Molecular Techniques for Genus and Species Determination of Fungi From Fresh and Paraffin-Embedded Formalin-Fixed Tissue in the Revised EORTC/MSGERC Definitions of Invasive Fungal Infection

Abstract

The EORTC/MSGERC have revised the definitions for proven, probable, and possible fungal diseases. The tissue diagnosis subcommittee was tasked with determining how and when species can be determined from tissue in the absence of culture. The subcommittee reached a consensus decision that polymerase chain reaction (PCR) from tissue, but not immunohistochemistry or in situ hybridization, can be used for genus or species determination under the new EORTC/MSGERC guidelines, but only when fungal elements are identified by histology. Fungal elements seen in tissue samples by histopathology and identified by PCR followed by sequencing should fulfill the definition of a proven fungal infection, identified to genus/species, even in the absence of culture. This summary discusses the issues that were deliberated by the subcommittee to reach the consensus decision and outlines the criteria a laboratory should follow in order to produce data that meet the EORTC/MSGERC definitions.

Correct diagnosis of fungal genus and species from histopathology is a vanishing art, and the diagnostic accuracy of identification from traditional histopathology alone, even to the genus, is generally below 80% [1, 2]. This lack of diagnostic accuracy can have many consequences, ranging from inappropriate antifungal therapy to exclusion from a clinical trial. In addition, there are problems with fungal diagnosis from tissue; often, the histopathology detects a fungus but the culture is negative, or is simply not possible, for instance, because the specimen has been placed in formalin and not sent for culture. A study from the MD Anderson Cancer Center showed that culture was positive for only 30% of their histopathology positive fungal cases, and a study from 2 Spanish hospitals found that only 56% of their histopathology-positive fungal cases were also culture positive [3, 4]. In this scenario, it is difficult to determine the genus, let alone the species of the offending fungus.

The EORTC/MSGERC have revised the definitions for “proven,” “probable,” and “possible” fungal infection [5]. It has been 10 years since the definitions were last revised, during which time there has been substantial improvement in our methods and understanding of diagnosis of fungal infections from tissue, especially using molecular techniques. The tissue subcommittee of this group determined that the changes in technology that have come about since the publication of the definitions in 2008 warranted an update to the definitions with regard to identification of fungi from tissue [6]. The 2008 definitions stated the following regarding the use of molecular methods for detecting fungi in tissue: “By contrast, molecular methods of detecting fungi in clinical specimens, such as [polymerase chain reaction] PCR, were not included in the definitions because there is as yet no standard, and none of the techniques has been clinically validated”…“We had hoped that nucleic acid–detection tests, such as PCR, would have improved enough to incorporate the results of these tests into the definitions. However, standardization and validation have not yet been attained for these platforms” [5]. In the intervening period since these definitions were published, there has been substantial work on the development and application of PCR to amplify fungal DNA from both formalin-fixed paraffin-embedded (FFPE) and fresh tissue [4, 7-31]. However, the success rate of the various laboratory-developed protocols varies greatly due to many methodologic variables such as the method of DNA extraction, inoculum, sequencing targets, primer selection, specimen variables (open biopsy vs fine needle aspiration, fresh vs fixed tissue), providing mixed results in the literature. For example, Buitrago and colleagues found that PCR of fungal DNA from tissue shown by histopathology to contain fungi was 89% successful, although the species determined by DNA sequencing was discordant with the culture results in 13% of cases [4]. In contrast, a study from Japan showed that PCR of fungal DNA from FFPE tissue shown histopathologically to contain fungi was only 23% successful [15]. The apparent discrepancy in success between studies would indicate caution should be exercised because successful amplification of fungal DNA from tissue is dependent upon many factors, including the amount of fungi in the tissue, the amount of tissue available, the amount of time the tissue has been fixed in formalin, and whether the formalin was buffered or unbuffered.

One of the inherent difficulties in determining the success rate of fungal species identification from tissue is the suboptimal success rate of the “gold standard,” which is the growth of a fungus from tissue [3]. If molecular detection and identification of fungi is more sensitive than the gold standard, as was seen with the recent fungal meningitis outbreak [32, 33], then the true sensitivity and specificity of the assay cannot be determined. A good example is the previously mentioned study from Spain, where histologic evidence of fungal infection was used as the standard, culture was only 56% sensitive [4]. Although tissue PCR was successful, there was no way to prove that the species identification was correct because the gold standard was less sensitive, leaving no isolate for comparison with many PCR positive cases.

The US outbreak of Exserohilum meningitis presented a good opportunity to directly compare culture with histologic evidence of infection and amplification of nucleic acids. In the study by Ritter and colleagues, patient FFPE tissues were stained using a panfungal polyclonal antibody directed to cell wall carbohydrates [33]. Tissue scrolls from both the histochemically positive and the histochemically negative cases were sent for PCR analysis. PCR from FFPE was not as sensitive as histopathological staining for the detection of fungi [33]. There were no cases where fungal nucleic acids were amplified in the absence of histological staining. By contrast, there were 16 cases where fungi could be detected by histochemical staining but no fungal DNA could be amplified. It is unclear if these results are fungus (Exserohilum) or tissue (meninges, brain)-specific or could be extrapolated to the plethora of other fungi causing invasive disease.

