Applications of Differential-Display Reverse Transcription-PCR to Molecular Pathogenesis and Medical Mycology

Abstract

The host-fungus interaction is characterized by changes in gene expression in both host and pathogen. Differential-display reverse transcription PCR (DDRT-PCR) is a PCR-based method that allows extensive analysis of gene expression among several cell populations. Several limitations and drawbacks to this procedure have now been addressed, including the large number of false-positive results and the difficulty in confirming differential expression. Modifications that simplify the reaction time, allow the use of minute quantities of RNA, or address unusual species- or gene-specific sequences have been reported. DDRT-PCR has been used to address biological questions in mammalian systems, including cell differentiation, cell activation, cell stress, and identification of drug targets. In microbial pathogenesis and plant pathogenesis, DDRT-PCR has allowed the identification of virulence factors, genes involved in cell death, and signaling genes. In Candida albicans, DDRT-PCR studies identified TIF-2, which may play a role in the upregulation of phospholipases, and the stress-related genes, CIP1 and CIP2. In Histoplasma capsulatum and C. albicans, genes involved in the host-pathogen interaction, including a member of the 100-kDa family in Histoplasma and an ALS and 14-3-3 gene in Candida, were potentially identified by DDRT-PCR. Although very few reports have been published in medical mycology, studies in mammalian, nonfungal microbial, and plant pathogen systems are easily applied to basic questions in fungal pathogenesis and antifungal therapeutics.

The incidence of fungal infections has dramatically increased in recent years, due in part to the onset of AIDS and improved medical techniques including the use of antibiotics, immunosuppressive therapy, and organ transplantation. In many instances, fungi such as Candida albicans and Aspergillus fumigatus are normally benign or even endogenous organisms, but they can lead to serious and indeed fatal infections in immunocompromised individuals. Additionally, international travel and exploitation of new habitats have led to the reemergence of old fungi and exposure to previously unknown fungi (54). However, the dramatic increase in fungal infections has not been paralleled by an increase in antifungal therapeutics or diagnostic techniques (142, 144, 145).

Although the predisposing factors associated with fungal infections are known, little information exists regarding the molecular events that occur when the fungal cell interacts with the host. Fungal surface components are clearly important in the initial interaction between fungus and host. Molecules associated with adherence and stage-specific antigens have been biochemically described, but the importance of these antigens to disease development is poorly described (18, 99, 105). Due to the complex nature of the fungus-host interaction, genes which respond to temperature and pH and those involved with cell wall morphology are all regulated during this process. When this transition occurs in the host, more variables are involved. Thus, it is implied that genes in both host and fungi are both activated and inactivated during pathogenesis. Identification of the genes differentially expressed during host recognition, attachment, internalization, and/or morphogenesis could lead to a better understanding of the mechanisms of disease. This in turn would provide insights into potential antifungal drug targets and diagnostic markers.

Historically, differential gene expression in fungi was indirectly studied by comparing antigen or protein profiles of two or more cell populations. This approach delivered new information but was time-consuming and rarely gave information on regulatory mechanisms. Additionally, while differential protein expression could be documented by two-dimensional electrophoretic patterns, it was difficult to isolate the proteins of interest from the gels. However, with the advent of proteomics, improvements in computer imaging and mass spectrophotometric analysis, and public access to a broadening fungal genomic database, two-dimensional protein analysis may soon be a practical approach (39, 141). With the onset of molecular biology techniques, subtractive library screening and differential hybridization became the method of choice for identifying differentially expressed genes. In subtractive library screening, a subtracted cDNA library is constructed from different RNA sources. For example, mRNA is isolated from cells before and after attachment to host tissue. In the attached population, cDNA is synthesized, denatured, and then allowed to hybridize with the RNA from the unattached cells. After exhaustive hybridization, cDNA that has not formed cDNA-RNA hybrids is purified by hydroxyapatite chromotography and used as a template to form double-stranded cDNA, resulting in a library that should consist of cDNAs that are expressed only upon attachment to the host (126). Differential hybridization is a bit simpler in that a cDNA library is constructed from one cell population and then duplicate filters are screened with probes made from two different sources. This technique was successfully used to identify genes in C. albicans that responded to a changed environment (9, 56, 127). The major disadvantages of these methods are (i) the need to construct and screen cDNA libraries, (ii) the requirement for large quantities of RNA, (iii) nonrecognition of subtle changes, (iv) the ability to identify changes in only one population, and (v) the ability to study only two cell populations or variables simultaneously.

In the last decade, PCR-based methods have been developed, including RNA fingerprinting, arbitrarily primed PCR (AP-PCR), and differential-display reverse transcription-PCR (DDRT-PCR), that allow the identification of differentially expressed genes in a less labor-intensive fashion than the methods described above. The aim of this review is to provide a resource for laboratories planning to study differential gene expression in medically important fungi and thus assess the feasibility of using DDRT-PCR. For this reason, a comprehensive methodology section is included. Very few reports of DDRT-PCR in medical mycology have been reported; therefore, the review summarizes pertinent approaches in relative fields. The review focuses on the modifications, adaptations, and optimizations made to DDRT-PCR since the procedure was first reported in 1992. Second, examples in mammalian and nonfungal microbial pathogenesis models that can be applied to medical mycology are discussed. Finally, examples of DDRT-PCR applied to fungal interactions, both nonmedical and medical, are reviewed.

DDRT-PCR

Method

Original method.

A fingerprinting technique to identify and compare mRNAs expressed during different cell processes was first described by Liang and Pardee (74). The requirements of this protocol were (i) reproducibility, (ii) comparison of all mRNA species in the cell populations of interest, and (iii) the ability to isolate the corresponding cDNA. The DDRT-PCR procedure consists of two major steps: (i) reverse transcription (RT) of RNAs isolated from different cell populations with a set of degenerate, anchored oligo(dT) primers to generate cDNA pools, and (ii) PCR amplification of random partial sequences from the cDNA pools with the original anchored dT primer and an upstream arbitrary primer (Fig. 1). The annealing temperature of the PCR is low to maximize the number of amplified mRNA species. α-³⁵S-labeled radioisotope is included in the amplification step so that the short sequences can be visualized on a sequencing gel (Fig. 1B). DDRT-PCR is performed with the same primer sets on multiple cell populations (74).

FIG. 1.

DD of C. albicans after adherence to host cells or ligands. (A) Schematic of DDRT-PCR. (B) RNA was isolated from Candida grown to mid-exponential stage in YPG (lanes 1), Candida which had adsorbed to plastic (lanes 2), Candida which had adhered to HET-matrix (lanes 3), or Candida which had adhered to matrix only (lanes 4). DDRT-PCR was performed with dMGT₁₂, dMAT₁₂, dMCT₁₂, or dMTT₁₂ and the arbitrary decamer dTGGTAAAGGG. (C) Bands which were the most clearly differentially expressed in at least one population of RNAs were excised from the gel and reamplified with the same primer set. Note that in some lanes, more than one band was amplified. (D) RNA was isolated from an overnight C. albicans culture (row 1), Candida incubated as in row 3 but with no extracellular matrix (row 2), Candida incubated at 37°C in minimal essential medium with extracellular matrix (row 3), HET-1A cells (no Candida) (row 4). RNA (0.5 and 1 μg) was spotted onto a nylon membrane. The membrane was probed with ³²P-labeled DD products (25a, 25b, 35a, or 35b) or BMH. BMH was subcloned from DD product 25b.

For optimal display, the primer sets are designed to amplify 50 to 100 mRNAs. The original method proposed that 12 oligo(dT) primers were necessary: 11 Ts to anchor the primer to the 3′ polyadenylation site and 2 additional bases (e.g., 5′-T₁₁GC anneals to mRNAs ending in GCA_n). In this way, each oligo(dT) primer recognizes 1/12 of the mRNA population, since there are 12 combinations of the last two 3′ bases excluding T at the penultimate spot. The 5′ primers are arbitrary sequences designed to anneal to cDNAs at random distances from the poly(A) tails. In theory, the primers should be short (6 to 7 bp) to allow frequent annealing near the end of cDNA strands; Liang and Pardee ascertained that the use of a 10-base arbitrary primer, together with the oligo(dT) primer, gave specific and reproducible results. Additionally, 10-mers amplified more bands than statistically expected, suggesting that degenerate hybridization does occur. The sequencing of selected DDRT-PCR products proved this to be true. This is advantageous since fewer primer pairs are necessary to display all mRNAs. For mammalian cells, it was calculated that 20 arbitrary 10-mers in conjunction with 12 anchored primers would statistically amplify all mRNA sequences.

Refined method.

Almost immediately, Liang et al. refined DDRT-PCR with a few key optimizations (73). These included reducing the number of anchored primers from 12 to 4. It was noted that the penultimate base from the 3′ end of the oligo(dT) primer exhibited considerable degeneracy. Band patterns with four anchor primers, T₁₂MG, T₁₂MA, T₁₂MC, and T₁₂MT (where M = G + A + T + C) and an arbitrary primer were comparable to those obtained using all 12 original anchor primers. This significantly reduced the number of RT reactions that had to be performed. Second, it was determined that the starting quantity of total RNA could be as small as 200 μg and that poly(A)⁺ RNA was not necessary. In fact, residual oligo(dT) matrix used to isolate poly(A)⁺ RNA may interfere with the differential display. Finally, bands that migrated at less than 300 bp often appeared as doublets or triplets. This appeared to be due to the unequal A tailing by the Taq polymerase. Interestingly, because DDRT-PCR may be more sensitive than cDNA libraries, it can be used to assess the quality of such library (73).

Figure 1B gives an example of a display. In this instance, genes that are differentially expressed in C. albicans after adherence to host cells or ligands in an in vitro pathogenesis model were identified (99). Four differentially treated populations were amplified with the same primer set. Note that several bands are displayed in each lane (only a portion of the acrylamide gel is represented) and that the dT₁₂MT primer gave very poor results. This was probably due to the anchored primer annealing nonspecifically to any poly(A) tail. DDRT-PCR was performed by the original modified method (73, 75). It should be noted that several companies sell DDRT-PCR kits, and one from GenHunter (Nashville, Tenn.) is referenced most frequently.

Once differentially displayed (DD) PCR products are identified, the cDNA species of interest are excised from the dried sequencing gel, extracted by boiling, and reamplified with the original primer sets (Fig. 1C). This PCR product can then be directly subcloned, sequenced, and used as a probe in Northern analysis to confirm differential expression or to screen libraries (Fig. 1D).

