New Strategy for Rapid Diagnosis and Characterization of Fungal Infections: The Example of Corneal Scrapings

Purpose The prognosis of people infected with Fungi especially immunocompromised depends on rapid and accurate diagnosis to capitalize on time administration of specific treatments. However, cultures produce false negative results and nucleic-acid amplification techniques require complex post-amplification procedures to differentiate relevant fungal types. The objective of this work was to develop a new diagnostic strategy based on real-time polymerase-chain reaction high-resolution melting analysis (PCR-HRM) that a) detects yeasts and filamentous Fungi, b) differentiates yeasts from filamentous Fungi, and c) discriminates among relevant species of yeasts. Methods PCR-HRM detection limits and specificity were assessed with a) isolated strains; b) human blood samples experimentally infected with Fungi; c) blood experimentally infected with other infectious agents; d) corneal scrapings from patients with suspected fungal keratitis (culture positive and negative) and e) scrapings from patients with suspected bacterial, viral or Acanthamoeba infections. The DNAs were extracted and mixed with primers diluted in the MeltDoctor® HRM Master Mix in 2 tubes, the first for yeasts, containing the forward primer CandUn (5'CATGCCTGTTTGAGCGTC) and the reverse primer FungUn (5'TCCTCCGCTT ATTGATATGCT) and the second for filamentous Fungi, containing the forward primer FilamUn (5'TGCCTGTCCGAGCGTCAT) and FungUn. Molecular probes were not necessary. The yields of DNA extraction and the PCR inhibitors were systematically monitored. Results PCR-HRM detected 0.1 Colony Forming Units (CFU)/µl of yeasts and filamentous Fungi, differentiated filamentous Fungi from yeasts and discriminated among relevant species of yeasts. PCR-HRM performances were higher than haemoculture and sensitivity and specificity was 100% for culture positive samples, detecting and characterizing Fungi in 7 out 10 culture negative suspected fungal keratitis. Conclusions PCR-HRM appears as a new, sensitive, specific and inexpensive test that detects Fungi and differentiates filamentous Fungi from yeasts. It allows direct fungal detection from clinical samples and experimentally infected blood in less than 2.30 h after DNA extraction.


Introduction
The frequency of fungal infections has been increasing for the last 30 years due to viral or iatrogenic immunodeficiencies, the efficiency in treating bacterial infections, the development of indwelling devices, and the massive use of contact lenses. [1][2][3][4][5][6][7][8] The incidence of fungal keratitis (keratomycosis) is also on the rise, and filamentous Fungi are the most frequently reported pathogens. [3,8] From yeasts, Candida albicans is the most frequently associated with disease. However, C. glabrata, C. tropicalis, C. krusei, and C. parapsilosis have gained greater significance. [1,2,4,8].
Fungal infection management requires timely diagnosis for rapid onset of treatments, but approximately one half of the samples remain culture negative and/or negative by fungal antigen detection using immunosorbent assays (ELISA). [9][10] Improved detection performances were reported by amplifying fungal genomic regions (polymerase chain reactions, PCRs). [11,12] However, the classic PCRs do not differentiate filamentous Fungi from yeasts and require post amplification procedures (restriction enzyme digestion and analysis; single-base extension; hybridization probes or molecular sequencing). [11][12][13][14][15] The ''gold standard'' for fungal characterization is DNA sequencing, but this method is laborious, expensive and cannot be performed routinely for daily diagnosis. [15].
The real-time Taqman PCR using fluorogenic labelled Taqman-probes facilitates the detection and partial characterization of Fungi but requires a series of expensive labelled probes (each probe detects a single fungal type or one species per reaction). [12,[16][17].
Because the first-line therapy is different for filamentous Fungi and yeasts as well as for different yeasts, rapid and accurate information is required to target the treatments according to natural fungal susceptibilities. [18][19][20].