The 2008 definitions did allow for the designation of a proven infection based upon direct microscopic analysis of tissue. However, the guidelines did not allow for a species designation based on such analysis. Correct diagnosis of fungal species from histopathological specimens remains difficult [34]. Sangoi and colleagues compared histopathologic designation of genera with the cultured species identification [1]. Although the correct genera were described in 79% of cases, major errors were encountered, such as describing Rhizopus or Scedosporium as Aspergillus and describing Histoplasma as Candida, pointing to the limitations of any designation beyond “yeast or hyphae seen” [2, 7]. In a similar study, infections identified by histopathology as mucormycosis (MCR) were subsequently identified by PCR to be Aspergillus and Scedosporium, and conversely, cases identified as aspergillosis were subsequently identified as Fusarium or Rhizopus by PCR [29].

Immunohistochemical stains are commercially available for the Mucormycetes, Aspergillus species, and Candida species [35-37]. Although the specificity and sensitivity are high when used in culture-proven cases, the specificities of these stains have not yet been fully defined against a wide variety of other closely related species, especially among the other hyalohyphomycetes. Encouraging data with Mucormycetes, Candida species, and Aspergillus species are comprehensive and robust [35-39]. In a recent study, Jung et al investigated the accuracy of histomorphologic diagnosis of MCR and invasive aspergillosis (IA), using fungus-specific immunohistochemistry (IHC) in patients with proven/probable MCR or IA that had FFPE tissues available [38]. In 7 proven cases of MCR, the sensitivity and specificity of MCR IHC were 100% and 100%, respectively. In 8 proven cases of IA, the sensitivity and specificity of aspergillosis IHC staining were 87% and 100%, respectively. For probable cases, the sensitivity and specificity are much lower. In the absence of fungal culture results, the IHC tests seem helpful in differentiating between IA and MCR. However, such approaches remain problematic in the uncommon infections caused by organisms other than Aspergillus, Mucormycete spp, and Candida, because an abundance of negative control staining remains elusive. Until a comprehensive clinical study that provides more negative controls is published, specific immunohistochemical staining for genus identification cannot be fully endorsed for inclusion in the EORTC/MSGERC definitions. Similarly, difficulties are encountered with in situ hybridization techniques [30, 34, 40]. Some of the published probes have shown cross-reactivity and there are no validated protocols available. Although promising, more work needs to be done before it can be recommended.

Metagenomic next-generation sequence analysis is emerging as a powerful tool because it could be used to detect any fungal pathogen, even when intact fungal cells are not present in the tissue [41]. However, this tool is still under clinical validation and one of the few studies to look at fungi in tissue showed low sensitivity and specificity [42]. In addition, because of the high sensitivity of the assay, there is not yet a firm grasp on the clinical implications of the identification of an organism, especially when histopathology and culture confirmation are absent [43]. As is the case with in situ hybridization and gene-targeted PCR techniques, rare fungi have few reference genomes and curated public databases, which complicates the analysis [44].

There are additional problems outside of the tissue-based molecular assays themselves that contribute to the difficulty of making a correct diagnosis. The first is that formalin treatment of tissue affects the integrity of the DNA and leads to shorter fragments, which may prevent amplification of the target sequence or may only allow the amplification of a fragment that is not long enough for species identification [45, 46]. Another is the paucity of sequences in publicly available validated and annotated fungal genome libraries [47, 48]. New fungal species are described on a regular basis but genome libraries are not consistently updated and annotated, often leaving sequences misidentified [49–51]. Although pan-fungal primers recognize most fungal taxa, they can also promote amplification of contaminating fungal DNA. Quantitative real-time PCR might obviate the problem of contamination, but its specificity limits the number of species that can be reasonably detected. Despite all of these difficulties, significant advances have been made. In a prospective, blinded study comparing histopathology, PCR and culture of aspergillosis and mucormycosis in FFPE tissue specimens, PCR identification fared quite well and even allowed the detection of mixed infection in 2 cases [52]. In a retrospective analysis to compare histopathology, culture, and PCR of MCR, PCR was positive for 10 of 12 culture confirmed cases and the sequence matched the cultured organism in nine of those cases [53]. In the 15 culture negative cases, the PCR was positive and a Mucormycete sequence was obtained for 12 cases.

A final problem related to assay performance is that many reagents used to process tissue, such as lyticase, proteinase K, and even the PCR master-mix, can be contaminated with fungal DNA during the manufacturing process [54–56]. In addition, fungal spores are ubiquitous in the indoor environment and the paraffin used for making blocks is not kept sterile in the pathology laboratory. The ubiquitous nature of fungal spores and fungal DNA greatly increases the chance of spurious amplification and false-positive PCR results.