Consequently, DDRT-PCR has several advantages over subtractive and differential hybridization. These include (i) rapidity and simplicity of assays, (ii) small quantities of RNA, (iii) increased sensitivity, (iv) reproducibility, (v) ability to identify differentially expressed genes in more than one population, and (vi) ability to compare several cell populations or variables simultaneously. However, the assay does have its limitations, including false-positive results and the inability to confirm differential expression. Additionally, it is unclear how well low-abundance mRNAs are represented (7). Since the first description of DDRT-PCR, numerous optimizations and potential adaptations that address these issues have been reported, and these are discussed below.

Modifications

Primers.

Many modifications of primer design have been suggested in order to simplify amplification reactions, reduce false-positives and facilitate cloning and sequencing (reviewed in Table 1).

TABLE 1.

Primer modifications

Primer	Modification	Reference(s)
2-base-anchored primer	Simplified reactions	73
1-base-anchored primer	Simplified reactions	77
Longer primer	Higher stringency, simplified cloning	155
	Direct sequencing	27, 79, 148
	Inclusion of coding region	67
1-anchor primer	Simplified reactions	125
Modified short primers	Increased number of products	2, 121
Special primer sets	Statistically chosen	13, 86
	Coding bias	13, 36, 106, 147
	Specific primers	68
	AT-rich primers	42
	End labeling	49, 139
Limiting RNA	3′ degenerate ends	154
	Solid support	123
	Unique 3′ motifs	31, 150

(i) Anchor primers.

The first modifications were further improvements of the anchor primers (77). The M at the penultimate site was omitted, leaving only a 1-base anchor: N₁₀T₁₁A, N₁₀T₁₁G, or N₁₀T₁₁C (T₁₁T is omitted) (N₁₀ indicates that 10 nucleotides were added at the 5′ ends, including a restriction site to aid in cloning [77]). Consequently only 3 anchored primers in conjunction with arbitrary primers are now considered sufficient for a comprehensive display.

(ii) Long primers.

Initially, short primers were considered optimal to more accurately represent mRNA populations; it is now considered advantageous to use longer primers. The use of 20-mer primers in combination with a two-step PCR procedure increases reproducibility and DD product size. Restriction enzymes sites are included at the 5′ end to facilitate cloning (77, 88, 155). A sequence of 10 bases, including an EcoRI site to facilitate cloning, was placed 5′ of both the original arbitrary decamers and 1-base-anchored primers. In the first cycle, primers were allowed to anneal under low-stringency conditions at 40°C for 4 min followed by 35 cycles at 60°C for 2 min. The higher annealing temperature increases stringency and reduces false-positive results (88, 155). These conditions were used to analyze normal and transformed rat fibroblast cells. The results were reproducible and at least as sensitive as the original method (155). Some researchers claim that successful DDRT-PCR can be performed with only one anchor primer, dT₁₂MN, which consists of equimolar quantities of dT₁₂MG, dT₁₂MA, dT₁₂MT, and dT₁₂MC (125). In general, three anchor oligo(dT) primers, either mono- or dinucleotide, are now routinely used in DDRT-PCRs. The dinucleotide anchored primers produce fewer bands per track but allow higher resolution (88).

An additional limitation of DDRT-PCR is that often little information is gained from sequencing the PCR product. The majority of the sequence is in the 3′ untranslated region (UTR). The lengthening of primers increases product size, resulting in longer sequences that can demonstrate homology to sequences reported in the database (67). Longer primers also improve direct sequencing of DDRT-PCR products (27, 67, 79, 88, 148). Either the anchor or the arbitrary primer can be end labeled and used to sequence the PCR product. Alternatively, the product can be labeled and used in Northern analysis. If the product is to be characterized further, the restriction enzyme sites on both ends facilitate directional cloning (148). Laboratories report between 10 and 50% success rates with direct cloning (67, 88).

(iii) Short primer modifications.

Although many workers support the use of increased primer length to reduce false-positive results, the original premise behind the use of short primers is that they more accurately represent the complexity of a mRNA population. The major limitation of short primers is lower stringency during PCRs and therefore inefficient hybridization. Afonina et al. suggested chemically modifying primers to improve hybrid stability (2). It is unclear whether this is time effective. Another limitation of short primers and low stringency is nonreproducibility due to imperfect annealing of primers. Since inosine can bind with all four bases at the same strength, arbitrary decamers can be synthesized with one to five inosines, allowing the annealing temperature to be increased (121). While this improves reproducibility, the primers are considerably more expensive and it is probably more useful to use longer primers.

(iv) Special primers.

In most instances, the 1-base-anchored 23-mer and arbitrary 20-mer primer sets (77, 148) are sufficient to address any number of biological questions. However, further refinements are possible depending on the questions being asked, balanced against time and cost restraints. Several laboratories have reported the efficient design of primer sets to increase screening potential (86). Primer sets are statistically chosen to reflect the coding bias of specific cell populations or species (13, 36, 106, 147). This is particularly useful for bacteria where mRNA is not polyadenylated (36). This approach could be useful in medical mycology by designing arbitrary primers based on codon usage in known genes of a particular fungal species.

Jurecic et al. (68) were successful in using specific primers (≥20 bp) as arbitrary primers at low annealing temperatures. The advantage of this method is that primers already present in the laboratory can be used. Specific primers can also be used in DD to target a particular class of cDNA (see “Targeted display” below) (68).

(v) Arbitrary primer amplification.

Other laboratories have reported that a limitation of DDRT-PCR is that many of the products are reamplified with the arbitrary primer alone (42, 49, 139). Several approaches have been developed to minimize this problem. One group reasoned that the G-C interaction in the arbitrary-primer annealing may result in stronger annealing than the AT-rich oligo(dT) does (42). Therefore, 12- to 14-mers were designed, with an A+T content of 60 to 80%, rationalizing that this would result in a PCR competitive equilibrium during the amplification reaction (42). Since some fungal genomes are AT rich, this may be a beneficial adaptation for fungal DDRT-PCR.

A second approach to overcome false-positive results due to amplification by the arbitrary primer alone is to end label the anchor primer (49, 139). cDNA products resulting from amplification of the two arbitrary primers will not be visible on an autoradiograph (139). On the other hand, some researchers claim that DDRT-PCR is improved by amplifying with two arbitrary primers instead of the anchored oligo(dT) primer (48). The use of arbitrary primers reduces amplification of the 3′ UTR. Haag and Raman also claim that the gel exhibits improved resolution (48). This is basically AP-PCR and is described below.

(vi) Limiting amounts of RNA.

Several laboratories have modified primers to enable them to perform RT-PCR with even smaller amounts of RNA than were originally proposed. Unlimited DD reactions were performed with 20 ng of RNA (10³ cells) (154). Degenerate ends were placed on the 3′ end of both the anchor and arbitrary primers. The degenerate sequence (N₄) at the 3′ end takes into consideration the mismatch tolerance of Klenow polymerase and increases binding 256-fold. Reamplification is performed with the same primers minus the degenerate 3′ ends. This protocol can be applied to cells isolated by microdissection or clinical needle biopsy (154). Another approach is to couple mRNA to a solid phase, e.g., streptavidin-coated magnetic beads. RT is performed with biotinylated anchor primers (123). Rosok, et al. were able to display differentially expressed fragments from less than 5 ng of RNA (10⁴ stem cells). Of interest, RT with biotin-dT₁₁ gave reproducible, clear results but RT with biotin-dT_18,25 resulted in smears. These methods may be useful in medical mycology, when only small numbers of cells can be obtained.

Several laboratories have designed primers to improve sensitivity or enrich for moderate- to low-abundance mRNAs by targeting specific sequences including the polyadenylation signal AATAAA or ATTTA, a motif in short-lived mRNAs (150), and AU sites (31). AU-rich elements are regulatory sequences found in the 3′ UTR of labile mRNAs encoding cytokines, proto-oncoproteins and transcription factors. All these genes are of pharmacological interest (31). Although it is unclear if either of these sequences would apply to fungal sequences, it does indicate that specific 3′-anchored primers can be used to generate DD of labile mRNAs.

Reamplification.

A major limitation reported for DDRT-PCR is the inability to selectively reamplify the majority of cDNA products in order to sequence and confirm differential expression. The amount of cDNA used in the reamplification step is important. The original protocol calls for 2 μl, and this is sufficient. Greater than 5 μl normally results in no product rather than more (30). One of the major problems is excision of more than one band from the gel. Using 40 cycles of PCR in the reamplification step greatly increases the concentration of contaminating or comigrating DNA. Normally this problem is addressed by sequencing between 5 and 10 individual clones per gel fragment. However, if large numbers of clones need to be screened, this approach may be very labor-intensive.

(i) Clone screening.

Several laboratories have developed methods to reduce the number of clones to be screened by prescreening the clones by restriction digestion or dot blotting. To limit the number of sequencing reactions, isolated cDNA fragments or clones can be digested with restriction enzymes, e.g., Sau3AI, to determine how many species are present in one DDRT-PCR. If one pattern is predominant, this clone normally represents the cDNA of interest (93, 133, 135). Alternatively, 5 to 10 colonies per insert can be analyzed by Southern or dot blot analysis. Duplicate filters are probed with ³²P-labeled cDNA from the original DDRT-PCRs to identify the correct colony (24, 146). Obviously, if multiple cell populations were compared, this would not be a very time-effective procedure.

(ii) Precloning.

At least three protocols have also been described which allow the elimination of contaminating cDNAs or enrichment of the desired cDNA prior to cloning. The cDNA of interest can be separated by single-strand conformation polymorphism (SSCP) (91, 95). SSCP differentiates between fragments that are similarly sized but have different sequences. The isolated cDNA is amplified for only five cycles before being cloned into a T vector (95). If the SSCP pattern is complex, the most intense band is cut out and reamplified. If there is no difference in intensity, this fragment is omitted from further analysis (91).

