cells (RBCs) [6, 7].
The low abundance of the pathogen RNA relative to that of the host at the infection site adds an additional hurdle to downstream RNA processing. For prokaryotic pathogens, protocols based on binding of the polyadenylated eukaryotic mRNA to an oligo-dT platform have been designed to remove mammalian RNA, leaving the prokaryotic mRNA behind [4].
The approach described here exploits a major difference between mammalian cells and eukaryotic pathogens. Many eukaryotic pathogens, including all fungi, possess a substantial cell wall that provides protection against mechanical, chemical, and osmotic environments [10]. In the present study, we sequentially lyse mammalian and pathogen cells to allow collection of RNA from the latter. The initial lysis of the mammalian cells allows removal of >75% of the mammalian RNA from complex tissue mixtures before pathogen RNA collection. We demonstrate successful use of the collected pathogen message for downstream analysis.
MATERIALS AND METHODS
Mice and C. albicans.
ICR mice weighing 2427 g were used. Mice were rendered neutropenic by administration of cyclophosphamide before infection (150 mg/kg 4 days and 100 mg/kg 1 day) to allow adequate growth of C. albicans. A clinical isolate, C. albicans K1, was used.
C. albicans growth in vivo.
Growth from subculture was diluted, grown to log phase at 37°C, washed, and resuspended in 0.85% NaCl to a concentration of 5.5 log10 cfu/mL. The inoculum was injected into mice via the tail vein. Groups of 2 mice were killed at 26 h intervals for 33 h. We chose the kidney as the end organ, since this is the most extensively characterized infection model for this pathogen. Kidneys were removed and processed for colony-forming unit enumeration. Results were expressed as the mean number of colony-forming units in 2 kidneys from 2 mice. Groups of 10 mice were killed at various time points for RNA isolation.
Organ processing to isolate pathogen RNA.
Kidneys from 10 mice were harvested per time point. Kidneys from each mouse were placed in 1 mL of water and homogenized with a mechanical grinder. Homogenates were passed through 4 plies of filter discs (2225-m pores, 25-mm diameter) (Miracloth; Calbiochem) to remove large tissue debris. Flow through from 10 animals was pooled and rapidly frozen with liquid nitrogen to stop transcription.
The flow through was treated with DNAse (4 U/mL; RQ1 RNase Free DNase) and Triton X-100 (final concentration, 1%) for 20 min at 37°C. The tissue suspensions were then centrifuged at 500 g for 5 min to further remove animal tissue. The supernatant was collected and further centrifuged (4000 g for 3 min). The pellet of yeast cells was resuspended in 269 L of RNase-free water.
RNAse digestion.
The pellet was then treated to digest contaminating mammalian RNA before the breakage of the yeast cell wall. Digestion was accomplished by addition of 30 L of 10× RNase buffer (300 mmol/L NaAc, 10 mmol/L TRIS, and 5 mmol/L EDTA) and 1 L of a commercial cocktail (Ambion Technologies) of RNase A (500 U/mL) and RNase T1 (20,000 U/mL); the mixture was then incubated for 20 min at 37°C.
RNAse inhibition and total RNA isolation.
Before isolation of pathogen RNA, the RNase enzymes were inactivated by the addition of 20 L of SUPERase RNase Inhibitor (20 U/L) and incubated for 20 min at 37°C. The yeast pellet was next processed for total RNA isolation and stabilization by use of the hot phenol method [11]. In vivoextracted RNA samples were analyzed using LabChip technology (Agilent Bioanalyzer 2100) [12]. The entire procedure, from tissue harvest to collection of yeast RNA, was completed in <2 h.
Organism growth in vitro.
C. albicans was also grown in vitro for comparison of gene expression. A subculture was diluted to an OD600 of 0.02 in 6× 100 mL of yeast peptone dextrose in 500-mL flasks and grown at 37°C and 250 rpm. The culture was sampled every 3 h for measurement of OD600 and plating for colony enumeration. At various time points, cells were harvested and frozen in liquid nitrogen. The cells were then processed in a manner identical to that used for the kidney tissue.
