DNA Microarray-Based Detection and Identification of Fungal Pathogens in Clinical Samples from Neutropenic Patients

MATERIALS AND METHODS

Design of capture probes and microarray preparation.The fungal oligonucleotide probes and two positive control oligonucleotides from human DNA were designed from the ITS1 sequences available in the GenBank database (www.ncbi.nlm.nih.gov/). Sequence alignments were performed using the Genomatix DiAlign (www.genomatix.de/cgi-bin/dialign/dialign.pl). By comparing the sequences of the ITS1 regions, oligonucleotides were generated from regions with a high level of variability between the different fungal species. One to three capture probes of various lengths (22 to 70 bases) per fungal species and two human ITS1 oligonucleotides as positive controls were designed (24 and 55 bases). As additional positive controls, we designed four capture probes (26 to 55 bases) based on the human glucose-6-phosphate dehydrogenase (G6PD) gene and two oligonucleotides (26 and 55 bases) based on the human ribosomal protein L19 (RPL19) gene.

Two plant capture probes of 55 bases from the A. thaliana genes AtrpL23a and ACT2 served as negative controls.

The ITS1 region of Candida parapsilosis, the third yeast responsible for candidemia, showed a high level of homology and cross-reactivity to other Candida species so it was not included.

To predict the potential cross-reactivities of the capture probes, additional database searches were performed by using the BLASTN program (www.ncbi.nlm.nih.gov/). Sequences of the capture probes and sequence sources used for primers and capture probes design are given in Table 2.

View this table:

• View inline
• View popup

TABLE 2.

Capture probe sequences derived from the ITS1 regions of the corresponding organisms and sequence sources

All oligonucleotides had a C12 spacer and were NH₂ modified at their 5′ ends for covalent coupling.

The synthetic oligonucleotides were diluted to 100 μM in 3× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate). Microarrays were prepared and treated as described previously (28). The microarray design is shown in Fig. 1. For hybridization data redundancy, two replicates were spotted onto the same slide.

• Open in new tab
• Download powerpoint

FIG. 1.

Schematic composition of glass microarray and grid location of fungal, human, and A. thaliana-specific oligonucleotides (capture probes). Grid locators (A1, A12, and D1 for Cy3) represented by a spotted Cy3-labeled arbitrary oligonucleotide are depicted as black boxes. Names of the fungal capture probes are derived from the first letter of the genus name and the first four letters of the species name and are numbered consecutively as follows: Calb1S (20), Candida albicans; Cdub1S (20) and Cdubl, Candida dubliniensis; Clusi, Candida lusitaniae; Cglab, Candida glabrata; Ctrop, Candida tropicalis; Rmicr, Rhizopus microsporus; Foxys, Fusarium oxysporum; Fsola, Fusarium solani; Mrace, Mucor racemosus; Sprol, Scedosporium prolificans; Tasah, Trichosporon asahii; Aflav, Aspergillus flavus; Aterr, Aspergillus terreus; and Afumi, Aspergillus fumigatus.

Chip evaluation.Hybridized glass slides were scanned with an Affymetrix GMS 418 array scanner (Affymetrix) by using the recommended settings for Cy3 fluorochrome. High-resolution images (10 μm/pixel) were saved in the TIF file format (16-bit) and further evaluated by the ImaGene 4.0 software tool package (BioDiscovery, Inc., Los Angeles, CA). The original signal and the signals of the two replicates from one hybridization were evaluated separately. The mean values and standard deviations were calculated. After background subtraction for each signal, false color mapping (BMP file format, 24-bit) was used to display resultant images as shown in Fig. 2. Dots with at least twice the signal intensities of the background intensities were evaluated as positive signals. Background intensities were calculated by the ImaGene 4.0 program.

• Open in new tab
• Download powerpoint

FIG. 2.

(A) Typical hybridization patterns using 5 ng genomic DNA of the different fungal species, presented by false color mapping. Glass microarrays were probed with Cy3-labeled DNA fragments amplified separately with the primer mixture for the fungal ITS1 PCR. (B) Hybridization patterns using 200 ng human DNA mixed with potential fungal DNA from clinical samples amplified separately with the primer mixture for the fungal and human ITS1 regions and for human housekeeping/A. thaliana genes and combined for hybridization. Fungus-specific signals are indicated by arrows. For assignment of the hybridization signals, compare Fig. 2B with Fig. 1 and Tables 4 and 5.