Other laboratories describe methods to reduce false-positive results from comigration, but these are labor-intensive because they require more complicated reamplification steps. These methods are not practical for most applications and so are not described in detail here. Methods to reduce false-positive results include incorporation of [α-³⁵S]dATP in the reamplification reaction mixture and running it in tandem with the original DDRT-PCR to determine whether the correct cDNA is reamplified (38). Second, the DDRT-PCR is repeated with the same arbitrary primer and the three nondegenerate oligo(dT) primers that are subsets of the original anchored primer. The rationale is that products of amplification by the arbitrary primer alone will be easily recognizable, since they will be of equal intensity in all three lanes. The ability of the comigrating DNA to outcompete the desired cDNA is reduced, since there will be approximately one-quarter the number of mRNAs amplified by the less degenerate anchored oligo(dT). In this manner, 80% of the fragments tested were confirmed to be differentially expressed by Northern analysis compared to 10% confirmed using the original primer pairs (136). Third, a subtraction hybridization step is included prior to cloning (103). This method is certainly more labor-intensive than the original method described but may be time effective if a large proportion of isolated cDNAs contain contaminants. The individual researcher must determine whether this protocol is time effective.

(iii) Vector sequences.

An interesting adaptation to reamplification is the transient ligation of the cDNA product to a vector or the use of vector sequences (10, 116). Bonnet et al. transiently ligated a T vector directly to the isolated DDRT-PCR product (10). They then reamplified this product by nested PCR with the original oligo(dT) primer and 5′ primers located in the T7 cloning site of the vector. Two more PCRs were performed with nested 5′ primers, which eliminated fragments primed by only one primer. Sufficient DNA was obtained in this method to perform sequencing reactions, RT-PCR, or reverse Northern analyses. This method allowed 57.5% successful reamplification versus 31% for the original method. Transient ligation of the extracted DNA to a T vector also allows direct sequencing with T7 (116). Instead of transient ligation, the 5′ ends of arbitrary and anchor primers were extended with half of the T7 or SP6 promoter site and a restriction endonuclease site (11). The reamplification primers contained the whole promoter sequence. The longer sequences allow PCR reamplification to be performed with higher stringency and direct generation of antisense riboproteins or sense RNA (11).

Screening.

DDRT-PCR allows the rapid identification of gene products and requires small amounts of RNA. Several screening protocols have been described, particularly to reduce the amount of time required to analyze a large number of clones or to confirm differential expression when limited RNA is available.

(i) Large screens.

The following method utilizes 96-well technologies to facilitate the screening of large numbers of potentially differentially displayed products (25). Rapid screening is achieved by (i) direct amplification of the products from the sequencing gel, (ii) ligations in a 96-well plate using reduced concentrations of reagents, (iii) screening of clones by colony PCR, and (iv) confirmation of altered expression by dot blot differential hybridization. DD products were directly amplified from dried gels and cloned in 96-well plates using the TA Cloning kit (Invitrogen, San Diego, Calif.) but with 12.5 ng of vector instead of 50 ng. Colony PCR (12 colonies per ligation) was performed using primers located in the TA vector to identify inserts of the appropriate size, and the amplified fragments were spotted onto membranes. Duplicate membranes were then probed with ³²P-labeled control and treated cDNAs (25). RNA was isolated from livers from control and WY (peroxisome proliferator)-fed rats and compared by DDRT-PCR. DD bands were cut directly from urea-free gels, and 80% were successfully amplified after 40 cycles of PCR. When more than one PCR product resulted, all bands were cloned. A total of 414 cloned fragments resulting from 26 primer pairs were screened, and altered expression was confirmed in 30 (25). Although there was a large number of false-positive results, the time needed to obtain the 30 products is considerably shorter than for the normal procedure, and the process is less labor-intensive (414 products were screened twice each in 6 weeks). Only one clone had previously been identified to be down regulated by peroxisome proliferators. The large number of false-positive results was thought to be due to cloning of multiple PCR products per DD band and selection of bands that were only marginally altered in expression (25). If a large number of clones needs to be screened, this is a cost- and time-effective approach.

(ii) Hybridization filter arrays, virtual Northern analysis, and reverse Northern assays.

The hybridization filter array, virtual Northern analysis, and reverse Northern assays use cDNA instead of RNA for the confirmation step and are particularly useful when only low concentrations of RNA are available (57, 87, 88, 114, 124). Hybridization arrays provide a valid, rapid, prescreen means of confirming differential expression; these arrays are more accessible to the scientific community than are microarray chips (87, 88, 114). Different approaches can be used to reamplify the DD products to construct gene tags. One method is to ligate adapter sequences (UNI-Amp; Clontech Laboratories, Palo Alto, Calif.) to cDNA sequences and PCR amplify with primers homologous to the adapter sequence (124). A second approach is to design nested primers from sequenced DD products (87, 88). The gene tags are then spotted onto membranes to form hybridization filter arrays. In one study, DD bands were directly sequenced and nested primers were designed to hybridize just inside the arbitrary primer site. The nested primer and the appropriate anchor primer were used to PCR amplify a gene tag for hybridization arrays. DNA gene tags were spotted on membranes in a 96-well format to form hybridization arrays. Each gene tag was spotted serially in two or four spots. Replicate membranes were hybridized with ³²P-labeled total cDNA from different cell populations. It is important to assay decreasing concentrations of the gene tag since differences are sometimes seen only at lower concentrations. The hybridization array was deemed valid for several reasons: (i) down regulation was confirmed for 95% of the genes selected by DD, (ii) 9 of 10 gave corresponding results in Northern analysis, (iii) appropriate expression was seen for control nondifferentially and differentially regulated genes, and (iv) loading controls indicated that the gene tags were reliably loaded (87). The use of hybridization filter arrays appears to be a practical way of prescreening candidate differentially expressed genes. Additionally, the filter can be reprobed with cDNA prepared from multiple cell populations. The laboratory must decide which approach to use to prepare gene tags. The use of adapter linkers considerably reduces the number of primers needed, and prior sequencing is not required. However, this method relies on efficient ligation of adapters to the cDNA ends.

(iii) Miscellaneous.

The methods described in this section claim to improve screening but are probably practical only in specific circumstances. Cloned DDRT-PCR products are used to isolate longer clones from high-quality cDNA libraries constructed from developmentally different cell populations (47). The longer clones are used in in situ hybridization to confirm differential expression during development. Although the success rate was only 25%, this method proved to be time effective (47). One potential approach to reduce the number of DDRT-PCR products screened is to use cytoplasmic instead of total RNA. This excludes nuclear RNA and therefore rRNA, introns, and immature transcripts (83, 136).

(i) Radioactivity.

Several factors determine which radioisotope to use or whether to use radioactivity at all. [α-³⁵S]dATP was originally chosen since it was routinely used in sequencing gels, offered high resolution, and necessitated minimal physical protection (74, 76). However, [α-³⁵S]dATP is volatile and can lead to contamination of the PCR machine during DDRT-PCR (22, 140). The amount of escaping ³⁵S-labeled gas (SO₂⁻) was decreased when heated lids were used. Alternatively, the labeled gas can be trapped by placing a filter soaked in lead acetate or barium chloride over the tubes (22). ³²P is an alternative, but the autoradiograph must often be developed twice (76). Reactions performed with end-labeled primers result in more intense labeling of small fragments, while nucleotide incorporation results in a stronger signal in larger fragments. The major disadvantage, however, is that increased physical protection is required. ³³P is a viable alternative because it combines the advantages of both ³⁵S and ³²P and is more sensitive (139). A 1-μCi portion of ³³P allows the same sensitivity and resolution as 5 μCi of ³⁵S, thus making it cost-effective (76).

(ii) Staining.

DDRT-PCRs can be displayed in 1.6% agarose gels and stained with ethidium bromide (13, 122). Although sensitivity is decreased and fewer bands are displayed, differentially expressed bands can be detected and validated. Therefore, it is a potential method to avoid radioactivity if identification of the total mRNA population is not essential. Although silver staining allows direct visualization and isolation of the product (41, 70, 81), it is normally less sensitive, procedures are difficult to control, and results are inconsistent (3).

(iii) Fluorescence.

There are several advantages of fluorescent labeling, including safety, stability of reagents, low cost and easy disposal, and high throughput. Primers labeled with JOE, FAM (blue), ROX (red), or TAMRA (yellow) gave reproducible results (60, 61, 63, 84). HEX (green) was not efficient. In contrast to radioactivity (76), direct incorporation of fluor-dUTP (labeled nucleotide) resulted in blurry bands (84). Analysis was also possible by simultaneous comparison of several primer combinations due to the multicolored banding patterns (63). However, the use of fluorescent arbitrary primers did not overcome the problem of amplification with decamer alone and, additionally, synthesis of fluorescent primers is expensive. Smith et al. proposed amplification with a fluorescently labeled universal primer (134). During RT, a common 20-mer sequence was added to the 5′ end of each cDNA strand. Amplification was performed with a decamer, which primes first, and the fluorescently labeled primer, which is analogous to the 5′ end of the oligo(dT), primes the second strand. The advantages to this method are as follows: (i) the fluorescent primer is a constant in every PCR; (ii) artifacts from only the 10-mer amplifying cDNA are not seen; (iii) the ease of sample throughput and data handling is increased; and (iv) a single fluorescent primer in all reactions is cost-effective (134). However, a major drawback to fluorescence is that cDNAs cannot be recovered from the automatic sequencer. Consequently, DDRT-PCR has to be reproduced exactly with a radioisotope in order to recover the bands of interest. Therefore, fluorescence DDRT-PCR would be useful when the purpose of the experiment is to screen a large number of mRNA species, obtain an overview of all changes in several cell populations, or identify sequence changes in mutant populations but not when the goal is to physically isolate differentially expressed genes.

(iv) Chemiluminescence.

Several laboratories have been able to use chemiluminescence (3, 20, 114). Anchor primers are labeled with digoxigenin (DIG) (20) or biotin (3). In the former case, the sequencing gel is transferred and the filter is blotted with anti-DIG alkaline phosphatase and developed colorimetrically. Bands can be amplified directly from the excised strip and reused up to five times in PCRs (20). For the biotin system, the gel can be treated with a SEQ-Light chemiluminescent detection system (3). While both these methods eliminate the need to cut out nonvisible bands, they require extensive handling of a thin acrylamide gel and are technically more difficult than drying and developing a gel. Therefore, the expertise of personnel and the routine procedures used in the laboratory should be considered when choosing a detection method.