Quantitative reverse-transcription polymerase chain reaction (RT-PCR).
Quantitative RT-PCR was used to compare mRNA abundance of the genes of interest [15]. TaqMan probe and primer sets were designed for the CDR1, CDR2, and actin (ACT1) genes of C. albicans and the glyceraldehyde-3-phosphate-dehydrogenase (GAPDH) gene of Mus musculus (table 1), by use of Primer Express (version 1.5; Applied Biosystems) [20]. The Qiagen RT-PCR Kit was used in an ABI PRISM 7700. Reactions were performed in accordance with the kit manufacturer's instructions. Data analysis was completed using a modification of the CT method of Livak and Schmittgen [21]. For each amplification run, the threshold cycle, CT, of the target was normalized to the CT of the ACT1 gene (CT = C - C). We compared the trend in CT over time from cells grown in vitro and from those isolated directly from the kidneys. Assays were also performed on three 10-fold cDNA dilutions to ensure similar primer efficiencies among the various targets (data not shown).
Impact of RNase treatment on mouse and C. albicans RNA.
We examined the abundance of a constitutively expressed gene from both C. albicans and the mouse to estimate the impact of the RNase step on relative quantities of total RNA from both species. The ACT1 and GAPDH genes were used for C. albicans and the mouse, respectively [13, 14]. mRNA abundance of the genes from each species was measured using quantitative RT-PCR [15].
In 5 experiments, 2 groups of 10 mice were infected and their kidneys harvested during midlog-phase growth. The first group was processed as described above. The second group was processed identically, except without the RNase digestion step. RT-PCRs were performed in duplicate.
Comparison of in vivo and in vitro expression by use of RT-PCR.
We next examined the mRNA abundance of C. albicans genes from cells grown in vitro and in vivo. We examined expression of 2 C. albicans genes responsible for drug efflux, CDR1 and CDR2 [16, 17]. The expression of both CDR1 and CDR2 has been shown to increase during log-phase growth and to decline during stationary-phase growth [18, 19]. We would not anticipate differential expression of these genes in vitro versus in vivo. We collected cells from the mid-log phase, late log phase, and stationary phase of growth, both from broth and from the kidneys of infected mice. RNA was extracted from cells grown under both conditions and processed as described above. Quantitative RT-PCR was used to compare the relative mRNA abundance of each gene for each growth phase. Two biological and 3 assay replicates were performed for each condition.
Comparison of in vivo and in vitro expression by use of microarrays.
Studies were then undertaken to compare the global response of C. albicans to growth in vivo, by use of microarrays. The in vivo transcriptional profile was compared with that of cells grown to a similar phase in vitro. Kidneys from 20 mice were collected during midlog-phase growth. Cells grown in two 100-mL cultures were also collected during midlog-phase growth (OD600, 0.7l). After initial collection of in vitrogrown cells by centrifugation, the cells were processed similarly to those grown in vivo. Two biological and technical replicates (hybridizations) were performed.
Microarrays were produced at the Biotechnology Research Institute, National Research Council, Montreal [22]. The microarray is based on C. albicans SC5314 and consists of 6737 PCR open reading frames (ORFs) (98% of ORFs).
cDNA synthesis, RNA amplification, and labeling.
Cell populations were collected as described above. Total RNA was used to amplify mRNA by use of the Roche cDNA Synthesis Kit. Amino-allyllabeled amplified RNA (aRNA) synthesis was performed using the MegaScript T7 Kit (Ambion). The quality of the aRNA was estimated as described above. The yield of each reaction was estimated using the NanoDrop spectrophotometer. The purified aRNA was then coupled to the appropriate Cy dye. In all hybridizations, RNA from the in vivo samples was labeled with Cy5, and RNA from the in vitro samples was labeled with Cy3. The uncoupled Cy dye was quenched and fragmented before hybridization.