RESULTS

The multiplex PCR for amplification of the ITS1 regions (137 to 245 bp) of the 14 fungal pathogens showed different sensitivity thresholds for the genomic DNAs of the fungal species (Table 3). Even large amounts of genomic bacterial DNAs (5 ng) used as negative controls (Enterococcus faecalis DSM 2570, Escherichia coli DSM 787, Klebsiella pneumoniae DSM 4617, Proteus mirabilis DSM 788, Pseudomonas aeruginosa DSM 50071, Staphylococcus aureus DSM 799, and Streptococcus pneumoniae DSM 20566) were not amplified in the ITS1 multiplex PCR and did not give any signals after hybridization to the fungal DNA microarray. The genomic DNA of Penicillium notatum, a fungal organism that we did not aim to detect, was amplified by PCR and gave hybridization signals with one of three capture probes of Candida glabrata and one of two capture probes of Scedosporium prolificans. Additionally, for a human, ITS1-specific, 226-bp fragment, few nonspecific amplification products were visible when large amounts of human genomic DNA (200 ng) were used in the ITS1 multiplex PCR. After the hybridization of the human PCR products to the DNA microarray, there was no cross-reactivity between human DNA and fungal capture probes detected. There were also no cross-reactivities between any fungal or human PCR products and the two spotted A. thaliana negative control oligonucleotides. Hybridization with two Cy3-labeled oligonucleotides (50 μM), which were complementary to the two spotted A. thaliana capture probes, resulted in intense signals (data not shown).

The fungal and human capture probes were located on the glass microarray as shown schematically in Fig. 1. To establish the fungal DNA chip, 5-ng amounts of each of 14 genomic fungal DNAs were amplified separately by ITS1 multiplex PCR and the products were then hybridized. Every fungal organism showed a specific hybridization pattern after evaluation of the hybridization signals (Fig. 2A). Details about the correlation of the hybridization patterns and the fungal organisms are described in Table 4. In spite of single cross-reactivities, each species has a unique signature hybridization pattern.

Different hybridization conditions were tested and compared. The best hybridization results and an appropriate ratio between specific binding, signal intensity, and cross-reactivity were achieved by using the following hybridization and washing conditions: hybridization at 42°C using 50% formamide for 12 h following a washing step for 10 min in 1×, 0.5×, and 0.1× SSC at room temperature, respectively.

For Candida tropicalis, Fusarium oxysporum, and Rhizopus microsporus DNA, the detection limit of the array is above the detection limit of the hybridization preceding fungal ITS1 PCR. Additional serial 10-fold dilutions of the three DNAs were used to determine the detection threshold of the array. The ITS1 PCR detection limits were 1 pg for C. tropicalis, 10 pg for F. oxysporum, and 8 pg for R. microsporus, and the detection limits of the array were 300 pg DNA for C. tropicalis, 500 pg for F. oxysporum, and 300 pg for R. microsporus.

For the other fungal species, the detection threshold of the ITS1 PCR marked the detection limit of the array (Table 3).

Two hundred nanograms of genomic DNA directly from samples of patients with known or predicted IFI were used in the ITS1 fungal multiplex PCR and in positive and negative control PCRs. The template of the PCR was composed of a mix of human DNA and potentially present fungal DNA. The PCR products were used for hybridization to the DNA microarray.

The results of microarray investigations of blood, BAL, and tissue samples from 46 neutropenic immunocompromised patients are given in Table 5. Microarray results were compared to culture, histopathology, imaging, serologic, and clinical findings, according to the EORTC/MSG 2002 criteria for invasive fungal infections, and to Aspergillus-specific nested PCR (29) results, in summary demonstrating the feasibility of this tool.

For 22 of 22 patients (5 without IFI and 17 with possible IFI), negative conventional diagnostic results corresponded with negative microarray data. For eight patients with proven (n = 5) or probable (n = 3) IFI, results for four or two patients were confirmed by microarray testing, respectively. For patients with possible IFI (n = 5), positive microarray data were validated by other diagnostic findings. For 11 of 11 patients with possible IFI, the microarray results provided additional information about fungal pathogens. For one patient with proven invasive aspergillosis (determined from a brain abscess sample), DNA microarray results from blood samples were negative under antifungal conditions; for one patient with probable invasive aspergillosis, DNA microarray results from a BAL aliquot were negative; this patient also received antifungal treatment.

