Evaluation of Fluorescent Capillary Electrophoresis for Rapid Identification of Candida Fungal Infections

Abstract

Early diagnosis of fungal infection is critical for initiating antifungal therapy and reducing the high mortality rate in immunocompromised patients. In this study, we focused on rapid and sensitive identification of clinically important Candida species, utilizing the variability in the length of the ITS2 rRNA gene and fluorescent capillary electrophoresis (f-ITS2-PCR-CE). The method was developed and optimized on 29 various Candida reference strains from which 26 Candida species were clearly identified, while Candida guilliermondii, C. fermentati, and C. carpophila, which are closely related, could not be distinguished. The method was subsequently validated on 143 blinded monofungal clinical isolates (comprising 26 species) and was able to identify 88% of species unambiguously. This indicated a higher resolution power than the classical phenotypic approach which correctly identified 73%. Finally, the culture-independent potential of this technique was addressed by the analysis of 55 retrospective DNA samples extracted directly from clinical material. The method showed 100% sensitivity and specificity compared to those of the combined results of cultivation and panfungal PCR followed by sequencing used as a gold standard. In conclusion, this newly developed f-ITS2-PCR-CE analytical approach was shown to be a fast, sensitive, and highly reproducible tool for both culture-dependent and culture-independent identification of clinically important Candida strains, including species of the “psilosis” complex.

INTRODUCTION

During the last several decades, the impact and frequency of fungal infections have gained importance mainly due to an increasing number of immunocompromised patients (1, 2). Fungemia cases are being caused mainly by Candida species, which are the fourth most common microorganisms isolated from the blood samples (3, 4, 5). Sepsis due to Candida spp. is a very serious condition and has a higher mortality rate that for bacterial pathogens; reaching 54 to 64% in Candida-associated septic shock (6, 7). Moreover, the current changes in the epidemiology of invasive mycoses has highlighted a shift in the Candida species involved with a reduced proportion of Candida albicans and an increase in non-albicans species, which can show different susceptibility to various antifungal therapies (8, 9, 10).

Early initiation of antifungal therapy is a critical step in the treatment of fungal infections. Therefore, quick, successful detection and identification of the etiological agents is crucial for early targeted therapy and favorable clinical patient outcome (11). Correct species identification is mostly based on phenotypic features and is usually time-consuming because a typical diagnostic workflow takes up to several days. Moreover, the phenotypic methods may lead to misidentification, particularly in the case of closely related species (12, 13). A significant improvement occurred when matrix-assisted laser desorption ionization−time of flight mass spectroscopy (MALDI-TOF MS) was introduced as a routine laboratory procedure, enabling rapid fungi identification. The only obvious limitation of this approach remains its culture dependence and the size and quality of the library used (14). On the other hand, culture-independent analysis based on molecular biological methods offers additional techniques that are alternatives to culture. Although PCR-based approaches for candidiasis diagnosis are currently not recommended in the clinical setting, mainly because of a lack of standardization (15, 16), several promising PCR-based methods have been developed and have shown good performance in detecting Candida infections in a clinical setting (10, 16, 17, 18, 19). A recently published meta-analysis of almost 1,000 specimens of invasive candidiasis indicated even higher sensitivity and specificity of PCR-based techniques than with cultivation-based approaches (10).

The PCR-based techniques developed so far include approaches based on real-time PCR targeting of specific fungal pathogens by using species-specific probes or primers (20, 21, 22). Another approach is based on PCR with panfungal- or genus-specific primers targeting conserved rRNA regions, followed mostly by sequencing (23) or also by other techniques like restriction analysis (restriction fragment length polymorphism [RFLP]) (24), high-resolution melting (HRM) analysis (25, 26, 27), microarray-based detection (28, 29), pyrosequencing (30), or determination of amplicon size by using capillary electrophoresis (31, 32, 33). Capillary electrophoresis has been proven to be superior to classical electrophoresis for separating DNA due to its higher efficiency, higher speed, better sensitivity, and better suitability for automation (34).

This study was focused on developing a new PCR-based approach for fluorescently labeled ITS2 rRNA amplicons detected by capillary electrophoresis (f-ITS2-PCR-CE), allowing fast identification of the most clinically relevant Candida species, with good data reproducibility and high resolution power. Moreover, we also determined the detection limit of this method and addressed the possibility of detection and identification of more than one species in a sample, as well as direct fungi identification in clinical material.