Hybridization.
Slides were prehybridized for 1 h at 42°C with a prehybridization solution (5× standard saline citrate , 0.1% SDS, and 50× Denhardt's solution), tRNA, and denatured genomic DNA. The 2 aRNA targets were resuspended with the hybridization solution, heat denatured, cooled, and applied to the DNA microarray slide for overnight hybridization at 42°C. The slide was then washed once for 10 min at 42°C with `1× SSC and 0.2% SDS; twice for 10 min at 37°C with 0.1× SSC and 0.2% SDS; and 4 times at ambient temperature in 0.1× SSC, for 3 min/wash.
Data analysis.
Microarrays were scanned using ScanArray 5000 (Packard BioScience). The intensity of the spots was quantified using QuantArray (GSI Lumonics). Data were analyzed using Excel (Microsoft) and Spotfire. The signal was normalized slide to slide on the basis of median intensity. To be included in the analysis, each spot had to satisfy 3 criteria [22]: (1) the signal intensity minus half of the SD had to be greater than the local background plus half of the SD, (2) the signal intensity had to be within the dynamic range of the photomultiplier tube, and (3) the raw intensities of the duplicate spots for each gene had to be within 50% of each other. For spots that met these criteria, the intensity of the expression from yeast cultured in vivo was divided by that from yeast cultured in vitro. Results presented are the averages of independent experiments.
Only ORFs that were modulated at least 2-fold in both replicate populations were included in the analysis. Because of the potential for homology between C. albicans and the mouse, we compared sequence homology by use of the Blastn algorithm for all genes that were highly expressed in vivo. The sequence of the C. albicans PCR product spotted on the array was compared by Blastn with the M. musculus genome.
RESULTS
Mammalian cell lysis and viable pathogen cell recovery.
Intact renal cells were not seen in staining of tissue homogenate (data not shown). The viable cell burden after each of the processing steps was similar to the cell count in the initial tissue homogenate.
Pathogen RNA quantity, quality, and mammalian contamination.
The mean ± SD quantity of presumed yeast total RNA isolated from the 10 pooled kidneys was 601 ± 116 ng. The in vivo extracted RNA samples analyzed by use of LabChip technology proved to be of a quality similar to that of the RNA extracted from cultures grown in vitro. Degradation, as determined on the basis of the ribosomal peak ratios (mean ± SD, 1.3 ± 0.23) and the bioanalyzer electropherogram, was not observed (figure 1). Spectrophotometer measurements of the 260/280 nm absorbance ratios ranged from 1.95 to 2.10.
The relative amounts of mouse and C. albicans RNA present in the final collection was estimated by comparing expression of constitutive genes from both M. musculus and C. albicans. Treatment of the tissue/yeast mixture with RNase before yeast cell RNA isolation resulted in depletion of mouse RNA and enrichment for C. albicans RNA (figure 2). Linear extrapolation of CT values from these experiments on the basis of the primer efficiency studies suggests that RNase treatment reproducibly eliminated 71%75% of murine RNA. Conversely, the samples were enriched for yeast RNA at least 70-fold.
Comparison of in vivo and in vitro expression by use of microarrays.
The microarray images were of high quality, and results were similar in the replicate experiments. Of the up-regulated C. albicans genes, only 10 genes with homology to genes within the mouse genome were identified by sequence analysis (table 2).
Adaptation of C. albicans to the renal infection site resulted in differential expression of 19% of all genes present on the microarray. Among the differentially expressed genes, the majority were down-regulated (85%). Fifty percent of the differentially expressed genes encoded a protein of known function in either C. albicans or S. cerevisiae. The functional categories of the differentially expressed genes are shown in tables 3 and 4 (array results available at: http://infectiousdisease.medicine.wisc.edu/andesinvivoarray). After infection of mouse kidney, the largest group of genes down-regulated in response to this in vivo environment was related to glucose utilization. Genes associated with transcriptional regulation were the most common functional group induced during in vivo growth. Of particular interest among the induced group were several genes previously shown to be important in C. albicans pathogenesis, including those encoding secreted enzymes and morphologic specific proteins.