Nonradioactive methods can also be applied to the screening protocol. The optimal conditions for using DIG-cDNA to probe hybridization filter arrays were carefully refined (114). Results were comparable to radioactive screening, shorter exposure times were required, and low-abundance transcripts were detected (1 to 15 transcripts per cell). However, disadvantages include the following: (i) strict conditions are required for detection with low backgrounds and high sensitivity and (ii) filters cannot be stripped and reprobed. On the other, hand the probe is reusable (114).

Enzymes.

The selection of reverse transcriptase (RTase) is important. RNase H-deficient RTases, e.g., Moloney murine leukemia virus (MMLV) RNase H RTase, are superior to avian myeloblastosis virus RTase (114, 133) Superscript RTase gave contradictory results. Rajeevan et al. found Superscript RTase to be superior to MMLV RTase (114), while Rosok et al. reported it to be inferior (123). However, conditions were not the same between the laboratories and contradictory results were accredited to the solid support used by Rosok et al. (123). Other laboratories report satisfactory results with Superscript RTase (Life Technologies) (99, 156). In most cases, any RNase H-deficient RTase can be used in DDRT-PCR.

Little difference was observed between cold- and hot-start DD-PCR with Taq polymerase, although a hot start is required with rTth DNA polymerase. Taq is reported to be better than Stoffel (123). In addition, the amount of optimal radioactivity to use was dependent on the type of polymerase (123). Again, the Taq polymerase routinely used in the laboratory can be used for most DDRT-PCR studies.

Summary

Regardless of whether the original DDRT-PCR method with no major modifications or one of the modified methods is used, the following steps should always be included to significantly reduce false-positive results. High-quality DNase-free RNA template is required. RNA must always be treated with DNase (73). To reduce false-positive results due to intersample variability, two uninduced and two induced populations or several time points during the course of induction should be run simultaneously (136). The product is selected only if it is differentially expressed in both induced populations compared to both uninduced populations. Alternatively, experiments are performed in duplicate using different cell donors and different sets of reagents in in vitro experiments (98).

The major limitation of DDRT-PCR is the large percentage of false-positive results. The major reasons for false-positive results are low-stringency annealing, isolation of comigrating PCR products (contamination), and amplification by only the arbitrary primers. Steps to avoid these problems can be taken during primer design, reamplification, or screening. For the average laboratory and for routine DDRT-PCR studies, it is advised to use longer primers and the two-step PCR cycle protocol described by Zhao et al. (88, 155). Differential expression can be confirmed by dot blot Northern analysis or, if the amount of available RNA is limiting, hybridization filter arrays or reverse Northern analysis. All these screens are amenable to 96-well plate configurations. Radioactive detection can be replaced by chemiluminescence or fluorescence depending on the expertise of the laboratory personnel.

METHODS DERIVED FROM DDRT-PCR

RNA Fingerprinting by Arbitrarily Primed PCR

RNA fingerprinting by AP-PCR (RAP-PCR) is essentially the same as DDRT-PCR except that the RT step is performed with an arbitrary oligonucleotide instead of an oligo(dT) primer. The amplification step is performed with the same or a different arbitrary oligonucleotide (92, 94). As observed with DDRT-PCR, the arbitrary primer normally only matches 6 to 8 out of 10 nucleotides at the 3′ end (94). RAP-PCR is normally used for RNA that is not polyadenylated, e.g., in bacteria (36), or to minimize the amplification of 3′ untranslated regions (48). The limitations of RAP-PCR is that both the primer match and the abundance of each RNA dictate the display of a product. An abundant RNA will undergo more amplifications with primers with three to four mismatches than will an RNA with no mismatches (94). While this also occurs in DDRT-PCR, it is accentuated in RAP-PCR by the use of two arbitrary primers. It is also more common that one RNA species will be displayed more than once, since both ends are based on an arbitrary sequence and not a poly(A) tail. Finally, more reactions must be performed, since fewer bands are displayed per primer pair. Similar to DDRT-PCR, RAP-PCR can be used to compare transcripts among many different cell populations or differentially treated populations. For example, lung epithelial cells were treated or not treated with transforming growth factor β (TGF-β), which halts cells in late G₁, and RNA was compared at various time points; cells treated with cycloheximide to inhibit protein synthesis were also compared (94). The mRNA display showed that 2.7% of the transcripts were induced by cycloheximide compared to 1.2% induced by TGF-β. Additionally, a substantial number of changes appeared to be the same for both cycloheximide- and TGF-β-treated cells. These observations were obtained without any further manipulations, such as subcloning and sequencing (94).

DDRT-PCR with Selected Primers

The purpose of DDRT-PCR with selected primers (SPR) is to increase the bias toward identifying moderate- to low abundance mRNAs (59). SPR differs from RAP-PCR and DDRT-PCR in that the arbitrary primers have been experimentally chosen so as to avoid the amplification of highly abundant transcripts. This has been addressed in four ways. (i) Primers were designed with a 50% G+C content overall but less than 50% at the 3′ end to reduce amplification of highly abundant rRNA and mitochondrial RNA. (ii) All PCR cycles were performed at a high annealing temperature (50°C). This increases selectivity and reproducibility but lowers band numbers. (iii) Band selection was based on amplification due to both primers, since highly abundant transcripts are preferentially primed with one primer. This was accomplished by comparing the display of PCR performed with both primers to that of PCR performed with 3′ primer only. (iv) Quantitative RT-PCR was used to confirm differential expression, since Northern analysis may not detect low-abundance mRNA (59). Some of these modifications had been adopted singly in other studies. SPR was used to characterize changes in gene expression during synapse formation in embryonic chicken ganglion. Previous studies had reported that the acetylcholine receptors were expressed at low levels, 200 to 1,000 copies per cell, and that this was reduced 1.5- to 2-fold under certain conditions. SPR provided the sensitivity to detect mRNA isolated from the equivalent of one ganglion (100 ng). Thirty-six cDNAs isolated were not homologous to highly abundant genes. Since low-abundance products were isolated, none could be confirmed by Northern analysis, and thus quantitative RT-PCR was used to confirm differential expression (59). A disadvantage of this system is that the primers must be experimentally designed. However, once chosen, they can be used in homologous systems.

Targeted Display

One useful adaptation of DDRT-PCR is targeted display, which allows the identification of members of a gene family (35), genes with specific domains (8, 62), or sequence motifs (12, 31) to be isolated under differential conditions. RT is performed with anchor primers and PCR is performed with degenerate primers homologous to conserved protein domains. To illustrate, Johnson et al. designed a primer with homology to the conserved amino acid sequence of the Cys₂-His₂ class of zinc finger proteins and analyzed a human ovarian cancer cell line (62). They cloned and sequenced 12 fragments; all 12 contained the zinc finger motif and 10 of 12 contained a poly(A) tail; the other 2 probably contained modifications or truncations generated by PCR (62). Normally, novel members of gene families are identified by screening cDNA libraries with degenerate oligonucleotides. More recently, PCR with two degenerate sequences has been used. DDRT-PCR incorporates this as well as providing a method to assess differences in expression in the gene family between different cells or following exposure to environmental stimuli. However, there are limitations to this procedure; (i) the 5′ primer should have 100% homology to the targeted domain; (ii) the primer should not be too degenerate so as to result in an overly complex pattern; and (iii) the targeted domain should be located near the poly(A) tail (62).

DDRT-PCR has also been used to target sequence motifs including AU-rich sequences and CAG repeat sequences (12, 31). Increased numbers of CAG repeats are found in DNA associated with neurological disorders. In this instance, genomic DNA amplified with bioavidin-(CAG)₁₂ discriminated between DNAs isolated from monozygotic twins and from normal and Huntington alleles (12).

Ordered Differential Display (READS)

Prashar and Weissman described a modification of DDRT-PCR originally named ordered differential display but now called READS (GenLogic, Gaithersburg, Md.) (for “restriction endonucleolytic analysis of differentially expressed sequences”) (110, 111). RT is performed with dinucleotide oligo(dT)s with a 20-base heel such that all cDNAs will have a common 3′ end (Fig. 2). These are digested with a restriction enzyme and ligated to a Y-shaped adapter. The Y-shaped adapter consists of three sections: (i) a 3′ overlap which is compatible with the restriction enzyme, (ii) a complementary middle region, and (iii) a noncomplementary 5′ end. In the amplification step, the 3′ primer anneals to the heel and the 5′ primer anneals to the upper branch of the Y adapter. Therefore, only the 3′ ends of restriction digested cDNA are amplified under stringent conditions. In lieu of using an arbitrary primer as in DDRT-PCR, the results of READS are dependent on the enzyme used to digest the cDNA, so that the lengths of the cDNAs differ. Therefore, cDNA prepared from total RNA from a cell population can be systematically resolved into a pattern of 3′-end restriction fragments by using different restriction enzymes. Digesting the cDNA with 12 6-base cutters approaches complete representation of the mammalian genome (110). The reproducibility between experiments is reported to be excellent (111). This technique was further modified by Matz et al. (93). Instead of digesting the cDNA with different restriction enzymes, the adapter-specific primers are extended by two randomly chosen bases at the 3′ end, thereby increasing the number of cDNA species. The double-stranded cDNA is digested with RsaI, and restriction fragments are ligated to adapter-specific primers (93).

FIG. 2.

Schematic of ordered differential display or READS.

Summary

Any situation that suggests altered gene expression can be studied by DDRT-PCR as well as RAP-PCR, SPR, or READS. All these procedures can be adapted to study a variety of cell populations or experimental situations. Both RAP-PCR and DDRT-PCR are straightforward, reproducible, and rapid to perform. However DDRT-PCR focuses on isolation of A-rich 3′ mRNA. Significant redundacies occur during RAP-PCR, but is not biased toward A-rich 3′ mRNA. SPR requires special primers and should be used only if it is known that the differentially expressed genes of interest are expressed at very low levels or that previous DDRT-PCR experiments resulted in no useful information. The limitation of READS is the production of the Y adapters. However, if the production of adapters is not a drawback for the laboratory, READS is a reproducible and relatively rapid method. Finally, targeted display is an excellent method if the aim of the study is to identify differentially expressed genes in a gene family or genes containing specific sequence motifs. For example, the relative expression of a protein family purported to play a role in virulence, i.e., serine proteases, in different host or environmental situations could be assessed. Additional methods to assess differential expression by analyzing cDNA have recently been published in Methods in Enzymology, volume 303, 1999.

