Comparison of Restriction Fragment Length Polymorphism, Microsatellite Length Polymorphism, and Random Amplification of Polymorphic DNA Analyses for Fingerprinting Aspergillus fumigatus Isolates

Abstract

Aspergillus fumigatus fingerprints generated by random amplification of polymorphic DNA (RAPD), restriction fragment length polymorphism (RFLP) upon hybridization with repeated DNA sequences, and PCR detection of microsatellite length polymorphism (MLP) were compared among 67 isolates. In contrast to RAPD, RFLP and MLP gave discriminating and significantly concordant genotyping results.

Aspergillus fumigatus is the main species responsible for invasive aspergillosis, an often-fatal infection occurring in immunocompromised patients (27). In nature, this fungal species thrives on decaying vegetation and releases large amounts of conidia, which are dispersed by air currents and which are present in all aerial environments (10). Patient infection is acquired upon inhalation of conidia. Identification of infective strains and contaminating sources has been a major epidemiological concern in the study of nosocomial aspergillosis due to A. fumigatus (14).

Three DNA fingerprinting techniques have been extensively used for typing A. fumigatus isolates (see reference 14 for early references). The most popular technique is based on random amplified polymorphic DNA (RAPD) PCR using short single primers and low-stringency conditions (1, 2, 15, 18). The second technique is based on restriction fragment length polymorphism (RFLP) analysis upon Southern blot hybridization with inactive retrotransposon Afut1 (20). By using this moderately repetitive species-specific sequence, extremely high levels of genetic diversity within A. fumigatus have been revealed (9, 11). The third technique is based on PCR amplification of (CA)_n repeats isolated from an A. fumigatus genomic library to detect microsatellite length polymorphism (MLP) (3, 4). In view of the discrepancies previously reported between RAPD and RFLP methods (1, 17, 26) and the recent development of MLP analysis, a comparison among these three fingerprinting systems was needed to identify the most accurate genotyping method currently available for A. fumigatus.

Sixty-seven A. fumigatus clinical and environmental isolates, including three subcultured reference strains, were randomly selected from two isolate collections (Table 1). With two exceptions, clinical isolates were recovered from different individuals cared for either in a single hospital at different time periods or in distinct institutions. All environmental isolates differed by their location and time of recovery, and none was related to a patient stay. DNA was purified at the Unité des Aspergillus (Pasteur Institute, Paris, France) as previously described (9). Aliquots from single DNA batches were given an arbitrary code and sent to each of the three collaborating laboratories. To obtain the most-reliable results, each laboratory carried out its familiar fingerprinting technique. RFLP typing (performed at the Institut Pasteur) consisted of electrophoresis of EcoRI restriction fragments, Southern blot hybridization with retrotransposon-like sequence Afut1, and computerized analysis of the banding patterns visualized on the autoradiographs using GelCompar software (8, 9). A visual examination of the autoradiographs was systematically performed to validate the software data (9). Two isolates were considered different when the hybridization patterns differed by at least two bands. MLP typing (performed at Hopital Henri Mondor) consisted of PCR amplification of four A. fumigatus CA-containing microsatellite sequences using specific primer sets and precise sizing of PCR products with an automatic sequencer (3). A single size difference among the microsatellite loci studied led to a different genotype. RAPD typing was performed as previously described at the Erasmus Medical Center by using enterobacterial intergenic consensus primers ERIC-1 and ERIC-2 and decamers RAPD-1281 and RAPD-1283 in two combination assays (15, 16). The patterns were compared visually; differences in ethidium bromide staining intensities were ignored, and a single band difference led to a different overall genotype (15). For each technique, data were obtained from two experiments. Four independent investigators compared the three series of data before scoring for isolate identity.

TABLE 1.

Genotypes obtained by RFLP, MLP, and RAPD for the 67 A. fumigatus isolates studied