The availability of improved fluorescent DNA binding dyes with highly predictable saturation properties allows precise assessment of sequence length by High Resolution Melting real-time PCR (PCR-HRM). [21,22] Recently, a diagnosis test based on PCR-HRM technology was reported for vaginal samples, detecting and identifying 8 Candida at species level. [23] Nevertheless, it was unable to differentiate Candida from filamentous Fungi and did not detect and characterise S. cervisiae and Trichosporon.
The goal of the present work is to develop a new test able to detect in 1 run the equivalent of at least 1 fungal colony forming unit (CFU) per reaction. In addition this molecular approach should simultaneously differentiate yeasts from filamentous Fungi and discriminate among relevant species of yeasts in clinical samples and in blood experimentally infected with fungal suspensions.

Materials and Methods
Investigations were conducted according to the principles expressed in the Declaration of Helsinki (http://www. wma.net/ e/policy/) and were approved by the Institutional Review Board of the Centre Hospitalier National des Quinze-Vingts (CHNO), Ministry of Public Health, Paris-France. Written informed consent was obtained from all participants for the use of each sample. Forms with written consent were drafted according to the requirements of the CHNO Review Board and the National Health Authorities were double checked, validated and signed by the physician in charge of the sampling and sent to the laboratory. The preliminary studies were performed with characterized strains isolated from patients presenting corneal ulcers in the National Eye Hospital in Paris (CHNO des Quinze-Vingts) or from strains isolated from blood stream infections (generous gift from Dr Christophe Hennequin's laboratory, CHU Saint-Antoine, Paris, France).
One colony of each fungal species was scraped from the surface after 48 h of culture on Sabouraud's dextrose agar, suspended in Phosphate Buffer Solution (PBS) and replated. To reduce the over representation of fungal DNA from non viable organisms, one colony was scraped from the second dish 48 hours later, suspended in PBS and tenfold diluted. Each dilution was divided in several aliquots; three were plated on Sabouraud's dextrose agar to assess the number of colonies (equivalent CFU/ml) and the others kept as calibrators. For each series of experiments the PCR-HRM detection limits were validated with serial dilutions of fungal suspensions diluted in PBS and simultaneously titrated by plating.
Blood samples were collected in 10 ml citrate tubes from vein puncture of healthy subjects and transported to the laboratory within 1 h. After white cell count to assess the cell load of inoculums (white cell counts .12.000/ml were excluded) randomized aliquots were spiked with different titrated fungal suspensions. Negative controls consisted in non infected blood or leukocyte suspensions from the same individuals.
For each fungal strain, haemoculture bottles were inoculated with 10 ml of saline or blood spiked or not with Fungi and incubated up to 12  . The confirmation of fungal species was carried out by sequencing the DNA from both ends using flanking vector primers. [12,15] Bacteria were cultured and characterized with routine diagnosis tests; Acanthamoeba and Herpesviridae were detected by real-time PCR. [24,25].
Corneal scrapings from 38 patients, 13 with proven fungal culture positive, 10 with suspected fungal keratitis culture negative and 15 with non suspected fungal keratitis (bacterial, viral or Acanthamoeba) were tested masked. Sampling from patients presenting corneal ulcers and requiring microbiological diagnosis was performed by deep corneal scraping by certified ophthalmologists with sterile stainless steel blades after rinsing of fluorescein and topical anaesthetic from the eye surface. [24] Slides with aliquots of scrapings were fixed and stained (Giemsa pH: 7.4) for direct microscopic examination and the presence of Fungi was confirmed by Grocott's methenamine silver reaction. The second aliquots were cultured within 30 minutes after collection up to 30 days before discarded as culture negative and remnants of blades were frozen dry at -80uC for further molecular diagnosis. PCR-HRM was carried out after thawing and addition of 200 ml of sterile Phosphate Buffer Solution (PBS) to the tubes containing the dry blades.