POTENTIAL APPLICATIONS

DDRT-PCR has been used to answer many biological questions in mammalian and other nonfungal systems. However, the same biological enigmas exist in medical mycology. Many of the approaches reported for mammalian systems are easily adaptable to fungal systems. A representative but by no means exhaustive sample of potential DDRT-PCR applications is described below. A detailed description of mammalian experiments is outside the scope of this review.

Cell Differentiation

A common application of DDRT-PCR is to identify genes that are regulated during differentiation or development, including hematopoiesis (17, 46, 57, 67, 68, 123), neuronal cell differentiation (13, 64, 153) senescence (79), and adipocyte differentiation (32). Normally, cell populations at different stages of development are compared. The advantage of DDRT-PCR is that several stages of development (47) or numerous cell populations undergoing similar differentiation pathways can be compared at the same time (46). These methods could easily be applied to medically important fungi; for example, RNAs isolated during morphogenesis or different morphogenetic pathways can be compared.

Cell Cycle and Life Stages

DDRT-PCR can be used to study cell and life cycles in order to identify genes which allow the cell or organism to inhabit different environments, initiate pathogenesis, and identify potential antifungal targets. DDRT-PCR was used to study the life cycle of Trypanosoma brucei (91) and the midgut-specific sporogonic cycle of Plasmodium falciparum (10). Three life stages of T. brucei were compared, and 22 DD products were analyzed by SSCP; 13 were analyzed further, and 4 of these were confirmed to be differentially expressed. One gene was identified, ESAG1, that is known to be stage specific, thus validating this procedure (91).

Meiosis was studied by DDRT-PCR in the yeast Saccharomyces cerevisiae (101). RNAs isolated from vegetative and sporulating cells were compared. Several meiosis-specific genes were identified, including genes already known to be induced during meiosis. Since S. cerevisiae is haploid and many mutants are available, the function of the gene identified by DDRT-PCR is more easily characterized. In this instance, the function of SSP1, a gene identified by DDRT-PCR to be expressed midway through meiosis, was determined. Knockout studies suggest the SSP1 functions in a pathway related to control of meiotic nuclear divisions but not related to microtubule or cytoskeletal rearrangements (101).

Some microorganisms, e.g., Neurospora crassa, undergo circadian rhythms. The molecular mechanisms which control this cycle are not well understood. Green and Besharse simultaneously studied eight time points which covered two cycles in Xenopus retinal cells to better understand the mechanistics of the circadian clock (43). This approach can be applied to fungi. In DDRT-PCR, several time points can be studied simultaneously, and thus would afford a more detailed study in N. crassa than is permitted by use of subtractive hybridization, which had been previously used to study circadian rhythms in Neurospora (82).

Cell Activation and Signaling

DDRT-PCR has been applied in many laboratories to identify genes involved in signal cascades. The importance of signaling cascades is to transfer an environmental signal through a series of proteins to eventually activate transcription and subsequent cellular functions. However, rarely are all of the components of the cascade identified. It is often of interest to identify either upstream or downstream targets. These targets have often been identified by comparing cells before and after stimulation of the cell by various factors. These studies have included stimulation of cells with cytokines (117, 130), antibody (42), and/or ligands (17, 64). Among the cell types studied have been T cells (42, 110, 117), macrophages (98), epithelial cells (130), and cell lines (64, 120, 147).

Markers and Mutations

A potential application of DDRT-PCR is the identification of cell surface or diagnostic markers including cell-, tissue-, or pathogen-specific molecules. Specific studies have included identifying metastatic genetic markers in prostate cancer (3), potential probes for rhabdysarcomas (107), a Mycoplasma fermentas species-specific marker (102), cell surface markers for fibroblasts (85), and molecular regional markers in freshwater planaria (93).

Mycoplasma is a significant contaminant in tissue culture laboratories. The detection of Mycoplasma is often difficult due to the inability to culture the organism. DDRT-PCR was used to analyze the M. fermentans 609-infected and -noninfected T-cell lymphoma cell line (102). One intensely expressed product was cloned, ftsZ, the prokaryotic homologue of tubulin. Primers were then designed to establish a species-specific PCR assay (102). This study demonstrated that DDRT-PCR can be used to identify markers for microorganisms as well as markers to discriminate between species. The application to medical mycology is evident, particularly with the great need for improvement in the diagnosis of aspergillosis and candidiasis.

Similar to identifying cell-specific markers is the identification of mutations. Since PCRs can be compared side by side on sequencing gels, small differences in sequence can be easily detected. DDRT-PCR was used to screen for genes involved in DNA repair by comparing Fanconia anemia and normal fibroblasts (108). DD identified deletions in the 3′ UTR of the human α tropomyosin (TPM1) gene. The deletions were 5 and 11 bp long in a tandem repeat, a region that has regulatory effects in cell growth and tumor suppression in transformed cells. It is postulated that these deletions affect mRNA stability (108). It is significant that DDRT-PCR can identify deletions in a single allele.

Drug Resistance and Targets for Drugs

A logical use for DDRT-PCR is to identify new drug targets and the mechanisms of drug resistance or toxicity. The potential applications in medical mycology are readily apparent. One approach is to treat sensitive and resistant cells with a drug and then analyze them by DDRT-PCR (33). The difference between resistant and sensitive cells may also be analyzed in the absence of drug in order to provide insight into why a cell is sensitive or resistant to drug action (72). The effect of drugs can also be examined in vivo by feeding animals the drug and then assessing differential expression in target tissues (25). An obvious adaptation to fungi would be to treat cells with drugs and analyze differential expression. Resistant and sensitive strains could be compared in the same way. The identification of DD products in these type of studies could provide insight into modes of drug action and potential therapeutic targets.

Nutritional and Environmental Stress

DDRT-PCR affords a new tool to identify effects resulting from the absence of specific nutrients (reviewed in reference 50). This is often approached by studying tissues isolated from nutrient-deficient versus nutrient-fed animals, e.g., those fed copper and lithium. Alternatively, two separate strains can be compared, e.g., normal and genetically obese mice. These approaches can easily be adapted to fungi by studying the effects of nutrients on the fungal cell or, alternatively, comparing strains that require a nutrient for growth. For instance, DDRT-PCR could be performed on auxotrophs which were generated by random mutagenesis to ascertain if other genes were mutated or to identify genes involved in metabolic and biosynthetic pathways.

Another beneficial approach of DDRT-PCR would be the ability to identify genes that are expressed only under certain environmental conditions in vivo. In this application, in situ RT would be necessary. Fleming et al. identified previously undescribed and/or cryptic microbial genes that were expressed in the soil but not under normal laboratory conditions. DDRT-PCR was successfully used to transcribe RNA in situ (37).

The stress response has been studied using DDRT-PCR in C. elegans and S. cerevisiae (26, 45, 137). These studies are normally based on observations made with mutant strains. For instance, by studying paraquat-sensitive mutants, it was determined that C. elegans development is inhibited by paraquat at a rate inversely proportional to the life span, of the organism, which emphasized a role for oxidative stress in aging (137). Therefore, DDRT-PCR was used to identify differentially expressed genes during oxidative stress induced by paraquat (137). In the larval stage, upregulation of cDNA products homologous to a detoxification enzyme, glutathione S-transferase, and potential zinc finger proteins, and one of no known homology were identified. Certainly, DDRT-PCR can be applied to identifying proteins involved in the cellular defense against oxidative stress, which is essential in all cells.

Low-Abundance Samples and In Vivo-Expressed Genes

In many instances, only a small number of cells can be harvested and consequently the quantity of mRNA is limiting. Several laboratories have reported adaptations of the original DDRT-PCR technique to address these situations. Some of these were previously described (13, 123, 150, 154). However, a few interesting applications that may be useful to medical mycologists are described below. DDRT-PCR was successfully performed on mRNA isolated from air-dried, snap-frozen mammalian tissue which had been adhered to a cryostat chuck. Tissue was subsequently microdissected to isolate individual cell types (16). DDRT-PCR was also performed on RNA isolated from snap-frozen biopsy specimens of skin tissues (53). These studies demonstrate that RNA can be selectively isolated from different cell populations or tissues. Therefore, these methods may be useful in identifying microbial genes expressed in vivo if clinical samples are collected properly. Mammalian studies have shown that sufficient RNA for DDRT-PCR can be isolated even if very few cells are available. Renner et al. prepared cDNA from only five cells, which had been obtained by magnetic cell sorting (117).

PATHOGENESIS

Differential gene expression is essential in normal development and pathological processes. The identification of differentially expressed genes aids in the understanding of the molecular mechanisms of pathogenesis. DDRT-PCR has been successfully used to identify candidate genes in disease. DDRT-PCR can serve a dual purpose by providing information about the pathogenesis of an illness and providing novel markers for diagnosis and therapy. Numerous laboratories have used this approach to gain insight into cardiovascular disorders (reviewed in reference 149), neurological disorders (reviewed in reference 80), cancer (see, e.g., references 14, 20, 44, 87, 96, and 134), and chronic idiopathic fatiguing illness (114). To study the events which occur during the initiation and pathogenesis of disease, approaches have included comparison of diseased and nondiseased tissues (90, 134), comparison of cells with different invasive capabilities (20), and expression of genes after infection (1, 78, 112, 113). Another approach to the study of pathogenesis is to identify genes expressed in the target tissue. This was applied to pancreatic islet cells to better understand the pathogenesis of diabetes (34). DDRT-PCR was chosen to identify genes present in islet cells but not in pancreatic exocrine cells. Over 2,000 genes were identified with 19 primer pairs, but only 42 were in islet and not exocrine cells (34). While this did not distinguish between genes that were specifically involved in pathogenesis, it quickly narrowed the candidate pool to be screened.

Microbial Pathogenesis (Nonfungal)

It is evident that some genes are activated and others are inactivated once the microorganism invades the host. A better understanding of these events would be efficacious in attempts to treat infections and/or develop new antimicrobial therapies. DDRT-PCR has been used to better define pathogen-host interactions. Studies with nonfungal pathogens have included comparison of virulent and nonvirulent strains (51, 118, 119), studies of the transition from colonization to disease (29, 115, 132), differential expression of pathogen genes (1) or host genes (78, 112, 113) after invasion, effects of drugs on pathogens (138), and pathogen cell death (151).

Virulence factors.