MATERIALS AND METHODS

Fungal strains and clinical specimens.

Twenty-nine reference Candida strains from the Belgian Coordinated Collections of Microorganisms (BCCM/MUCL) (Louvain-la-Neuve, Belgium), the Czech Collection of Microorganisms (CCM) (Brno, Czech Republic), and the Culture Collection of Yeasts (CCY) (Bratislava, Slovakia) were analyzed. All reference strains are indicated in Table 1.

TABLE 1.

Length of the ITS2 amplicons of reference strains and clinical isolates determined by sequencing and capillary electrophoresis

Species	Reference strain				Clinical isolate f-ITS2-PCR-CE range of length (nt)
	Name	ITS2 length by sequencing (nt)	f-ITS2-PCR-CE range of length (nt)a
			Non-Ab	Plus-Ab	Non-A	Plus-A
C. lipolytica	CCY 29-26-42	229	228.89	230.00	229.01	230.07
C. haemulonii	MUCL 27866	238	238.43	239.49	238.38	239.45
C. pulcherrima	CCY 29-2-128	243	244.13	245.22	244.39	245.00
C. lusitaniae	CCY 29-59-1	244	243.41	244.49	243.60	244.64
C. intermedia	CCY 29-12-10	247	247.99	249.11	248.26	249.28
C. catenulata	CCY 29-17-3	259	258.07	259.01	258.10	259.01
C. pararugosa	MUCL 29869	260	261.59	262.56	261.64	262.52
C. rugosa	CCY 29-15-1	263	259.52	260.46	NDc	ND
C. inconspicua	CCY 26-26-11	295	292.22	293.14	292.26	293.17
C. lambica	CCY 29-97-12	295	288.94	289.80	289.18	290.13
C. parapsilosis	CCM 8260	300	297.46	298.40	297.46	298.40
C. orthopsilosis	MUCL 49939	300	297.80	298.73	297.78	298.71
C. metapsilosis	MUCL 46179	304	300.85	301.92	300.78	301.91
C. norvegensis	CCY 29-47-2	314	310.05	311.14	312.25	313.32
C. tropicalis	CCM 8264	317	314.74	315.73	314.87	315.85
C. albicans	CCM 8320	326	326.64	327.67	326.75	327.77
C. dubliniensis	CCY 29-177-1	331	331.52	332.55	331.60	332.70
C. krusei	CCY 29-9-17	336	329.85	330.87	329.68	330.64
C. utilis	CCY 29-38-74	352	352.04	352.96	352.11	352.99
C. fabianii	CCY 38-20-1	361	359.82	360.76	359.75	360.70
C. zeylanoides	MUCL 27735	363	362.52	363.50	ND	ND
C. pelliculosa	CCY 29-6-7	364	362.91	363.95	362.88	363.87
C. guilliermondii	CCY 39-23-6	368	366.92	367.88	366.78	367.67
C. fermentati	MUCL 49145	368	366.92	367.87	ND	ND
C. carpophila	MUCL 49144	368	366.94	367.92	ND	ND
C. famata	MUCL 30003	370	369.02	370.00	369.16	370.04
C. glabrata	CCM 8270	408	405.38	406.39	405.14	406.18
C. robustad	CCM 8191	409	408.37	409.41	408.41	409.45
C. kefyr	MUCL 29857	422	418.66	419.72	418.60	419.80

^aReference strains were analyzed five times. The SD was not >0.1 nt for nay reference strain.

^bNon A/Plus A, double peaks which were connected with incomplete/complete 3′ A nucleotide addition during amplification.

^cND, not determined.

^dAnamorph of Saccharomyces cerevisiae.

In addition, 143 clinical Candida strains from the collection at the Department of Microbiology, St. Anne's University Hospital Brno, Czech Republic, which were cultured from clinical samples over the last 5 years, and an additional 55 retrospective DNA samples, which were isolated directly from clinical material in the course of routine molecular microbiological analyses in the Molecular Genetics Laboratory, Centre for Cardiovascular Surgery and Transplantation Brno, Czech Republic, over the last 3 years, were included in the clinical validation of the f-ITS2-PCR-CE method.