Isolate	Origina	Type by:
		RFLP	MLP	RAPD
IP5	Env (Hosp.1)	A	a	I
IP6	Env (Hosp.1)	A	a	I
IP8	Pat (Hosp.2)	B	b	I
IP7	Pat (Hosp.2)	B	b	II
IP24	Env (Hosp.2)	C	c	III
IP23	Env (Hosp.1)	C	c	IV
IP30	Env (Hosp.1)	D	d	V
IP31	Env (Hosp.1)	D	d	V
IP33	Env (Hosp.1)	D	Unique	Unique
IP38	Pat (The Netherlands)	E	e	I
IP35	Env (Hosp.1)	E	e	II
IP36	Env (Hosp.1)	E	e	II
IP37	Pat (Hosp.2)	E	e	Unique
IP39	Pat (Germany)	E	e	Unique
IP27	Env (Hosp.2)	E	f	VI
IP32	Env (Hosp.1)	E	Unique	II
IP42	Env (Hosp.1)	F	g	V
IP43	Env (Hosp.2)	F	g	V
IP48	Env (Hosp.1)	G	h	Unique
IP49	Env (Hosp.2)	G	h	Unique
IP18	Env (Hosp.1)	H	i	IV
IP17	Env (Hosp.2)	H	i	XIII
IP41	Bovineb	I	j	VII
IP50	Bovineb	I	j	VII
Cr53	Env (Hosp.3)	J	m	II
Cr52	Env (Hosp.3)	J	m	X
Cr60	Pat (Hosp.3)	K	n	Unique
Cr59	Pat (Hosp.3)	K	n	XIII
Cr66	Pat (Hosp.3)b	L	o	XI
Cr67	Pat (Hosp.3)b	L	o	XI
Cr55	Env (Hosp.3)	Unique	b	Unique
IP26	Pat (Germany)	Unique	f	VI
IP40	Pat (Hosp.2)	Unique	j	Unique
Cr71	Pat (Hosp.3)	Unique	j	III
IP3	Pat (Hosp.2)	Unique	k	I
IP2	Pat (CBS 144-89)	Unique	k	II
Cr51	Pat (Hosp.3)	Unique	n	I
IP45	Env (Hosp.1)	Unique	l	VIII
IP46	Env (Hosp.1)	Unique	l	VIII
Cr61	Env (Hosp.3)	Unique	n	XII
IP16	Env (Hosp.1)	Unique	n	IV
IP1	Pat (Hosp.2)	Unique	Unique	II
IP14	Env (Hosp.1)	Unique	Unique	II
IP47	Pat (Germany)	Unique	Unique	II
Cr54	Pat (Hosp.3)	Unique	Unique	II
Cr73	Pat (Hosp.3)	Unique	Unique	II
Cr74	Env (Hosp.3)	Unique	Unique	II
IP4	Env (Hosp.1)	Unique	Unique	III
IP21	Env (Hosp.2)	Unique	Unique	IV
IP11	Env (Hosp.1)	Unique	Unique	V
IP12	Env (Hosp.1)	Unique	Unique	IX
IP13	Env (Hosp.1)	Unique	Unique	IX
Cr70	Pat (Hosp.3)	Unique	Unique	X
Cr72	Pat (IP 2279-94)	Unique	Unique	X
Cr62	Pat (Hosp.3)	Unique	Unique	XII
Cr63	Env (Hosp.3)	Unique	Unique	XII
IP15	Env (Hosp.1)	Unique	Unique	XIV
Cr65	Env (Hosp.3)	Unique	Unique	XIV
IP10	Pat (Hosp.3)	Unique	Unique	Unique
IP22	Bovine	Unique	Unique	Unique
IP25	Pat (CBS 143-89)	Unique	Unique	Unique
Cr56	Env (Hosp.3)	Unique	Unique	Unique
Cr57	Pat (Hosp.3)	Unique	Unique	Unique
Cr58	Env (Hosp.3)	Unique	Unique	Unique
Cr64	Env (Hosp.3)	Unique	Unique	Unique
Cr68	Pat (Hosp.3)	Unique	Unique	Unique
Cr69	Pat (Hosp.3)	Unique	Unique	Unique

^aHosp.1, Hôpital Hôtel-Dieu, Paris, France; Hosp.2, Hôpital Trousseau, Paris, France; Hosp.3, Hôpital Henri-Mondor, Créteil, France; env, environmental; pat, patient; IP, Institut Pasteur Collection; Cr, Creteil Collection; CBS, Centraalbureau voor Schimmelcultures. 

^bIsolates IP41 and IP 50 are from the same animal; isolates Cr66 and Cr67 originated from the same patient. 

Typing results are shown in Table 1. Among the 67 isolates studied, RFLP analysis generated 49 genotypes, of which 37 were unique. MLP analysis yielded 43 genotypes including 28 unique types. RAPD analysis detected 31 distinct genotypes, with 17 unique types. Both RFLP and MLP methods generated an identical discrimination for 53 isolates and 33 genotypes (Tables 1 and 2). Statistical analysis, based on the comparison of nonunique genotypes (Table 2), indicated a significant concordance between RFLP and MLP (coefficient of contingency [C] = 0.91; P < 0.001). Among the 14 discrepancies observed, 11 corresponded to isolates displaying unique profiles only by RFLP; 8 of these unique types were assigned on the basis of differences in two discrete bands. The other three discrepancies (isolates IP32, IP33, and IP27) resulted from a higher differentiation by MLP due to one marker. Levels of concordance between RAPD and either RFLP or MLP were similarly low: perfectly concordant genotypes were only found for 13 and 12 genotypes respectively (corresponding to only 15 and 14 isolates, respectively) (Tables 1 and 2). In addition, two (or more) genotypes defined by RFLP or by MLP could not be related to one RAPD genotype. For instance, RAPD genotypes I to VI included isolates displaying both unique and nonunique RFLP patterns (Table 1). As a result, no significant concordance was obtained, either between RFLP and RAPD (C = 0.84; P > 0.05) or between MLP and RAPD (C = 0.83; P > 0.05). Additionally, six and seven isolates were found unique by RAPD but not by RFLP and MLP, respectively (isolates IP37, IP39, IP48, IP49, IP33, Cr60, and Cr55). In all cases, the uniqueness of the RAPD types was due to differences in two or more RAPD bands.