An obvious use of DDRT-PCR is to hunt for virulence factors. This was approached in Mycobacterium tuberculosis by comparing virulent and avirulent mutant strains (118, 119). DDRT-PCR was performed with modified primers to take into account the lack of poly(A) signal in M. tuberculosis genes. The use of nonspecific arbitrary primers resulted in amplification of rRNA and multiple priming of the same RNA species. However, with 32 primer sets, nine DD products were identified and sequenced. A second RT-PCR with gene specific primers confirmed that four of the nine were up regulated in the avirulent strain. All four demonstrated high homology to M. tuberculosis sequences in GenBank. Two of these genes contain polymorphic repetitive elements, including the AT10S repeat and a member of the PPE family. Members of the PPE family are named for a Pro-Pro-Glu near the N terminus and are glycine rich. It is possible that since genes with repetitive sequences are normally less stable in the genome, these genes may play a role in the loss of virulence in M. tuberculosis (119). Another study with the same M. tuberculosis strains but 55 different primer sets revealed 52 differentially expressed bands in the virulent strain and 15 in the avirulent strain. Six bands that were expressed in the virulent but not the avirulent strain were analyzed further. Differential expression was confirmed by dot blot analysis and RT-PCR. The gene fragments which were apparently down regulated in the avirulent strain included those encoding members of the PPE family, the polyketide synthase family (lipid synthesis), and the family containing the 6-kDa early secretory antigenic target (ESAT-6). While homologous to M. tuberculosis protein sequences, three products were of unknown function. The three known gene products have been correlated with immunological or pathogenetic aspects of M. tuberculosis. ESAT-6 is a potent T-cell antigen, PPE proteins are antigenically polymorphic and may cause antigenic variation or inhibit antigenic processing, and polyketide synthase is essential in M. tuberculosis pathogenesis potentially by the synthesis of new lipids (118). It is curious that members of the PPE family were found to be both down regulated (118) and up regulated (119) in the avirulent strain. These were not the same genes according to their GenBank accession numbers. Avirulent and virulent strains were also compared in Leishmania mexicana, and virulent and avirulent specific cDNAs were identified (51).

Drug and environmental influences.

A second approach to understanding pathogenesis is to identify how virulence is attenuated by either drug action or environmental influences of the host. Chloroquine (CQ) resistance is an obstacle in the successful treatment of malaria. To better understand the molecular mechanisms of CQ action, DDRT-PCR was performed with RNA isolated from CQ-sensitive and CQ-resistant strains of Plasmodium falciparum that were treated with CQ and other stresses including high temperature, high partial oxygen pressure, and absence of serum (138). Two DD products were identified that were expressed in CQ-treated CQ-resistant cells. Both products were helicases, an RNA helicase (encoded by Pfhe1) and an ATP-dependent helicase (encoded by Pfhe2). This suggests that strains that can produce high levels of helicases can repair damage inflicted by chloroquines. Therefore, this study identified potential therapeutic modes of action of CQ, protein translation and mitotic control, but not the putative targets of CQ (138).

To assess the gene regulation involved in initiation of infection, researchers have mimicked new host conditions in vitro. Fasciola hepatica has a complex life cycle in the host, and biochemical data indicate that this is correlated with large changes in gene regulation. Immature and adult parasites were compared by DDRT-PCR (115). Eight stage-specific DD products, seven from the immature stage and one from the adult, were chosen for further analysis. The authors found that the ability to confirm differential expression by Northern analysis depended on the intensity of the original signal in the DDRT-PCR gel; however, all bands were confirmed by RT-PCR. The absence of contaminant host cDNA was assessed by analyzing lamb liver RNA. The gene fragments showed no homology to known sequences, but this is not surprising since there is a limited database for helminths (115). Some pathogens may proceed from benign colonization to disease. A better understanding of this event could lead to the identification of virulence genes or novel therapeutic targets. In this light, RNA from Enterococcus faecalis subjected to aerobic and nonaerobic conditions, mimicking colonization and disease states, respectively, were compared. Several products unique to aerobic E. faecalis were identified (132).

Pathogen control was approached in a novel way by Dimopoulos et al. (29). Anopheles gambiae is an important vector for Plasmodium, and the midgut is a potential barrier to malaria transmission. If the parasite can be killed in the insect, transmission to the mammalian host will be reduced. RNA was isolated from female midguts at various times after blood feeding and compared to that isolated from gutless carcasses and bacterially infected larvae. This optimized the identification of genes up regulated in Anopheles specifically in response to Plasmodium and not to infection in general. DD products were run on ethidium bromide-stained agarose gels. Six clones that were found predominantly in or unique to the female midgut were analyzed further. The genes were mapped in situ to polytene chromosomes or chromosomal division-specific cDNA pools. None of the six were expressed in the gutless carcasses, but all six were expressed in the infected gut. Three products, a member of the maltase family, a V-ATPase subunit, and a low-molecular-weight GTP binding protein, were switched off after feeding. The other three were enhanced and included a GAL-lectin, a C-lectin, and a serine protease. The serine protease was also enhanced in the bacterially challenged larvae, suggesting that it is a nonspecific defense mechanism (29).

Gene expression in vivo.

An even more relevant approach to the study of virulence would be to identify genes differentially expressed in vivo. Abu Kwaik and Pederson described a very comprehensive study reporting an early-stage macrophage-induced locus (eml) in Legionella pneumophila identified by DDRT-PCR (1). RNA was isolated from L. pneumophila 4 h after infection of macrophages (4 h p.i.) and compared to that from in vitro-cultured cells. One of the DD bands (eml) showed no homology to the GenBank database. Dot blot analysis confirmed that it was up regulated 4 h p.i. but down regulated by 12 h p.i. eml was not induced by stress or related to growth phase. Transposon insertional mutagenesis demonstrated that eml was essential for virulence, and complementation of the mutated strains restored its ability to invade macrophages. eml was not necessary for initial association with the macrophage but was required for intracellular survival. The mutated strains were also defective in invading amoebae (1). This is a very elegant study that correlated a product identified by DDRT-PCR with virulence.

Gene expression in the host.

In addition to examining genes that are differentially expressed in a pathogen, it may be of interest to identify changes in the host. Gastric cells infected or not infected with Helicobacter pylori were subjected to DDRT-PCR. Of 16 clones, 4 were confirmed to be up regulated in human gastric cells after infection. This study demonstrated that H. pylori did affect gene expression in gastric cells and that DDRT-PCR can be used to study bacterial pathogenesis (152). Two approaches have been used to study host expression induced by Mycobacterium species (78, 112, 113). Human peripheral monocyte-derived naïve THP-1 cells were infected with live or heat-killed Mycobacterium bovis BCG or immunoglobulin G-coated beads. RNA was then isolated from THP-1 cells and compared by DDRT-PCR. One DD product that was up regulated after infection with live or heat-killed M. bovis demonstrated 100% homology to the human ferritin heavy-chain gene product (78). In a separate study, murine peritoneal macrophages were infected with M. tuberculosis. One DD product that was down regulated 6 h p.i. was sequenced and identified to be murine cytochrome c oxidase subunit VIIc (COX VIIc) (112, 113). This product continued to be down regulated for 5 days, the length of the assay. Confirmation that down regulation of COX VIIc was specifically due to M. tuberculosis was assessed by the RNase protection assay and Northern analysis. Limiting-dilution PCR was performed to demonstrate that down regulation was due to M. tuberculosis and not to phagocytosis by comparing RNA isolated from phagocytes after incubation with heat-killed M. tuberculosis latex beads, or BCG. Ex vivo adherent splenocytes from M. tuberculosis-infected mice were also analyzed. It was noted that these macrophages exhibited apoptopic features, and thus COX VIIc may be involved in the first step of programmed cell death due to perturbation of the mitochondrial membrane (113).

Cell death.

An understanding of the mechanisms of apoptosis in pathogens may result in novel antimicrobial therapies or pathogen control. Trypanosoma brucei rhodesiense undergoes apoptosis after stimulation with concanavalin A, and this process is associated with de novo gene expression (151). Therefore, apoptosis may be directly controlled by the induction of several genes that are normally expressed at a very low level; however, it is also possible that housekeeping genes are also important in this event. DD is a rational way to identify genes involved in cell death. Welburn et al. modified DD to study T. b. rhodesiense, and they named this method RADES (for “randomly amplified developmentally expressed sequences”) (151). RT-PCR was performed with one oligo(dT) primer, and single-stranded cDNA was amplified with the oligo(dT) and a 16- to 39-mer containing the miniexon spliced leader sequence of all trypanosome mRNAs. Therefore, double-stranded cDNA served as the template for arbitrary primers in the display reactions. A total of 22 cDNAs were cloned and sequenced; 5 of the cDNAs represented ribosomal genes. It is not known whether ribosomes are involved in cell death or if the increased expression was due to metabolic disturbance of the dying cell. Similar differential expression of ribosomal genes was seen during apoptosis of rat glioma cells (5). One cDNA was homologous to a peptide chain release factor. Three products were related to cell signaling, including TRACK, a protein kinase C receptor, a serine threonine protein kinase, and a leucine zipper protein. One product was a mitochondrial protein, prohibitin, involved with the cell cycle. The involvement of mitochondria in concanavalin A-induced cell death was further revealed by the identification of cDNAs homologous to a mitochondrial RNA-splicing protein, a mitochondrial transporter, and cytochrome c₁. A histone H3 was also identified. Although 13 of these cDNAs are homologous to mammalian or invertebrate genes, only cytochrome c₁ has been implicated in apoptosis (151).

Summary.

The DDRT-PCR studies in the nonfungal microbial systems discussed above were chosen because similar or identical approaches can be applied to medically important fungi. DDRT-PCR was used to identify potential virulence factors by comparing virulent and avirulent strains (118, 119, 151), by identifying differentially expressed genes during the transition from the nonpathogenic to the pathogenic state (115, 132), and by identifying differentially expressed genes in fungi or hosts after invasion (1, 29, 78, 112, 113, 152). Although genes were identified, none of the studies were complete, and further studies are warranted to examine the function of the proteins encoded by the isolated genes. However, DDRT-PCR did identify genes that may play a role in virulence and thus are potential drug targets, including lipid biosynthetic metabolites in M. tuberculosis (118, 119) and mitochondrial proteins in trypanosomes. DDRT-PCR has also shed light on the mode of action of drugs (138). The outcome of these experiments should enhance our understanding of the microbe-host interaction in order to improve diagnosis and antimicrobial therapy.