The DNA extraction was carried out in a vertical safety laminar flow cabinet in a dedicated room. To monitor the extraction yields and the absence of PCR inhibitors the internal control (IC) consisting of 5 ml of a whole virus preparation of seal herpes virus (gift from G. J. van Doornum, Dept. of Virology Erasmus MC, Rotterdam, The Netherlands) was added to 200 ml of each suspension (scraping, blood, leukocytes or saline) before extraction (final concentration of 1000 to 2000 viral particles/ml). [24,25] In order to obtain spheroplasts, each specimen (sample + IC) was mixed with tris-EDTA buffer and 10 U recombinant lyticase (Sigma-Aldrich, France)/100 ml of suspension and incubated at 37uC for 60 min. After incubation, the suspensions were vortexed thoroughly and 100 ml were used for DNA extraction using the MagNA Pure compact nucleic acid isolation kit IH as described by the manufacturer in the MagNA Pure Compact automateH (Roche Diagnostics, Meylan, France) and eluted in 100 ml of elution buffer. To monitor the DNA extraction yields and the PCR inhibitors the seal herpes virus internal control (IC) was amplified in an independent real-time PCR run. [24,25] The primer sequences were respectively: 59GGGCGAATCACA-GATTGA ATC and 59GCGGTTCCAAACGTACCAA and VIC-TTTTTATGTGTCCGCCACCATCT GGATC-TAMRA for the probe. Amplification and detection of the IC was carried out in a separate tube containing 18.5 ml of the TaqManH FAST Universal PCR Mastermix (2X no AmperaseH UNG) (Applied Biosystems-France ABI Ref. 4352042), the forward and the reverse primers (0.5 uM each) with or without the fluorophorelabelled TaqManH probe (0.5 uM). This solution was mixed with 5 ml of the DNA eluted in DNA and RNA-free solution. The PCR cycling program consisted of one cycle at 95uC for 20 sec and 45 cycles at 95uC for 3 sec and 30 sec at 60uC. [24].
Kits (stable at 220uC for at least 12 weeks) for fungal detection consist of 2 tubes, the first for detection, semi quantification and identification of yeasts; the second for detection and semi quantification of filamentous   For PCR-HRM the DNA extracts (10 ml) were introduced in 2 tubes, the first containing CandUn + FungUn in the MDHRM the second FilamUn + FungUn. During PCR-HRM, the amplicons were automatically measured in a closed tube format using integrated cycler/fluorimeter ABI 7500 upgraded equipment and monitored using fluorescent DNA intercalating dyes present in the MDHRM. The PCR program started with a denaturation of 10 min at 95uC, followed by 55 cycles of amplification (15 s at 95uC, 30 s at 60uC and 30 s at 72uC). The PCR-HRM curve was obtained by denaturation at 95uC for 15 sec, cooling to 50uC for 1 min and a temperature increase until 60uC for 15 sec with a 2.2uC/s ramp rate. Samples with fluorescence of less than the 100% of the maximum were excluded from the analysis. Each run contained negative controls with no template and DNA extracts from the reactants. Linearity, sensitivity and detection limit (Equivalent CFU/ml from the Ct versus dilution curves) and reproducibility were assessed by diluting fungal suspensions in distilled water before DNA extraction. The melting temperature (Tm) at which 50% of the DNA is in the double stranded state was assessed by taking the derivative of the melting curve. The melting curves shapes depend on PCR product (amplicon) length. The DNA patterns of the derivative plot (difference plot) were used for amplicon analysis.