All Candida isolates cultured from various clinical specimens were identified using a conventional biochemical test, CANDIDAtest 21 (Erba-Lachema, Czech Republic), in all cases and the API/ID32C method (Bio-Mérieux, France) in cases where the results were ambiguous. Candida dubliniensis was detected using an additional Bichro-Dubli Fumouze latex agglutination test (Fumouze Diagnostics, France). The identities of all reference strains and clinical isolates were confirmed using panfungal PCR followed by ITS2 sequencing as described previously (27). The same panfungal PCR followed by sequencing was used to analyze the DNA extracted directly from the clinical material.

Sample processing before DNA extraction. (i) Cultured reference and clinical strains.

Single cultures were resuspended in 100 μl of sterile injection water (B. Braun Medical, Inc., Germany).

(ii) Clinical samples.

A 500-μl blood sample was incubated at −80°C for 10 min, 1 ml of injection water was added, and the solution was mixed and centrifuged at 14,000 rpm for 10 min. The supernatant was removed, and the sediment was used further. Two-milliliter samples cerebrospinal fluid, urine, or bronchotracheal secretions were centrifuged at 23,000 rpm for 20 min at 4°C, the supernatant was removed, and the sediment was used further. Small pieces of tissue specimens (up to 5 by 5 mm) were mechanically disrupted under sterile conditions and used directly for DNA extraction.

DNA extraction.

The genomic DNA of single-reference clinical strains and specimens was isolated by a QIAamp DNA blood minikit (Qiagen, Germany), according to the manufacturer's instructions with the following modification: specimens processed as described above were incubated with lysis buffer (500 mM EDTA [pH 8.0], 1 M Tris [pH 8.0], 43.2 μl · ml⁻¹ Triton X-100), 20 μl lysozyme (180 mg · ml⁻¹; Sigma-Aldrich, USA), and 20 μl lyticase (5 U · μl⁻¹; Sigma-Aldrich) for 60 min at 37°C.

PCR amplification by using a fluorescently labeled primer (f-ITS2-PCR).

The internal transcribed spacer region ITS2 of fungal genomic rRNA was amplified by using the previously described panfungal primers UNF1 (5′-GCATCGATGAAGAACGTAGC-3′) and UNF2 (5′-AACTATACGAATTCAAGTCGCC-3′) (35). Due to the final amplicon detection with capillary electrophoresis, the UNF1 primer was 5′ fluorescently labeled with 6-carboxyfluorescein (6-FAM). Amplification was performed in a total volume of 20 μl of HotStarTaq master mix (Qiagen, Germany) with 0.5 μM each primer and the addition of Mg²⁺ ions to 2 mM Mg final concentration. The PCR amplification conditions were as follows: initial denaturation step at 96°C for 10 min, 35 cycles of denaturation step at 98°C for 10 s, annealing at 56°C for 10 s, extension at 72°C for 30 s, followed by a final extension step at 72°C for 4 min. Amplicon detection was performed using capillary electrophoresis. Negative controls were included using sterile DNA-free water (Qiagen, Germany) as a template.

f-ITS2-PCR-CE.

FAM-labeled ITS2-PCR amplicons were separated and detected in terms of their size using capillary electrophoresis on an ABI Prism 3130 Avant genetic analyzer (Life Technologies, USA). The instrument was adjusted according to the manufacturer's instructions.

One microliter PCR amplicon was mixed with 0.25 μl GeneScan 500 ROX dye size standard (Life Technologies) and 9.75 μl deionized Hi-Di formamide (Life Technologies). After denaturation at 95°C for 2 min and cooling, the samples were injected into a 36-cm capillary column containing the high-performance polymer POP-7 (Life Technologies). The electrophoretic parameters were set as follows: injection time, 16 s; injection voltage, 1.2-kV; electrophoresis voltage, 1.2-kV; and oven temperature, 60°C. Finally, PCR product lengths were analyzed using GeneMapper v.4.1 software. Successful analysis was derived from the red peaks being assigned to the ROX-labeled size standard 500 (Life Technologies).

Determination of assay sensitivity.

A detection limit of f-ITS2-PCR-CE was determined for C. albicans CCM 8320 and C. tropicalis CCM 8264 by serial dilutions of genomic fungal DNA and cell concentrations ranging from 100 ng to 1 fg and 10³ to 1 CFU · ml⁻¹, respectively, in triplicate experiments. The accurate cell concentration was determined by the assessment of CFU on Sabouraud agar plates.