Biodefense

Although there are a paucity of reports involving DDRT-PCR and medically relevant fungi, there are numerous studies involving plant- and insect-pathogenic fungi. These studies address the same issues relevant to human disease. The first step in understanding pathogenesis is the elucidation of initiation of disease or the events which occur immediately upon pathogen interacting with host. In plant studies, this has been addressed in both symbiotic and parasitic relationships.

Proteases.

Genes have been identified in plant and insect studies which may be relevant to medically important reactions. For instance, a subtilisin-like protease (PrlB) and a metallocarboxypeptidase (MeCPA) were induced in Metarhizium anisopliae after contact with the host (cockroach) cuticle (65, 66). Both proteases were secreted into the host cuticle during invasion, although subtilisin is the major secreted protein. The subtilisin-like protease rapidly degrades cuticle into peptides. The peptides can then be further degraded by MeCPA to amino acids as a source of nutrition for the fungus (65, 66). Other proteases also are differentially expressed in fungal pathogen interactions with plant or insect hosts. Enolase and Hsp30 were slightly up regulated in the white rot fungus after application of chemical stress (58). Up regulation of a cDNA product homologous to a fungal cutinase was identified in Phytophthora capsici after interaction with pepper seedlings (100).

A definite advantage of DDRT-PCR is that differentially expressed genes can be identified from both pathogen and host. A cDNA product homologous to chitinase was found to be induced in a plant root after infection with the fungus Azorhizobium caulinodans ORS571 (40). Benito et al. identified differentially expressed transcripts from both the fungal pathogen Botrytis cinerea and its host, the tomato, that were induced during the plant-fungus interaction (6). Of course, it is important to confirm the source of the DNA. Benito et al. discriminated induced B. cinerea genes from plant defense genes by comparing expression patterns from healthy tomatoes and tomatoes infected with two nonrelated pathogens (6). The initial interactions between host and pathogen have also been studied in normal symbiotic organisms, e.g., mycorrhizial fungi (69, 89).

Environmental signals.

In pathogenic interactions, morphological and physiological changes in cells are induced that result from both perception or production of new signals in the environment. An approach to understanding pathogenesis involves a better understanding of how pathogens react to new environmental stimuli or stresses. These stimuli often correlate with phenotypic and morphological changes in the organism. In fact, if there is a visual change after application of the stimuli, DDRT-PCR is a logical choice to identify differentially expressed genes that may correlate with these changes. For instance, the alga Haematococcus pluvialis loses its flagella, increases its cell volume, produces a tough cell wall, and accumulates pigment in response to stress, e.g., nitrogen deficiency, or high light levels (52). The transition from a swimming green vegetative stage to red nonmotile aplanospores was analyzed by DDRT-PCR (52). A cDNA was identified that was up regulated in developing aplanaspores and whose product was homologous to the serine protease heat shock protein Haematococcus HtrA. It was further demonstrated that three transcripts were evident by Northern analysis, suggesting that this transcript is differentially spliced and is dependent on the developmental stage of the aplanaspore. Previously, HtrA family members were found only in bacteria, but they have since been found in mammalian tissue. Thus, these proteases may also participate in the stress response in other organisms (52). Another way to address stress-inducible genes is to analyze gene expression after wounding. In this manner, lipoxygenase cDNA clones were isolated from soybean (128).

Altered culture conditions can also be a form of stress to which organisms respond. The secondary-metabolite production by filamentous fungi is important for biotechnology purposes. An important class of secondary metabolites consists of the gibberellins, which are used in agriculture and the brewing industry. Therefore, it was of interest to identify genes involved in gibberellin biosynthesis to better identify the conditions needed for optimal production (4). This was accomplished by controlling the levels of gibberellic acid, which allowed repression or derepression of gibberellins during culture of Gibberella fujikuroi. Northern analysis confirmed differential expression of 16 transcripts under the derepressed condition. Of potential interest to the medical mycologist is that genetic information regarding essential pathways could be supplied by DDRT-PCR from a genetically poorly characterized organism (4). Thus, DDRT-PCR can give useful information about pathogenic interactions even from fungi whose genetic systems have not yet been worked out.

Signaling.

A most interesting application of DDRT-PCR was described by Chen et al. (19). Inhibition of fungal virulence in chestnut trees was accomplished by infection of the chestnut blight fungus, Cryphonectria parasitica, with a hypovirus, CHVI-713. To identify the effect of the hypovirus on gene expression and to assess the degree of disruption of the CPG-1 signaling pathway, fungal transcripts from C. parasitica strains that were infected (EP713), noninfected (EP155), or disrupted in CPG-1 (G1310) were compared by DDRT-PCR. The rationale behind studying the role of the CPG-1 signaling pathways was that strain G1310 is also hypovirulent. Eighty primer pairs displayed 139 DD products with a fourfold difference in expression and over 400 if differences in excess of twofold were included. Many observed changes in fungal gene transcripts were attributed to disruption of the CPG-1 signal pathway. Because 65% of DD products identified in EP713 were also observed in the G1310 display, hypovirulence might be linked with the CPG-1 pathway. CPG-1 encodes a GTP-binding protein α subunit. Mammalian G₁α subunits reduce cyclic AMP (cAMP) levels by inhibiting adenylylcyclase. Since CPG-1 levels were reduced in hypovirulent strains, virus infection may result in elevated cAMP levels. This was demonstrated in both the virus-infected and cpg-1 hypovirulent strains. The authors further verified the role of cAMP by using two products identified by DDRT-PCR as reporter gene products. EP155 (noninfected) was treated with compounds that elevate intracellular cAMP levels, and the expression of the two products was up regulated, as observed in hypovirulent strains. Therefore, DDRT-PCR demonstrated that hypovirus infections induced a reproducible change in fungal gene expression, i.e., that the levels of transcripts increased rather than decreased in the infected strain, and that a large number of the changes were due to the disruption of the CPG-1 signaling pathway (19).

DDRT-PCR AND MEDICALLY IMPORTANT FUNGI

Relatively few reports of applications of DDRT-PCR to medically important fungi have been published (23, 55, 97, 99, 156). The majority of studies have looked at differential expression of fungal genes after exposure to the host, both in vitro (23, 99) and in vivo (156).

Histoplasma capsulatum

Histoplasma capsulatum is a successful intracellular parasite. It is theorized that specific genes are activated which allow it to evade the natural killing abilities of the macrophage. Colonna-Romano et al. studied the expression of genes induced during attachment and internalization of H. capsulatum in 7-day-old murine bone-marrow derived macrophages (23). It was preliminarily determined that the majority of yeast cells had adhered after 15 min and were internalized after 1 h. Therefore, DDRT-PCR was performed with RNA isolated from H. capsulatum 15 min and 1 h after infection. Steps were first taken to eliminate contamination by murine RNA. Three treatments of macrophages isolated from mice 1 h after infection were assessed using mouse-specific probes: (i) lysis with double-distilled H₂O, (ii) lysis with 0.2% sodium dodecyl sulfate (SDS) in the presence of a DNase-RNase cocktail, and (iii) lysis with 0.5% SDS. Only lysis with 0.5% SDS eliminated contamination of murine nuclear or mitochondrial RNA. Total RNA was isolated from H. capsulatum mycelia, free-living yeasts, and yeasts after infection of macrophages (23). DDRT-PCR analysis was performed using 13 arbitrary primers and 3 1-base-anchored primers (77). Each primer combination displayed an average of 60 bands, and therefore approximately 2,300 different mRNA species were screened. At least 20 cDNAs were subcloned, and six colonies were sequenced per product. Of 20 products tested, 6 were confirmed to be differentially expressed. Three products gave no signal, and the change in expression for the remainder was not significant. Four clones, pHc12, pHc16, pHc2, and pHc23, were significantly up regulated during the first hour of infection. One clone, pHc20, was significantly up regulated at 15 min (attachment) but was reduced to normal levels after 1 h (internalization). pYs-1 was significantly down regulated after 1 h of incubation with the macrophage. Only one clone, pHc12 demonstrated significant homology to any of the genes in the database (23). This clone has been further characterized (109). The 508-bp cDNA insert was approximately 30% homologous to the p100 family (23). This insert was used to isolate the full gene from an H. capsulatum cDNA library and fully sequenced (pHc12-E). The entire sequence and hydrophobic cluster analysis indicates that pHC12 is a member of the p100 family and is designated Hcp100. Northern analysis confirmed that its expression increases approximately fivefold after infection of the macrophage. p100 proteins consist of four similar domains that are related to the staphylococcal nuclease protein (SN). However, p100 proteins lack the catalytic site, and therefore it is proposed that p100 binds DNA without cleaving it. The human p100 protein coactivates gene expression mediated by the EBNA2 protein. Therefore, Hcp100 may also act as a coactivator during infection of the macrophage (109).

Candida albicans

An in vitro method was also used to identify differentially expressed genes in C. albicans after exposure to host cells or ligands (99). The adherence of Candida to epithelial and endothelial cells is a crucial step in the initiation of mucocutaneous and systemic infections. Environmental conditions which favor adherence are well documented, but the genes which are activated or inactivated are not known. In this study, adherence assays were performed under conditions that were defined in vitro but bore relevance to human pathogenesis. The initial interactions between C. albicans and HET-1A, an immortalized, nontumorigenic esophageal epithelial cell line derived from normal human tissue, were studied. HET-1A cells were grown on Matrigel basement membrane matrix (Collaborative Biomedical Products), which is a solubilized extract from an Engelberth-Holm-Swann mouse tumor rich in extracellular matrix proteins. Candida will bind not only to HET-1A but also to exposed matrix. Adherence assays were performed with Candida grown in yeast extract-peptone-dextrose (YPD)–500 mM galactose. It was apparent that Candida adhered to both matrix and HET-1A cells. Additionally, yeast cells germinated in both situations. After 90 min, matrix-adherent, plastic-adherent, and nonadherent Candida cells were isolated. DDRT-PCR was performed with RNA isolated under various conditions, including Candida grown in YPG to mid-exponential stage (starting culture), Candida that had been incubated in Hanks balanced salt solution for 90 min at 37°C with matrix but had attached to the plastic, Candida adhered to HET-1A and matrix, and Candida adhered to matrix only. A small subset of differentially expressed bands (four bands) were subcloned and sequenced, and differential expression was confirmed by Northern dot blot analysis (Fig. 1D). None of these four bands demonstrated significant homology to sequences in the data bank. However, three clones from each product were sequenced, and one of the products resulted in the isolation of two mRNA species. Both dot blot and Northern analysis confirmed that the second mRNA species was down regulated after interaction with matrix and HET-1A matrix (Fig. 1C). The sequence of this clone showed remarkable homology to the 3′ end of the BMH genes of S. cerevisiae and Schizosaccharomyces pombe. The BMH genes encode proteins belonging to the 14-3-3 family. Members of this family have been isolated from several organisms and assigned a range of functions which may be involved in signaling and/or regulatory pathways. Additionally, they are essential for pseudohypha formation in S. cerevisiae. Interestingly, a 14-3-3 protein was also identified by DD in soybean after infection with Pseudomonas (131). In this instance, the product was identified in the host and the transcript was up regulated. However, many conditions were different, and it will be interesting to see if this gene is involved globally in host-pathogen interactions.