Results
Preliminary experiments were performed to assess the best conditions for extraction of DNA from spores: a-heat for 10 min at 94uC; b-proteinase K at 37uC for 60 min and heat at 94uC for 10 min; c-proteinase K at 37uC for 60 min, heat at 94uC for 10 min and extraction with the MagNA Pure compact nucleic acid isolation kit IH as described by the manufacturer in the Table 1. High-resolution melting analysis (PCR-HRM) detection limits (fungal spore suspensions titrated by plating) and discrimination among fungal species using the primers CandUn + FungUn and FilamUn + FungUn. MagNA Pure CompactH automate (Roche Diagnostics, Meylan, France); d-shaked in presence of beads and extraction with MagNA Pure; e-shaked in presence of beads with or without proteinase K at 37uC for 60 min, heat at 94uC for 10 min and extraction with Magna Pure; f-shaked in presence of beads with or without lyticase at 37uC for 60 min and heat at 94uC for 10 min; g-shaked in presence of beads with lyticase at 37uC for 60 min, heat at 94uC for 10 min and extracted with Magna Pure; or hlyticase at 37uC for 60 min, heat at 94uC for 10 min and extraction with Magna Pure. The highest fungal DNA extraction rates were obtained using the procedures g or h (results not shown). The detection limits have been obtained by dilution of fresh titrated fungal suspensions. PCR-HRM with the primers CandUn + FungUn detected 0.1 CFU/ml of Candida albicans, C. krusei, C. glabrata, C. tropicalis, Saccharomyces cervisiae and Trichosporon and 1 CFU/ml of filamentous Fungi suspended in PBS. As shown in Table 1 PCR-HRM detection capacities were repeatedly higher for yeasts (10 to 100 times) using the set CandUn + FungUn, and more than 10 times higher for filamentous Fungi using the set FilamUn + FungUn (Aspergillus nidulans, A. niger, A. versicolor, A. terreus, Penicillium piccum and Fusarium solani). These results suggest that the optimization of fungal detection requires the simultaneous amplification of DNA extracts in 2 tubes, the first with the set CandUn + FungUn and the second with FilamUn + FungUn. Under these conditions, the PCR-HRM coefficient of variation for the interassay reproducibility in the complete linear range of detection [10 5 to 10 21 colony forming units (CFU)/ml] for 5 runs was less than 10%. The patterns of the first derivative (difference plot) permitted differentiation of yeasts from filamentous Fungi and the divergence between amplicon sizes of closely related species allowed PCR-HRM to easily discriminate among yeasts (Figure 1). According to the amplicon sizes bracketed by the set of primers FilamUn + FungUn (generally $194 nucleotides for Fusarium sp. versus #192 for most species of Aspergillus) the melting curve shapes were repeatedly different for Fusarium solani (Figure 2).

Set of primers
Sensitivity and specificity of PCR-HRM (while comparing with corneal scraping cultures) was of 100%. PCR-HRM allowed rapid diagnosis of keratomycosis differentiating clinical relevant species of yeasts, and produced negative results for all the samples obtained from patients with non suspected fungal keratitis and for all the negative controls [DNA extracted from 10 6 CFU/ml of Bacteria, 10 6 PFU (plaque forming units)/ml of Herpes simplex virus type 1 or 10 6 PFU/ml of Herpes simplex virus type 2, and 10 5 Acanthamoeba cysts/ml suspended in saline] ( Table 2). Human cells (10 6 human epithelial cells or fibroblasts) did not interfere with the PCR-HRM performances, confirming the in-silico specificity predictions. In patients with clinically suspected fungal keratitis (samples 1; 3; 5; 6; 8; 10; 12; 13; 17; 18; 20; 23; 24) culture was positive in 55% and images evoking Fungi were detected in 65% of the clinical samples by direct microscopic examination (Table 2). In 4 out 10 patients with clinically suspected fungal keratitis and culture negative (samples [29][30][31][32][33][34][35][36][37][38], Fungi were detected by direct microscopic examination of deep corneal scrapings (filaments in 2; budding yeasts in 1 and pseudohypha in 1) (samples 29; 31; 32 and 36). In addition, for these same 10 culture negative patients, PCR-HRM detected and characterized Fungi in 7 out of 10, including the 4 detected by direct