Genomic DNA and specific hybridization patterns of Candida albicans, Candida dubliniensis, Candida glabrata, Aspergillus fumigatus, Aspergillus flavus, Fusarium solani, Rhizopus microsporus, Scedosporium prolificans, and Trichosporon asahii were detected. One fungal pathogen was identified in 11 patients with IFI, and two or more fungal pathogens were likewise identified in 11 patients with IFI.

We also compared microarray findings to the results of a previously described and clinically validated Aspergillus-specific nested PCR assay (6, 29). For one patient, the Aspergillus nested PCR assay was positive for Aspergillus DNA, whereas the DNA microarray results were negative. For three patients, the Aspergillus PCR results for blood and tissue samples were positive, whereas the DNA microarray results were negative for Aspergillus DNA but positive for additional fungal species (C. albicans, C. dubliniensis, and C. glabrata). Differences between nested Aspergillus PCR (29) findings and the microarray results are caused by the higher sensitivity of the nested PCR assay for Aspergillus DNA. For five patients, positive Aspergillus PCR findings were in line with positive DNA microarray results for Aspergillus DNA.

Three representative hybridizations of DNA from a total of 46 patients (Table 5) isolated from clinical samples are shown in Fig. 2B. For the blood sample from patient 3, hybridization resulted in distinct hybridization signals for all three C. glabrata capture probes without any cross-reactivities. A corresponding blood culture was positive for Candida species that could not be further differentiated. A second example (patient 1) shows the hybridization results using DNA from a BAL sample. A blood culture from this patient identified A. fumigatus. The hybridization gave positive signals for A. flavus, A. fumigatus, F. oxysporum, and F. solani. The third example (patient 5) showed signals with the C. glabrata capture probes and the A. flavus probe. Additionally, the sample was positive by Aspergillus antigen testing (galactomannan enzyme-linked immunosorbent assay). The result of the nested PCR assay for Aspergillus species was negative for this patient.

Hybridization of the DNA from a blood sample of a healthy volunteer did not give any signal with the fungal capture probes.

DISCUSSION

In immunocompromised patients, the spectrum of clinically relevant fungal pathogens causing systemic infections with considerable morbidity and high mortality rates is changing (25). Therefore, we set out to establish a rapid, specific, and sensitive molecular method to detect and identify the most emerging fungal pathogens (26) from noncultured clinical samples in one assay.

Our DNA microarray is able to detect clinically relevant fungal pathogens at low detection thresholds. Results using clinical samples from immunocompromised patients with hematological malignancies show the usefulness of the microarray in the clinical context. The benefit of the fungal microarray is the potential to detect the most important 14 fungal pathogens in a specific IFI high-risk setting with one test; due to feasibility concerns, other fungi were not included. In view of the changing spectrum of emerging clinically relevant fungal pathogens causing IFI and the advent of novel antifungals, this new diagnostic approach meets urgent clinical needs, at least concerning the group of patients treated for hematological malignancies.

For our assay, sensitive and specific detection and identification of the fungal pathogens were achieved by designing primer pairs complementary to the highly conserved 18S and 5.8S regions of the fungal rRNA genes and oligonucleotide capture probes complementary to the more variable ITS1 regions which enabled a differentiation of fungal species (13, 14, 15, 16, 20). As part of the rRNA genes, the ITS1 regions were chosen as targets because they are present in numerous copies in the fungal genome. The design of the primer pairs from highly conserved regions of a multicopy gene lead to a high sensitivity of the PCR and to an amplification of the target sequences with only a few primer combinations. Reasons for the different detection thresholds of ITS1 PCR for the fungal DNAs could be different primer binding capacities and possible secondary structures of single primers. The existence of two to three oligonucleotides per fungal organism (2) resulted in species-specific hybridization patterns.

As a proof of principle and for clinical validation, hybridization of DNA of blood, BAL, and tissue samples from a total of 46 patients demonstrated the applicability and feasibility of our diagnostic assay (Table 5).

For 22 patients, negative microarray data were confirmed by negative conventional diagnostic results. On the other hand, positive microarray data were confirmed by other diagnostic results for 11 patients with proven (n = 4), probable (n = 2), or possible (n = 5) IFI.

For 11 of 11 patients with possible IFI, the group of patients with the highest diagnostic uncertainty, the microarray results provided additional information about the pathogens, which were mainly Candida species. Among the clinical samples of this patient group were five BAL samples positive for Candida species; from a clinical point of view, the detection of Candida species in BAL is not accepted as proof for IFI. The rate of sample contamination with Candida species is still unclear.