RESULTS

Reference strains: ITS2 length library preparation.

Twenty-nine reference strains belonging to the Candida genus were analyzed using optimized capillary electrophoresis with fluorescent detection in multiple runs (n = 5), resulting in an internal laboratory database of the ITS2 lengths of the reference strains (Table 1). The lengths of the ITS2 amplicons of the Candida strains ranged from 229 nucleotides (nt) (C. lipolytica) to 420 nt (C. kefyr). Double peaks (non-A and plus-A peaks) appeared variably, which was also confirmed by sequencing (data not shown). The run-to-run variations resulted in a standard deviation up to 0.1 nt for both non-A and plus-A peaks, which indicated the excellent reproducibility of this method.

In total, 26 out of the 29 Candida reference strains analyzed were clearly distinguished by scoring the ITS2 rRNA region length. An ambiguous species determination was observed in the case of C. guilliermondii versus C. fermentati versus C. carpophila due to the same ITS2 amplicon length (Table 1). Of note, C. orthopsilosis and C. parapsilosis, having the same length of 300 nt according to sequencing, showed ITS2 length values constantly differing by 0.4 nt. Similarly, C. zeylanoides and C. pelliculosa, having lengths of 363 nt and 364 nt, respectively, determined by sequencing, showed a constant difference of 0.4 nt using f-ITS2-PCR-CE.

Sensitivity of the assay.

The reproducible detection limit of f-ITS2-PCR-CE, tested on C. albicans and C. tropicalis reference strains, was set to 5 fg DNA and 10 CFU · ml⁻¹ (Fig. 1).

FIG 1.

Illustration of the sensitivity assessment of the fragment length determined by the capillary electrophoresis. Serial dilutions were performed for genomic DNA isolated from C. albicans CCM 8320.

Clinical strain analysis.

In total, 143 clinical fungal strains, identified by cultivation-based approaches and also evaluated by high-resolution melting (HRM) analysis in our previous study (27), were analyzed by f-ITS2-PCR-CE in a blinded manner to validate an optimized technique. All of the clinical isolates analyzed were grouped according to the ITS2 length determined by the f-ITS2-PCR-CE approach and evaluated by comparison with the prepared ITS2 reference strain library. The identification was considered successful if the length of the analyzed sample appeared within the interval of ±0.3 nt of the respective reference strain's length. Because of the presence of non-A, plus-A, or both variants of ITS2 amplicon length during capillary electrophoresis, the clinical isolates were evaluated by comparison with the reference strain ITS2 library with respect to the occurrence of these plus-A/non-A peaks, accepting the interval of ±0.3 nt for both non-A and plus-A peaks. Despite the plus-A peaks appearing randomly with run-to-run difference, the signal intensity of the dominant non-A peak was always at least 5 times higher than that of the minor one.

All 143 monofungal clinical samples were clearly distinguished into 24 groups, and each group corresponded to one specific Candida species (Table 2). In total, 126 clinical strains (88%) were identified correctly. All C. africana strains (n = 7; 5%) were identified as C. albicans, and, furthermore, unlike phenotypic identification, six C. guilliermondii isolates (4%) were evaluated as C. guilliermondii/fermentati/carpophila. The remaining four monofungal samples (3%), identified both phenotypically and by sequencing as C. norvegensis, could not be classified because they differed by 2 bp in length compared to the reference strain, even though they all had the same unique ITS2 length. Apart from C. norvegensis, we did not observe any significant differences in the ITS2 amplicon lengths of the clinical strains compared with those of the reference strains. A small but constant divergence in fragment length (0.4 nt) was observed also between the C. parapsilosis and C. orthopsilosis clinical strains (20 samples in total, 14%).

TABLE 2.