DISCUSSION
The infection site encountered by a microbe exposes the organism to a unique environment. The outcome of the host-pathogen interaction is in large part due to selective gene expression in both the host and pathogen. Consequently, understanding pathogen gene expression in vivo is central to understanding how pathogens interact with or disrupt host cell function and produce disease. A number of technical obstacles have precluded routine use of in vivo models for the study of pathogen gene expression. Thus, most investigations monitoring pathogen transcriptional signatures have been performed in vitro.
There are few examples of differential gene expression examined using in vivo, compared with in vitro, test systems. In vivo bacterial models have been adapted to investigate in vivo host-driven genotypic events [5, 2325]. Recently, a biotechnology company released a product for isolation of prokaryotic RNA from mammalian tissues [4]. Technology for the use of infection models to investigate the transcription profile of eukaryotic pathogens has not been developed.
C. albicans is the most common fungal pathogen in humans and has a unique ability to adapt to numerous human microenvironments [26]. Nearly all pathogenesis investigations have utilized in vitro model systems. However, recently, several groups have examined gene expression in response to the host. The first group collected organisms from an oral mucosal infection and identified the importance of a family of aspartyl proteinases [8]. The majority of organisms at the mucosal interface are removed by lavage, resulting in minimal mammalian cell contamination. During systemic infection, however, organisms invade tissues of most internal organs. Unfortunately, collection of organisms from these organs without mammalian cell contamination is difficult. To circumvent this problem, 2 groups have investigated the interaction between C. albicans and single mammalian cell types ex vivo. Lorenz and Fink examined C. albicans gene expression associated with exposure to macrophage cell culture [7]. Similarly, Fradin et al. observed the response of C. albicans after exposure to RBCs and plasma [6]. Both studies identified important pathogenesis information and point to the relevance of the investigation of the host-pathogen interaction for this organism.
The current investigation describes a simple protocol that allows for observation of the transcriptional response of C. albicans directly from an internal organ. The approach takes advantage of the resistance of the cell wall of fungi to lysis, relative to mammalian cells.
After lysis, the passage of tissue mixture through a large-pore filter allowed fungal organisms to pass through and retained large particulate tissue debris. Quantitative culture after each step in the isolation procedure detected neither a loss of viable C. albicans cells nor a change in the relative cell morphotypes. After filtration, the homogenate was rapidly frozen to halt transcription. Additional particulate material was subsequently removed by differential centrifugation. The fungal pellet was then treated to digest the host RNA. After this enzymatic step, it was necessary to inactivate the RNase before further yeast processing. In the absence of inhibitor, yeast RNA yield was small, relative to total mixture RNA, and variable. Replicate studies with a similar number of viable C. albicans cells resulted in a remarkably similar RNA yield and integrity. Most importantly, the quality of the RNA allowed downstream transcriptional analysis.
The C. albicans genes CDR1 and CDR2 were compared in the initial expression study by use of RT-PCR. Two experimental variables were examined. The first variable involved the cell growth environment (in vitro or in vivo). The genes, CDR1 and CDR2, were specifically chosen because we did not anticipate differences in expression relative to growth in vitro or in vivo. The second variable examined was phase of cell growth. The abundance of both CDR1 and CDR2 mRNA has been demonstrated to increase during log-phase growth and to decline during stationary-phase growth. In the present study, the trend in relative mRNA abundance related to growth phase was similar for both genes in vitro and in vivo. The expression of both CDR1 and CDR2 declined during stationary-phase growth in vitro as well as in the cells grown in mice.