While in vitro models afford cleaner systems, to truly understand virulence interactions it would be desirable to identify differentially expressed genes in vivo. This was addressed using C. albicans and the rat model of oral candidiasis (156). RNA was isolated from C. albicans grown in vitro and from the oral cavity of the rat. RNAs from separate infections were not pooled but were compared separately. This model is rather complex in that there are two potential sources of contaminating RNA, rat and bacteria. Bacterial contamination was reduced but not eliminated, since RT-PCR was performed with oligo(dT) anchor primers. Rat epithelial cells were lysed with buffer containing 0.1% Triton X-100. However, this may not have been sufficient, as demonstrated for murine macrophage cells in the H. capsulatum model (23). Forty-five primer combinations using 15 arbitrary primers and 3 1-base-anchored primers have been analyzed; 70 differentially expressed fragments were observed and 35 of these were further characterized. These fragments were chosen because they represented genes expressed only in vivo, were detected in duplicate PCRs, and were observed in RNA isolated from two separate infections. The 35 clones were analyzed by Southern hybridization against Candida and rat DNA. A total of 7 fragments demonstrated homology to C. albicans DNA, 2 demonstrated homology to rat DNA, and 26 demonstrated homology to neither. The seven Candida-specific products were further analyzed. One fragment demonstrated homology to the ALS gene family which is involved in agglutination and potentially in adherence. A second fragment was homologous to the bacterial gidB gene. The remaining five fragments demonstrated no homology to sequences in the database. Differential expression of the ALS fragment was confirmed by RT-PCR. The gidB-related gene gave no signal after 30 cycles of PCR but could be detected by Southern analysis. The other five fragments were false-positives. This study demonstrates the potential of using DDRT-PCR in in vivo models even in the presence of contaminating cell populations (156).

Another way to approach virulence mechanisms is to compare virulent and avirulent isogenic strains. This was done by Mirbod et al. to elucidate the role of phospholipase in the pathogenesis of candidiasis (97). Two C. albicans strains, potent (virulent) and weak (avirulent) phospholipase producers, were compared by DDRT-PCR with five primer sets. Ten differentially expressed bands were directly sequenced. Only one showed homology to the database, and this demonstrated 90% homology to TIF-2. The full gene was isolated and cloned. Northern analysis confirmed that its expression was increased 1.4-fold in the potent producer strain. If TIF-2 in C. albicans is functionally as well as sequentially homologous to yeast TIF-1 and mammalian eIF-4A proteins, the up regulation of TIF-2 may lead to induction of proteins related to virulence, perhaps phospholipases.

A fourth study with C. albicans involved the identification of genes induced by stress (55). C. albicans cells were treated or not treated with the toxic heavy metal cadmium, and DDRT-PCR was performed to isolate genes expressed only in the cadmium-treated cells. Two products that were induced were further characterized (the CIP1 and CIP2 gene products) (55, 104). Northern analysis confirmed that the products were absent (the CIP1 product) or expressed at very low levels (the CIP2 product) in normal cells and that expression was specifically induced by cadmium and not other heavy metals or stresses including copper, mercury, lead, manganese, and zinc or heat shock treatment. CIP2 is also activated by the oxidant diamide. CIP1 encodes a 32-kDa hydrophobic protein with limited homology to isoflavone reductase of plants (55). These studies demonstrate that DDRT-PCR can be used to study the stress response in medically important fungi. Additionally, it is conceivable that several stress responses could be studied simultaneously in order to discern how many stress response pathways may exist in fungi.

RECENT DEVELOPMENTS AND PROSPECTS

Since DDRT-PCR was first described, newer methods have been introduced that have both advantages and disadvantages over DDRT-PCR; these are discussed below.

Subtractive Cloning

Subtractive cloning is a relatively new method that modernizes subtractive hybridization. Many of the limitations of subtractive hybridization have been eliminated, including quantity of RNA, speed to finished product, and amplification of rare transcripts. The major disadvantage, however, is that like differential and subtractive hybridization, only two cell populations can be compared. cDNA is synthesized from 0.5 to 2.0 μg of poly(A)⁺ RNA isolated from the two populations of interest (this is still more RNA than is required for DDRT-PCR). The reference cDNA is called the driver, and the cDNA where the differentially expressed transcripts are to be found is called the tester. The tester cDNA population is split, and different adapters are ligated to each of the two pools of tester cDNA; adapters are not added to the driver cDNA. Common cDNAs between driver and tester populations anneal in a hybridization reaction in the presence of excess driver cDNA. This results in a single-stranded DNA population enriched for differentially expressed products. In a second hybridization reaction, all cDNAs are mixed and annealing occurs only between single-stranded DNAs, since the pools are not first denatured. PCR amplification is performed with primers complementary to each of the adapter sequences, so only hybrids with two adapter sequences will be amplified (15, 28). An obvious application of this method is the identification of downstream target genes in signaling cascades by comparing wild-type and mutant strains. Carulli et al. used subtractive cloning to identify genes regulated by the transcriptional factor PAX-6, a gene involved in development in humans and mice. Genes were isolated that were present in wild-type but not mutant cells (15). The transcription factor can also be placed under the control of an inducible promoter, and the induced population (tester) can be compared to the noninduced population (driver). Although this method has been applied to C. albicans knockout strains, no published reports are available. This method may be more practical than DDRT-PCR for the applications described above. However, comparisons are unidirectional (i.e., differential expression in only one of the two populations can be assessed), and it is not applicable for studies requiring comparison of more than two cell populations.

Serial Analysis of Gene Expression

Serial analysis of gene expression (SAGE) is basically a rapid method to obtain short sequences of all expressed genes in a given situation (143). A unique sequence tag is generated for each transcript in the cell of interest. Double-stranded cDNA is digested with a 4-base restriction enzyme, linkers are ligated, and the population is amplified. The linkers contain a restriction site for a type II restriction enzyme which cuts the DNA 20 nucleotides away from the recognition site. The cDNA pool is then digested with this enzyme, which results in a pool of tagged cDNA of 13 to 20 bp defined by the original 4-base cutter and the adjacent DNA sequence. The sequence tags are ligated, and the final result is a library of clones where each clone contains 20 or more tagged genes. The expression profile is obtained by sequencing each clone. Advantages of this method include high throughput and the ability to compare SAGE data from numerous populations (143). However, preparation of SAGE libraries can be technically difficult, and bioinformatic tools are required for analysis of data.

cDNA Micorarrays

Genome sequencing projects have resulted in complete or substantial sequencing of several genomes, which has allowed the manufacture of cDNA microarrays that contain most if not all the expressed genes in a genome (71, 129). Therefore, with the rapid sequencing of C. albicans and the initiation of sequencing projects for Aspergillus fumigatus and Cryptococcus neoformans, it is likely that cDNA microarrays will be available in the future. This will most probably be the preferred method. However, its accessibility to the normal laboratory in the near future is doubtful. cDNA clones are arrayed in duplicate on glass slides at a high density (≥1,000 clones/cm²). Poly(A)⁺ RNA is isolated from two populations and labeled by incorporation of two separate fluorescently labeled dCTPs (129). In this manner, the slide can be simultaneously hybridized with both probes and scanned with a laser. The advantages of the cDNA microarray include the ability to compare the same set of genes under many experimental conditions and the ability to quickly analyze a large set of clones (96 clones/slide). Disadvantages include the lack of a large set of clones of known genes to array, the need for large amounts of RNA to prepare probes, and the need for special equipment to prepare slides and analyze results. To address these issues, a method based on the microarray is described which is more accessible to the average laboratory. This method is an enzyme-linked colorimetric detection assay, and cDNAs are spotted on filter membranes (21). The results of the microarray-colorimetric detection system were comparable to those of Northern dot blot analysis when the products of Northern analysis were detected by chemiluminescence. The advantages of this system is the simultaneous display of thousands of gene and the ability to perform analysis with a scanner or digital camera that can be bought from consumer electronic stores (20, 21). As in the microarray, the preparation of the filters is the time-consuming and technically limiting step.

CONCLUSION

It is hoped that this review accentuates the potential areas where DDRT-PCR can be applied to medical mycology and shows that it is a suitable method to approach these issues. The optimizations and adaptations made to DDRT-PCR now make it a valid option when it is known that gene transcription plays an important role in the biological process under study. Since DDRT-PCR was first described, many laboratories have modified the protocol in order to simplify reactions and reduce false-positive results. Virtual Northern analysis and hybridization filter arrays simplify and increase the output for confirming differentially expressed products even when RNA is limiting. Many modifications are time-consuming and may not be time effective or cost-effective for the purpose of the study. Additionally, some adaptations are specific for the organism or parameters of the experiment, and therefore the expertise and subject matter of the individual laboratory should dictate the experimental protocol. Although there are very few published reports of DDRT-PCR with medically important fungi, the potential of this technique is well documented by studies in mammalian systems, nonfungal microbe-host interactions, and plant pathogens. DDRT-PCR has been successfully applied to a wide range of biological activities which are easily applicable to problems in medical mycology, including cell differentiation, drug resistance, and host-pathogen interactions. It must be emphasized that DDRT-PCR only identifies potentially differentially expressed genes and that confirmation of the expression and the role of the gene in the biological process is still necessary. The applications of these techniques should lead to an improved understanding of the host-pathogen interaction, identification of antifungal targets, and new treatment strategies.