Although the use of PCR for the detection of fungi from tissue has greatly advanced, the 2 issues outlined in the 2008 EORTC/MSG guidelines remain: there is no standardized technique for detection of fungal nucleic acid from tissue and there have only been a few validation studies to show that the obtained results are correct. The protocols used for this procedure are laboratory specific; there are no International Organization for Standards or Clinical and Laboratory Standards Institute guidelines to assure that protocols and interpretations are reproducible between laboratories [57, 58]. However, given that a validated consensus protocol is not forthcoming, we offer guidance on essential criteria while allowing some deviation in protocol, similar to those developed for quantitative real-time PCR [59]. For the purposes of the EORTC/MSGERC definitions for the identification of fungal infections, laboratory-validated protocols should rely on a common set of rules that may suffice until an international standard can be established. The subcommittee for tissue diagnosis recommended that PCR for species identification of fungi from tissue be adopted with the following rules and caveats:

1. PCR is appropriate for detection of invasive fungal infection from tissue samples only when fungal elements or structures have been detected by histopathology. PCR in tissue is not recommended in cases where fungal staining is negative.

2. Laboratories performing PCR-based identification of fungi from tissue must have a unidirectional workflow with strict separation of DNA-extraction, PCR preparation, amplification, and detection of PCR products. These processes should take place either in separate rooms or in separate isolation cabinets.

3. Quality control procedures are highly recommended to check for potential contaminating fungal DNA.

4. Primers should be panfungal, targeting the fungal barcoding sequences of either the ITS region or the D1/D2 region of ribosomal DNA [60]. Alternatively, other ribosomal or mitochondrial genes that target unique, species- or genus-specific proteins or antigens that have been fully validated must be used [52, 53, 61].

5. Every PCR reaction must include separate negative and positive amplification controls accompanied by a separate or tandem PCR reaction targeting a human housekeeping gene, like beta-globin, to indicate successful fungal DNA extraction and rule out inhibitors.

6. Every PCR product must be sequenced for identification—neither size of the PCR product nor hybridization such as in real-time PCRs, are regarded as sufficient for identification.

7. The length of sequence/PCR product should not be less than 150 base pairs, including primer-binding regions.

8. For species identification, a homology of the PCR product with the sequence in the database should be ≥98%. Use of a quality-controlled database, like the new database of the International Society for Human and Animal Mycology is strongly recommended [60].

9. When the species identification matches more than 1 species in the database to the same percentage, only the genus name should be used. Care should be given to periodic changes in fungal nomenclature [62].

10. The genus and species identification from PCR amplification should be consistent with the key histological features of the organism in tissue.

11. This test should be performed only in reference centers or high-volume centers that meet these requirements and not in small volume, community hospitals where volume might be low and expertise cannot be maintained or requirements met [63].

Although such tissue-based molecular assays may not provide a cost-effective tool for routine clinical diagnosis and management of fungal infections [64], in cases where there is histopathological evidence of a fungal infection without confirmation by culture, molecular tools may assist in identification of the etiological agent if the outlined criteria are met. Thus, fungal elements seen in tissue samples by histopathology and identified by PCR followed by sequencing should fulfill the definition of a proven fungal infection, even in the absence of culture. Again, because of the ubiquitous nature of fungal DNA and fungal spores, amplification of specific fungal DNA without histopathologic evidence of fungal infection is insufficient to confidently prove an invasive fungal infection. In cases of endemic mycoses where the only proof of infection is histopathology, molecular tools are recommended as proof of the diagnosis, especially in laboratories that have limited experience with these fungi [65, 66]. Laboratories that intend to perform fungal identification by PCR should first perform both a validation of their methodology as well as a validation of the DNA sequence database that they intend to use.

Because of the promise of advancement using newer molecular techniques, there is an urgent need for a multicenter study using prospective evaluation of a consensus protocol or set of guidelines [67]. Because this type of study is not immediately forthcoming, the criteria developed should be used as a framework for the correct use of PCR for the amplification of fungal DNA from tissue in individual laboratories.

Notes

Financial support. D. P. K. acknowledges the Texas 4000 Distinguished Professorship for Cancer Research and the NIH-NCI Cancer Center CORE Support grant no. 16672. D. P. K. reports research support from Astellas Pharma and honoraria for lectures from Gilead, Merck, and United Medical. He has served as a consultant for Astellas Pharma, Cidara, Merck, Amplyx Pharmaceuticals, and Mayne. M. C. E. has, in the past years, received support from Astellas Pharma, bioMerieux, Gilead Sciences, Merck Sharp, Dohme, Pfizer, Schering Plough, Soria Melguizo SA, Ferrer International, F2G, Amplyx, Basilea, and Cidara y Scyntex. He is a founding partner and holds shares of Micología Molecular SL.

Supplement sponsorship. This supplement is sponsored by the MSGERC/EORTC.

Disclaimer. The findings and conclusions in this manuscript are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.