Adaptation of C. albicans to the renal infection site resulted in differential expression of 19% of all genes present on the microarray. Of the differentially expressed genes, only 14% were up-regulated. This general trend of transcript repression is similar to that described when in vivo bacterial models are used [5]. Among the most highly expressed genes were those associated with hyphal growth. Expression of genes associated with this pathway at a host site of invasive C. albicans growth is not surprising, given the proposed invading nature of the hyphal morphotype. Several of the genes found to be differentially expressed in the present study have also been identified by use of other technologies, supporting the validity of the current protocol. For example, similar to Lorenz and Fradin, we found that C. albicans responded by down-regulating the glycolytic pathway, which suggests a glucose-poor environment at the kidney infection site [6, 7]. We found a number of other similarly expressed genes, including several related to stress response and amino acid synthesis and transport, further validating our results and the use of mRNA from the infection site for microarray analysis. Future studies should examine expression profiles at other organ sites and may identify important site-specific differences.
We have identified potential limitations of our approach. Some of the genes expressed in vivo could reflect the handling procedure during RNA extraction [5]. To account for the conditions of tissue processing, the environments were reproduced for the in vitrogrown cells. Thus, the predominant variable for both cell populations was the location of growth (in vivo vs. in vitro). We did attempt to halt transcription very early in the protocol, by means of rapid freezing, and stabilized RNA after the mammalian RNase step. The comparable abundance of the CDR genes in vitro and in vivo suggests that the processing did not remarkably impact RNA stability. Furthermore, total RNA degradation was not detected by use of the sensitive RNA bioanalyzer method.
We were also concerned that the hypo-osmotic lysis condition might result in up-regulation of genes associated with osmostic stress. In an attempt to circumvent this, we chose to expose the in vitrogrown cells to identical conditions after growth in media, to ensure that the samples differed only by the initial growth condition. Importantly, we did not identify differential expression of osmotic stress genes in the microarrays. It is also possible that detection of up-regulated genes may have resulted from the hybridization of remaining mouse transcript for genes with enough homology. Among the up-regulated genes, we identified <6% of the genes with any appreciable homology between the C. albicans and the mouse sequences.
In sum, understanding the nature of host-pathogen interactions at different stages and sites of infection is central to strategies for the development of new anti-infective therapies. The approach described here provides the means to survey the transcriptional profile of eukaryotic pathogens during the course of infection directly from mammalian host tissues. The protocol could be completed in most laboratory settings. The RNA isolation protocol should allow in vivo sampling from other end-organ infection sites and other eukaryotic pathogens.
Acknowledgment
We thank Jon Woods for his review of the manuscript and helpful comments.
References
1. Altamura M, Casale D, Pepe M, Tafaro A. Immune responses to fungal infections and therapeutic implications. Curr Drug Targets Immune Endocr Metabol Disord 2001; 1:18997. First citation in article
2. Ishibashi K, Miura NN, Adachi Y, et al. DNA array analysis of altered gene expression in human leukocytes stimulated with soluble and particulate forms of Candida cell wall -glucan. Int Immunopharmacol 2004; 4:387401. First citation in article
3. Mullick A, Elias M, Harakidas P, et al. Gene expression in HL60 granulocytoids and human polymorphonuclear leukocytes exposed to Candida albicans. Infect Immun 2004; 72:41429. First citation in article
4. MICROBEnrich Kit. Manual version 0408. Austin: Ambion. First citation in article