Comparison of the phenotypic approach, HRMA, and newly employed f-ITS2-PCR-CE in Candida species identification

Species	No. of isolates	Phenotypic methods			HRM analysisa			f-ITS2-CE
		Correct identification	Group ranging	Misidentification/no identification	Correct identification	Group ranging	Misidentification/no identification	Correct identification	Group ranging	Misidentification/no identification
C. albicans	13	13			13			13
C. catenulata	1	1			1			1
C. dubliniensis	5	5			5			5
C. fabianii	20			20b		20c		20
C. glabrata	10	10			10			10
C. guilliermondii	6	6				6c			6d
C. inconspicua	1	1			1			1
C. intermedia	1	1			1			1
C. kefyr	6	6			6			6
C. krusei	7	7			7			7
C. lambica	3	3					3	3
C. lusitaniae	5	5			5			5
C. metapsilosis	9			9e		2c	7	9
C. orthopsilosis	5			5e		5c		5f
C. parapsilosis	15	15			15			15f
C. norvegensis	4	4			4					4g
C. pararugosa	3			3h	3			3
C. pelliculosa	3		3b		3			3
C. pulcherrima	1	1			1			1
C. tropicalis	6	6			6			6
C. utilis	1		1b		1			1
C. famata	4	4			4			4
C. lipolytica	3	3			3			3
C. robusta	3	3			NDi			3
C. africana	7		7j		ND				7j
C. haemulonii	1	1			ND			1
Total no. (%) of isolates	143/132a	95 (66)	11 (8)	37 (26)	89 (67)	33 (25)	10 (8)	126 (88)	13 (9)	4 (3)
Total no. (%) of species	26/23a	19 (73)	3 (12)	4 (15)	18 (78)	4 (17)	2 (9)	23 (88)	2 (8)	1 (4)

^aIn the case of HRM analysis, a lower number of clinically important Candida strains were tested in comparison with phenotypic and f-ITS2-PCR-CE methods. HRM analysis was used in our previous study (27).

^bIdentified as C. utilis/C. pelliculosa (could not be distinguished from each other).

^cCould not be distinguished from each other.

^dEvaluated as C. guilliermondii/carpophila/fermentati.

^eIdentified as C. parapsilosis.

^fSmall difference (0.4 nt) in fragment length appeared constantly.

^gThe ITS2 fragment length was unique and differed from that of all particular reference strains.

^hIdentified as C. rugosa.

ⁱND, not determined.

^jEvaluated as C. albicans. C. africana strains were identified by specific PCR.

Polyfungal sample analysis.

We also evaluated the ability of our approach to distinguish more than one fungal species in a specimen. Thus, we prepared mixtures of the particular Candida reference strains paired with C. albicans and an equal quality of all pairs of strains with a similar fragment length (for details, see Fig. 2). All strains tested in the artificially prepared polymicrobial samples were distinguished if they differed by at least 1 nt in length. Interestingly, C. inconspicua and C. lambica, which have no differences in ITS2 length determined by sequencing (295 nt) were distinguishable by capillary electrophoresis, either running separately or in the mixture (Fig. 2). Moreover, we did not observe any change in fragment length when more than one species was present in a sample (data not shown). Hence, C. parapsilosis versus C. orthopsilosis or C. pelliculosa versus C. zeylanoides present in one sample was not distinguished.

FIG 2.

Examples of final electropherograms for more than one species identified in the sample. (Top) C. krusei and C. dubliniensis; (center) C. lambica and C. inconspicua (for both species, 295 nt after sequencing); (bottom) C. glabrata and C. robusta (408 nt and 409 nt, respectively, after sequencing).

Clinical sample culture-independent analysis.

In order to test f-ITS2-PCR-CE in a culture-independent arrangement, we analyzed 55 retrospective DNA samples from various clinical materials (Table 3). The samples were treated in the same way as the DNA from clinical isolates, and the results were compared with those obtained using both a phenotypic approach and panfungal PCR targeted on the ITS2 region followed by sequencing in the routine setting. Twenty-seven samples were negative and the remaining 28 were positive using the combined results from culturing and panfungal PCR followed by sequencing as a gold standard. We observed 100% sensitivity and specificity of the optimized f-ITS2-PCR-CE approach (Table 3).

TABLE 3.

Culture-independent clinical sample analysis^a

^aShading of columns indicates concordance.

^b CSF, cerebrospinal fluid; ET, endotracheal tube.

^c NA, data not available.

^d C. robusta was evaluated as a probable contamination.

^e The sensitivity and specificity of the f-ITS2-PCR-CE approach compared with the combined results for cultivation and panfungal PCR plus sequencing.