For one patient with proven invasive aspergillosis (determined from a brain abscess sample), DNA microarray results from blood samples were negative and possibly caused by intensive antifungal treatment (18); for one patient with probable invasive aspergillosis, DNA microarray results from a BAL aliquot were also negative under antifungal treatment and/or due to a nonrepresentative sample.

Genomic DNA and specific hybridization patterns of Candida albicans, Candida dubliniensis, Candida glabrata, Aspergillus fumigatus, Aspergillus flavus, Fusarium solani, Rhizopus microsporus, Scedosporium prolificans, and Trichosporon asahii were detected. One fungal pathogen was identified in 11 patients with IFI, and two or more fungal pathogens were likewise identified in 11 patients with IFI. The microarray findings of these patients show the potential of the assay to detect more than one fungal organism as a cause of infection in the same patient. From a clinical point of view, our knowledge about fungal infections caused by multiple pathogens is small.

Comparing results of the microarray analysis to the Aspergillus-specific nested PCR assay, positive Aspergillus PCR findings for five patients were in line with positive DNA microarray results for Aspergillus DNA. For one patient, the Aspergillus nested PCR assay was positive, whereas the DNA microarray results were negative. For three patients, Aspergillus PCR results from blood and tissue samples were positive, whereas the DNA microarray results were negative for Aspergillus DNA but positive for additional fungal species (C. albicans, C. dubliniensis, and C. glabrata). Differences between nested Aspergillus PCR (29) findings and the microarray results are caused by the higher sensitivity of the nested PCR assay for Aspergillus DNA; generally, nested PCR assays are more sensitive than multiplex PCRs because of an additional partial amplification of the PCR product in the second step.

Three other groups previously described the application of the DNA microarray technology for the detection of fungal pathogens from cultured clinical isolates. In the study of Hsiao et al. (16), the capture probes were designed against the ITS1 and ITS2 regions of the fungal rRNA genes of filamentous fungi but not of yeasts. The group determined the detection threshold of the assay by using two serially diluted fungal DNA samples for all pathogens. In contrast, our assay defined different species-specific detection thresholds of both PCR and microarray hybridization for all 14 fungal species. The second study describing the detection of fungal pathogens from cultured clinical isolates was published by Leinberger et al. (20). In this study, capture probes complementary to the ITS1 and ITS2 regions of the fungal rRNA genes were spotted on epoxy-coated glass slides. This DNA microarray contained capture probes from 12 Candida and Aspergillus species, but not from other clinically relevant fungal pathogens. Although universal fungal primers exist (31), we decided to design primer pairs amplifying the ITS1 regions only, because of the good binding capacity of shorter DNA fragments of up to 300 bases during hybridization, leading to a high sensitivity of the assay. In the study by Leinberger et al., the question of PCR and hybridization detection thresholds was not addressed. In contrast, we focused our interest on a high sensitivity for our assay because we were investigating noncultured clinical samples containing only small amounts of fungal DNA. Huang et al. (17) described another application of the DNA microarray technology that uses capture probes from the fungal ITS2 regions of the rRNA genes to identify pathogenic fungi from standard strains and cultured clinical isolates. The detection of the fungal DNA was carried out using isolates and additionally spiked blood samples of 16 patients by negative culture, microscopy, or PCR. Their DNA microarray contains capture probes of 20 fungal species. Other pathogens are less relevant than Candida, Aspergillus, and Mucor species for the clinical investigation of immunocompromised patients with hematologic malignancies.

In addition to those of DNA chip technology tests, the results of a small number of multifungal PCR-based assays for the detection and identification of fungal pathogens in clinical samples have been published (8-10, 14, 19, 21, 27). These assays cover a different spectrum of targeted fungal organisms using diverse methodical detection procedures after an initial PCR step, in order to answer distinct clinical questions. However, a molecular diagnosis gold standard of IFI has not yet been defined.

The established multifungal DNA microarray detects clinically relevant fungal pathogens specifically and at low detection thresholds from noncultured clinical samples, detecting different fungal pathogens with one test. In view of the changing spectrum of clinically relevant and emerging fungal pathogens causing IFI, this new diagnostic approach meets urgent clinical needs, at least concerning the high-risk group of patients with hematologic malignancies. This evaluation by a prospective multicenter study, testing blood, BAL, and tissue samples from immunocompromised patients, especially patients with acute leukemia, and, comparing microarray results with findings from conventional diagnostic studies, is ongoing.