TABLE 2.

Concordance between RFLP, MLP, and RAPD upon analysis of 67 A. fumigatus isolates

Methods compared	No. of concordant genotypes			Cc (P)
	Total (%)a	Unique	Nonuniqueb
MLP and RFLP	33 (67.5)	26	7	0.91 (<0.001)
RAPD and RFLP	13 (26.5)	11	2	0.84 (>0.05)
RAPD and MLP	12 (27.9)	10	2	0.83 (>0.05)

^aPercentage calculated according to the method generating the highest number of genotypes. 

^bEach of these genotypes included two isolates. 

^cC was calculated from χ² established on a contingency table with k columns and r lines, and the Yates correction was used (21). The significance of C was estimated from the χ² values using (k − 1)(r − 1) degrees of freedom. 

This study is the first that compares the three completely different DNA fingerprinting methods with the highest discriminatory levels so far reported for A. fumigatus (3, 8, 16). The wide genetic diversity of this fungal species is reflected not only in strains from distinct sources and geographical locations but also within isolate populations sampled in one hospital, as 85% of the environmental isolates may be recovered only once (8). In the present study of 67 isolates originating from various environmental and clinical sources, each fingerprinting system was expected to detect a consistent level of genetic polymorphism. This polymorphism was best detected with RFLP and with MLP, not with the RAPD system, which generated the lowest variety of fingerprints. Concordance among these fingerprinting systems was also expected, and MLP and RFLP provided identical differentiation for 78% of the isolates and concordant types for 67%. RAPD had a considerably lower discriminatory ability with identical differentiation for 21 and 22% of the isolates when compared, respectively, with RFLP and MLP and did not generate data significantly concordant with that generated with MLP or with RFLP. This was either because RAPD did not differentiate isolates defined as unique by RFLP or by MLP or because it differentiated isolates considered nonunique by the other two systems. Thus, our results suggest that, in most cases, isolate discrimination by RAPD fingerprinting will be erroneous.

Discordant results between RAPD and RFLP hybridization have indeed been previously reported (1, 26), and many factors can be implicated in the explanation. First, PCR components and parameters and DNA quality and gel electrophoresis duration are known to affect the reproducibility of RAPD, as does the discriminatory strength of selected primers (17, 19, 24). In our study, DNA quality does not seem to interfere with the RAPD patterns since no modification of the patterns was seen when a DNA extraction protocol currently used for RAPD was tested on 10 isolates (26). Second, visual comparison of the banding patterns of different gels may be subjective, even in the presence of internal standards (25). Third, a band may consist of different RAPD products with equal electrophoretic mobilities (21). Fourth, the appearance or disappearance of a RAPD band is more likely to result from nonspecific primer hybridization to the DNA template during amplification than from some mutation event at a specific priming site (5, 28).

In contrast to RAPD fingerprinting, the MLP and RFLP genotyping systems are based on specific DNA hybridization between either a probe or primer(s) and the target DNA. The reproducibility of these techniques is therefore high. Moreover, both methods detect definite phenomena: either mutations in enzyme restriction sites for the RFLP system or differences in the number of CA repeats for the MLP system (3). Under the assumption of a repartition of these phenomena in the A. fumigatus genome, the congruence between the two techniques is not surprising. Presently, RFLP remains the more discriminating system, although complex RFLP patterns can be difficult to interpret (9). In the future, the discriminatory power of the MLP system will be probably improved by adding polymorphic microsatellite sequences to be identified in the ongoing sequencing project of the A. fumigatus genome, as seen recently for Saccharomyces cerevisiae (13). MLP analysis is easy since a single PCR band product is obtained, in accordance with the haploidy of the A. fumigatus genome. An additional advantage of the MLP typing system over the RFLP and RAPD methods is its capacity to detect a mixture of isolates (3). Indeed, two original isolates were rejected from our study upon detection of distinct PCR band products at a single locus and confirmation of their cross-contamination upon monospore subculturing. However, some artifacts may occur during amplification of dinucleotide repeats and must be overcome by optimization of PCR conditions (7, 12) and monitored by including well-characterized strains in each typing (3). MLP and RFLP have been also successfully used to type other human pathogens (6, 23).

Regarding our results, the very convenient RAPD approach should be not favored for A. fumigatus, even as a first-line typing strategy. In contrast, because of the methodology used, MLP analysis can give quick, reliable, computerizable data and may be considered the first-line choice for typing A. fumigatus isolates, despite its high cost. If more discrimination is needed, RFLP, presently the most powerful technique of the three tested in this study, should be used.