DISCUSSION

The mortality associated with invasive Candida infections remains high, and early initiation of targeted antifungal therapy has been shown to be of major value. As various medically important Candida spp. differ in their susceptibility to therapy, early, fast, and accurate identification of particular Candida species is needed (10, 15). Therefore, we employed and evaluated the efficiency of fluorescent fragment length analysis of the variable ITS2 region from the multicopy rRNA, followed by detection using capillary electrophoresis. Thus, the method was optimized on 29 reference Candida species and validated on 143 clinical Candida strains. We assessed the efficiency of the approach by comparing the technique developed with phenotypic identification as well as HRM analysis used in our previous study (27). Finally, the culture-independent potential of this approach was tested using 55 retrospective DNA samples extracted directly from clinical material.

Despite an excellent correlation (R² = 0.999) between the ITS2 amplicon length determined by capillary electrophoresis with the ITS2 length established by sequencing (see Fig. S1 in the supplemental material), we observed deviations in fragment lengths between capillary electrophoresis-determined PCR fragments and ITS2 amplicons after sequence analysis (Table 1) which is in agreement with Landlinger et al. (33) and Chen et al. (36). This can be attributed to the specific properties of capillary electrophoresis and migration shifts during electrophoresis caused by the fluorescent dye component (33). Moreover, in reference and cultured clinical strain analysis, we observed an artificial additional peak of lower size and intensity that we assigned as a plus-A or non-A peak connected with complete or incomplete 3′ A nucleotide addition during amplification as was reported by Smith et al. (37). However, the additional peak always had a lower signal intensity and did not complicate the final identification in any sample. Furthermore, no intraspecies variabilities were observed among the species in which more than one clinical strain was tested.

The analysis of reference species indicated the very good resolution power our technique; all reference species except the very closely related C. guilliermondii versus C. fermentati versus C. carpophila (38) were determined unambiguously with excellent data reproducibility. Furthermore, in clinical Candida species isolates, 88% of the strains were evaluated correctly by blinded analysis. An ambiguous resolution was observed only in three cases: (i) C. guilliermondii, because closely related and indistinguishable C. fermentati and C. carpophila were included in the reference library; (ii) C. africana, closely related to C. albicans (39) (and even not recognized as an individual species by some authors [40]); and (iii) C. norvegensis, differing by 2 nt compared to the reference strain. Thus, the clinical strains of C. norvegensis could not be classified, although they showed a constant and unique pattern. Interestingly, two tested C. norvegensis strains identified by capillary electrophoresis in a previous study (33) also differed by 2 nt in length. Based on our results, we added the C. norvegensis strain from the clinical material as a reference strain to our library. On the other hand, our approach was able to distinguish all tested strains of C. parapsilosis, C. metapsilosis, and C. orthopsilosis. As Horn et al. (41) stated, accurate species identification of the “psilosis” complex is becoming crucial in clinical management and the rapid detection of these species is now clinically relevant because of variations in antifungal susceptibility (42, 43). In our hands, C. metapsilosis was easily distinguishable among the species of the psilosis complex and, interestingly, C. parapsilosis and C. orthopsilosis also showed no overlaps and were clearly identified, despite the fact that the lengths of the sequenced ITS2 regions were the same for both species. By using f-ITS2-PCR-CE, the lengths of the C. parapsilosis and C. orthopsilosis ITS2 lengths constantly differed by 0.4 nt in all reference and clinical strains tested. Small differences might be caused by the primary ITS2 sequence; alignment of the ITS2 sequences showed one base mismatch: C (in C. parapsilosis) versus T (in C. orthopsilosis) (nucleotide position at 142 bp compared with that of the reference strain; data not shown). Since we observed deviations not higher than 0.1 bp within one species in all the sets of samples tested, we presume that this small nucleotide difference in fragment length is significant for the resolution of these two closely related species using the capillary electrophoresis approach. It is likely that separation by capillary electrophoresis is also influenced by the nucleotide composition of the amplified region, which might be useful for determining closely related species such as C. parapsilosis and C. orthopsilosis.