microscopic examination. The relative high number of cornea negative cultures could be partially the result of residual eye drop preservatives carried with the samples from the eye surface. PCR-HRM was negative for all the controls carried out with the samples obtained from the air, surfaces, reactants and the blood of healthy donors (total blood or buffy coats) ( Table 2). Table 3 shows the recovery and detection time for haemoculture bottles spiked with Fungi. For fungal inoculums containing 100 CFU/bottle, the time for positivity was ranged between 16 and 36 hours of incubation for yeasts and between 36-48 hours for filamentous Fungi. Positivity was obtained after 24 to 48 hours of incubation of blood spiked with yeasts and after 24 to 72 hours with filamentous Fungi. For inoculums of 10 CFU/bottle the time for positivity was ranged between 24-48 hours for yeasts and between 48 and 72 hours for filamentous Fungi. For yeasts suspended in blood cultures were positive after 24-48 hours and after 24-72 h for filamentous Fungi. Only C. albicans, Trichosporon and Penicillium piccum could be detected for inoculums containing 1 CFU per bottle (after 96 h of culture) in saline. Inoculums containing 1 CFU per bottle or less of 7 different yeasts or 4 filamentous Fungi suspended in blood were negative. PCR-HRM detected 100% of the samples containing the equivalent of 0.1 CFU/ml of yeasts and filamentous Fungi and generated reproducible melt-curves. When challenged against profiles obtained from referenced strains run in parallel the melting profiles obtained with all the samples allowed differentiating yeasts from filamentous Fungi and discriminating among 7 different strains of yeasts. PCR-HRM performances were equivalent for Fungi suspended in saline or blood at concentrations significantly lower (10 times or more) than the detection limits of fungal cultures (Table 3).

Discussion
The automatic melting analysis of fungal sequences amplified with 2 sets of primers diluted in a mix containing a DNA intercalating dye (SYTO9) allowed rapid detection of Fungi. The differences in amplicon sizes between species were suited to fungal differentiation of yeasts from filamentous Fungi and to speciation among yeasts. By adapting the existing real-time PCR instrumentation for data acquisition it was possible to carry out reproducible diagnosis in less than 2.30 h after DNA extraction, with detection limits of at least 0.1 CFU of filamentous Fungi and yeasts per ml of sample with no need for molecular probes (radioactive, enzymatic or fluorogenic) or post amplification procedures (sequencing, amplicon restriction enzyme analysis, etc.). Corneal samples could  be readily assayed after DNA extraction without interference from other DNAs found in the specimens. Fungal culture performances depend on the type of agent, the fungal load and the mass of material that can be processed, the microbiology laboratory capacities and the presence of antiseptic or antifungal in the samples. Moreover, cultures rely on the ability of the organism to grow ex vivo and may require long incubation periods (may become positive late in the course of the infection), are time consuming and prone to contamination. [1,4,11].
In 89% of patients with suspected fungal keratitis PCR-HRM was able to detect Fungi differentiating filamentous Fungi and yeasts. Interestingly, in all patients with positive PCR-HRM the corneal infiltrates were dramatically reduced after antifungal treatments, suggesting that PCR-HRM signals represented true positive infections (results not shown). Local and topical antifungal treatments are long lasting, potentially toxic, and first-line therapies have to selectively target yeasts or filamentous Fungi. [4][5][6][27][28][29] On this matter, yeasts but not filamentous Fungi require Fluconazole and/or 5-Fluorocytosine (5FC). Aspergilli require Voriconazole and/or Caspofungin; Candida krusei are intrinsically resistant to Fluconazole and Candida glabrata and Candida tropicalis have unpredictable susceptibilities to this latter. [4][5][6][27][28][29] Hence, false negative diagnosis or lack of discrimination among major fungal types may have a direct impact on infection management (morbidity and mortality).