5. Hinton JCD, Hautefort I, Eriksson S, Thompson A, Rhen M. Benefits and pitfalls of using microarrays to monitor bacterial gene expression during infection. Curr Opin Microbiol 2004; 7:27782. First citation in article
6. Fradin C, Kretschmar M, Nichterlein T, Gaillardin C, d'Enfert D, Hube B. Stage-specific gene expression of Candida albicans in human blood. Mol Microbiol 2003; 47:152343. First citation in article
7. Lorenz MC, Fink GR. The glyoxylate cycle is required for fungal virulence. Nature 2001; 412:836. First citation in article
8. Naglik JR, Newport G, White TC, et al. In vivo analysis of secreted aspartyl proteinase expression in human oral candidiasis. Infect Immun 1999; 67:248290. First citation in article
9. Rubin-Bejerano I, Fraser I, Grisafi P, Fink GR. Phagocytosis by neutrophils induces an amino acid deprivation response in Saccharomyces cerevisiae and Candida albicans. Proc Natl Acad Sci 2003; 100:1100712. First citation in article
10. Ioannou YA, Chen FW. Quantitation of DNA fragmentation in apoptosis. Nucl Acids Res 1996; 24:9923. First citation in article
11. Kohrer K, Domdey H. Preparation of high molecular weight RNA. Methods Enzymol 1991; 194:398405. First citation in article
12. Mueller O, Lightfoot S, Schroeder A. RNA integrity numberstandardization of RNA quality control [publication no. 59891165EN]. Agilent Application Note. Palo Alto, CA: Agilent Technologies, 2004. First citation in article
13. Sabath DE, Broome HE, Prystowsky MB. Glyceraldehyde-3-phosphate dehydrogenase mRNA is a major interleukin 2-induced transcript in a cloud T-helper lymphocyte. Gene 1990; 91:18591. First citation in article
14. Losberger C, Ernst JF. Sequence of Candida albicans gene encoding actin. Nucl Acids Res 1989; 17:9488. First citation in article
15. Bustin SA. Absolute quantification of mRNA using real-time reverse transcription polymerase chain reaction assays. J Mol Endocrinol 2000; 25:16993. First citation in article
16. Sanglard D, Ischer R, Monod M, Bille J. Cloning of Candida albicans genes conferring resistance to azole antifungal agents: characterization of CDR2, a new multidrug ABC transporter. Microbiology 1997; 143:40516. First citation in article
17. Prasad R, De Wergifosse P, Goffeau A, Balzi E. Molecular cloning and characterization of a novel gene of Candida albicans, CDR1, conferring multiple resistance to drug and antifungals. Curr Genet 1995; 27:3209. First citation in article
18. Lyons CN, White TC. Transcriptional analyses of antifungal drug resistance in Candida albicans. Antimicrob Agents Chemother 2000; 44:2296303. First citation in article
19. Krishnamurthy S, Gupta V, Prasad R, Panwar SL, Prasad R. Expression of CDR1, a multidrug resistance gene of Candida albicans: transcriptional activation by heat shock, drugs and human steroid hormones. FEMS Microbiol Lett 1998; 160:1917. First citation in article
20. Livak KJ, Flood SJ, Marmaro J, Giusti W, Deetz K. Oligonucleotides with fluorescent dyes at opposite ends provide a quenched probe system useful for detecting PCR product and nucleic acid hybridization. PCR Methods Appl 1995; 4:35762. First citation in article
21. Livak KJ, Schmittgen TD. Analysis of relative gene expression data using real-time quantitative PCR and the 2-CT method. Methods 2001; 25:4028. First citation in article
22. Cowen LE, Nantel A, Whiteway MS, et al. Population genomics of drug resistance in Candida albicans. Proc Natl Acad Sci USA 2002; 99:92849. First citation in article
23. Dozois CM, Daigle F, Curtiss R 3rd. Identification of pathogen-specific and conserved genes expressed in vivo by an avian pathogenic Escherichia coli strain. Proc Natl Acad Sci USA 2003; 100:24752. First citation in article
24. Xu Q, Dziejman M, Mekalanos JJ. Determination of the transcriptome of Vibrio cholerae during intraintestinal growth and midexponential phase in vitro. Proc Natl Acad Sci USA 2003; 100:128691. First citation in article
25. Zhu G, Xiao H, Mohan VP, et al. Gene expression in the tuberculous granuloma: analysis by laser capture microdissection and real-time PCR. Cell Microbiol 2003; 5:44553. First citation in article
26. Pappas PG, Rex JH, Sobel JD, et al. Guidelines for treatment of candidiasis. Clin Infect Dis 2004; 38:16189. First citation in article