An overall comparison of capillary electrophoresis with HRM analysis and classical phenotypic identification, which is still considered the gold standard, indicated a higher resolution power for f-ITS2-PCR-CE (Table 2). In our study, correct Candida species identification was achieved in 23/26 (88%) cases, compared to 18/23 (78%) and 19/26 (73%) cases for HRM analysis and phenotypic identification, respectively. Capillary electrophoresis offers the possibility of distinguishing all Candida species which may be incorrectly identified by routinely used phenotypic identification (C. metapsilosis, C. orthopsilosis, and C. rugosa strains). Furthermore, C. pelliculosa, C. utilis, or C. fabianii, which were also indistinctly identified by HRM analysis (27), were identified correctly by capillary electrophoresis (Table 2). Moreover, HRM analysis, which had better resolution than phenotypic identification, was not able, unlike capillary electrophoresis, to clearly distinguish between C. fabianii and C. guilliermondii and to identify some other species like C. lambica or C. metapsilosis correctly (Table 2).

Furthermore, the pilot study of 55 clinical samples permitted unambiguous species identification of all positive specimens while generating no false-positive results compared to standard cultivation and sequencing followed by panfungal PCR (Table 3). It confirmed both the excellent sensitivity and specificity of this approach and the applicability in culture-independent identification of Candida spp. in various clinical materials. Although three samples tested positive using molecular methods, they were culture negative. We consider the possibility of false-positive results very unlikely in these cases for the following reasons: (i) we maintained a standard diagnostic workflow by performing sequential sample manipulation procedures (DNA extraction, PCR preparation, and post-PCR operations) in separate rooms and in running both extraction from sterile water and PCR with sterile water instead of a DNA template as negative controls in each set of samples; (ii) there was a concordance between both molecular methods (f-ITS2-PCR-CE and panfungal PCR followed by sequencing) performed independently at different times; and (iii) the amplification signal was very strong in all three samples, while the controls remained negative using both molecular methods.

Thus, the f-ITS2-PCR-CE approach represents the possibility of rapid identification of clinically important Candida species without the need for cultivation, which is necessary for phenotypic identification as well as identification by MALDI-TOF MS, considered the most rapid and reliable tool for the identification of fungi (44). Moreover, unlike most current molecular diagnostic tools, f-ITS2-PCR-CE showed excellent species discrimination power in polyfungal samples, even in the case of very similar ITS2 fragment lengths. Interestingly, all strains were identified correctly in the clinical polyfungal samples despite their presumptive different quantities. Our preliminary experiments showed that strains in mixtures can be identified up to the ratio of 1:100 (data not shown). Mixed infections represent a diagnostic challenge (45), and since simultaneous infections with multiple fungi may be misinterpreted as monomicrobial infections by many other molecular diagnostic tools, f-ITS2-PCR-CE represents a very useful, inexpensive, and fast alternative to sequencing followed by analysis with the online tool RipSeq (Pathogenomix, Inc., Santa Cruz, CA, USA).

Capillary electrophoresis has been previously described in some studies (31, 32, 33, 36) as a promising method for clinically important fungi detection, identification, and differentiation of closely related species. Turenne et al. (31) first reported the application of this approach for a large number of commercially available strains and approximately 26 clinical strains, including 12 different Candida species. Landlinger et al. (33) applied the technique directly on clinical specimens and used seminested PCR, which is more sensitive but is also more time-consuming and has a higher risk of contamination. Although Chen et al. (36) identified a large number of clinically important strains; only 16 different Candida species were included in their study. Moreover, these authors did not report the ability of this approach to distinguish the three species of the psilosis complex and did not compare it with the performance of phenotypic identification or other techniques (like HRM analysis). Finally, in addition to the above-mentioned publications, the ability of capillary electrophoresis to distinguish species from polyfungal samples was addressed in this study.

Thus, developing rapid and accurate amplification-based ITS2 assays to identify different Candida species might have an impact on care and improve the outcome for affected patients. The newly developed and optimized f-ITS2-PCR-CE analytical approach has proved to be fast, taking 4 h (including DNA isolation). This is a difference of more than 3 h compared with the time for the currently used panfungal PCR followed by sequencing, which is a sensitive and efficient tool for culture-independent identification of a substantial number of clinically important Candida species. Although cultivation-based techniques are considered the gold standard, molecular biological approaches such as f-ITS2-PCR-CE can supplement these culture techniques and can also supplement panfungal PCR followed by sequencing to enable rapid, reproducible, and sensitive detection and identification of a large number of clinically relevant Candida species.