The yeast detection capacities of the haemoculture system used in the present study are equivalent to those reported for other automated systems (inoculums containing the equivalent of 100 and10 CFU/bottle). Nevertheless, all the blood samples containing fungal loads of 1 CFU/bottle were negative. For automated blood culture bottles inoculated with 1000 yeasts it was shown that the BacT detected growth of 90% of Candida, while Bactec 66%. [6,7] The mean time to growth detection was between 25.62 h and 27.30 h and both detected experimentally infected blood (simulated candidemia) only when additional specialized mycology media were used. [6] For filamentous Fungi the BACTEC plus Aerobic/F and the BACTEC Mycosis-IC/F automated systems followed by subculture in solid medium detected inoculums of A. fumigatus at concentrations of .3 conidia per 10 ml after 21-40 h, with longer incubation periods for A. flavus and A. terreus. [7].
For the detection and identification of Candida and Aspergillus species a variety of real-time PCR assays based on species specific probes were developed targeting the 18 S rDNA [16,30], the mitochondrial cytochrome b [31], the 18 S and 28 S rDNA [32], the CaMP65 (65-kDa mannoprotein) gene [33], the RNase P RNA gene [34], and the ITS 2 region [35]. These nucleic-acid amplification techniques are independent of microorganisms' growth and detect stressed, injured, and fastidious Fungi and those suspended in antimicrobials. [11][12][13][14] Despite this, expectations for routine direct fungal diagnosis are not fulfilled with PCRs because they a) cover either yeasts or filamentous Fungi, b) do not to detect C. lusitaniae and c) may require complex post amplification procedures and/or the synthesis and stability testing of labelled molecular probes. [11][12][13][14][15][16][17]26] Therefore, DNA sequencing was almost the only tool for molecular species assessment of PCRpositive samples. [36][37].  For diagnosis of candidemia in subjects with haematological malignancies or various forms of immunodeficiency a real-time PCR targeting the 18 S rRNA gene and requiring a series of labelled molecular probes, yielded positive results in 58.3% of blood culturepositive samples and detected in blood the genomes of Candida 3 days earlier than culture. [37] In this series 27% of whole-blood were PCR positive compared to 15% of haemoculture (92% of correlation of positives). Other studies indicate that numerous pairs of primers and labelled probes were required for each sample to identify 72% of species of positive cultures. [38] For vaginal specimens it was reported a test based on PCR-HRM technology, but it was unable to differentiate yeasts from filamentous Fungi. [22] However, the strategy developed here is different because PCR-HRM is able to detect and differentiate yeasts and filamentous Fungi in one run with detection limits of 0.1 CFU/ml or less. In addition to differentiating filamentous Fungi from yeasts, PCR-HRM discriminated among relevant clinical species of yeasts in less than 2.30 hours after DNA extraction. This only required upgrading the available real-time PCR software of the thermocyclers used in the routine microbiology laboratory. Because PCR-HRM was able to produce consistent results without the need for synthesizing and labelling molecular probes and without post amplification procedures (restriction enzymes, electrophoresis, gel analysis, hybridisation, sequencing reactants) the cost for reactants for testing one DNA extract could be reduced to less than 2 USD (2 primers and HRM mix). Compared to classic PCR (amplification followed by electrophoreses and/or hybridization and/or sequencing) this new PCR-HRM has the additional advantage of minimizing risks for false positive results due to cross contamination, because the targeted sequence amplification, the signal detection, and the DNA melting analyses are carried out in closed tubes. Moreover, PCR-HRM minimizes risks for false negative results because the yields of extraction of the DNA and the potential interference of PCR inhibitors are systematically monitored in each run and for all the samples. According to the results obtained in this study, if the future runs generate reproducible melt-curves over time in different settings, a reference database could be built to store PCR-HRM calculations and shapes of the melting profiles for each family or species to be challenged against profiles. Larger prospective multicentric trials testing different types of samples (clinical and environmental) are necessary to validate PCR-HRM usefulness as a diagnosis tool and for environmental studies.  Haemoculture bottles were inoculated with 10 ml; S: saline; B: blood; #: results from 2 independent experiments; POS: Positive; NEG: Negative after 240 hours; ##: reference strains and negative controls were extracted and tested for each run;*: melting-curve profiles consistently superimposed on those obtained with Fusarium solani. doi:10.1371/journal.pone.0037660.t003