A high density COX1 barcode oligonucleotide array for identification and detection of species of Penicillium subgenus Penicillium

Abstract

We developed a COX1 barcode oligonucleotide array based on 358 sequences, including 58 known and two new species of Penicillium subgenus Penicillium, and 12 allied species. The array was robotically spotted at near microarray density on membranes. Species and clade-specific oligonucleotides were selected using the computer programs SigOli and Array Designer. Robotic spotting allowed 768 spots with duplicate sets of perfect match and the corresponding mismatch and positive control oligonucleotides, to be printed on 2 × 6 cm² nylon membranes. The array was validated with hybridizations between the array and digoxigenin (DIG)-labelled COX1 polymerase chain reaction amplicons from 70 pure DNA samples, and directly from environmental samples (cheese and plants) without culturing. DNA hybridization conditions were optimized, but undesired cross-reactions were detected frequently, reflecting the relatively high sequence similarity of the COX1 gene among Penicillium species. Approximately 60% of the perfect match oligonucleotides were rejected because of low specificity and 76 delivered useful group-specific or species-specific reactions and could be used for detecting certain species of Penicillium in environmental samples. In practice, the presence of weak signals on arrays exposed to amplicons from environmental samples, which could have represented weak detections or weak cross reactions, made interpretation difficult for over half of the oligonucleotides. DNA regions with very few single nucleotide polymorphisms or lacking insertions/deletions among closely related species are not ideal for oligonucleotide-based diagnostics, and supplementing the COX1-based array with oligonucleotides derived from additional genes would result in a more robust hierarchical identification system.

Introduction

The Barcode of Life aims to provide a large, freely accessible DNA database, enabling effective and efficient identification of living organisms based on sequences of standardized mitochondrial and/or nuclear gene regions. The COX1 gene (accepted as the ‘default’ Barcode of Life region) was first proposed as the principal barcode region for animal species, and a 648-bp region is used for species-level identification (Hebert et al. 2003a, b). The use of COX1 as a fungal barcode was expected to be problematic, because for a broad range of fungal taxonomic groups, (i) universal primer pairs for the COX1 region are unavailable, and (ii) multiple copies and introns with considerable length variation have been detected occasionally (Seifert et al. 2007; Nguyen & Seifert 2008; Gilmore et al. 2009). The resolution of COX1 varies significantly among species complexes and therefore, may not allow species to be identified in some fungal groups (Geiser et al. 2007; Seifert 2008, 2009). Currently, mycologists are preparing a formal proposal to the Consortium for the Barcode of Life (CBOL) to designate ITS as the fungal barcode region. The Barcode of Life Data Systems (http://www.barcodinglife.org, Ratnasingham & Hebert 2007) has become the first online database accepting ITS sequences as barcodes (Seifert 2008).

Penicillium subgenus Penicillium (Trichocomaceae, Eurotiales, Eurotiomycetes, Ascomycota) includes 58 known species, several of great economic importance in food and the medical sector. Traditional identification of species involves culturing on diagnostic media and microscopic examination performed by experienced taxonomists (Frisvad & Samson 2004). In subgenus Penicillium, only a few gene regions have been sequenced for most or all species. The nuclear ribosomal internal transcribed spacer (ITS) and other nuclear rDNA regions examined have low variability among morphologically distinct species, and therefore lack the resolution to distinguish species (Skouboe et al. 1999; Peterson 2000). However, partial β-tubulin (BenA) sequences yield a more resolved phylogeny (Samson et al. 2004) and were referred to as the ‘most promising’ gene for a one-gene phylogeny in this subgenus (Frisvad & Samson 2004). Despite this, several species complexes had species sharing identical or highly similar sequences, e.g. P. camemberti, P. caseifulvum and P. commune. Seifert et al. (2007), presented a complete COX1 data set for subgenus Penicillium, based on multiple strains per species. They demonstrated that COX1 was easily amplified and aligned well because of the absence of indels. Approximately two-thirds of the species had unique COX1 sequences, and divergences between species are comparable to ITS but less than in BenA.

DNA arrays can be spotted on various supporting platforms, such as silicon slides and nylon membranes. Because of their low cost, higher sensitivity and reusability, membrane-based arrays are a more practical choice for biodiversity studies, such as determing species profiles from environmental complexes (Summerbell et al. 2005). Membrane-based DNA arrays can be printed either manually (Lévesque et al. 1998; Fessehaie et al. 2003; Tambong et al. 2006) or robotically (Lappin et al. 2001; Tung et al. 2006; Leaw et al. 2007; Zahariev et al. 2009). They combine hybridization of labelled polymerase chain reaction (PCR) products with membrane-bound taxon-specific oligonucleotides, and are powerful, sensitive tools for detection of species using molecular techniques. This technology, originally named reverse dot blot hybridization (RDBH), was first developed for detecting human gene mutations (Zhang et al. 1991; Cuppens et al. 1992). It has been adapted for detecting fungi and fungus-like organisms in complex samples without culturing (Lévesque et al. 1998; Tambong et al. 2006; Zhang et al. 2007), and virus, bacteria and nematode species from environmental samples (Ehrmann et al. 1994; Uehara et al. 1999; Wu et al. 2007). Based on taxonomically complete data sets for specific genes, species- or clade-specific oligonucleotides can be designed, synthesized, and then immobilized on a solid membrane support. Target gene regions are then amplified and labelled (e.g. using mainly digoxigenin) from total DNA derived either from pure cultures or from environmental samples. The profile of species in a sample is determined by hybridizing a population of labelled amplicons with the DNA array. Chemiluminescent signals from the positive reactions are captured either by exposure to film or by a digital camera in a dark room, and then analysed by computer. Oligonucleotides specific for any kind of organism can be included on one membrane (which can be expanded into a higher density array, as in the current study), allowing the development of substrate- or host-specific assays with broad taxonomic coverage.

The objective of the current study was to exploit the availability of a complete COX1 data set to design an oligonucleotide hybridization array for detection and identification of species in Penicillium subgenus Penicillium. Because of the lack of indels that would offer obvious sites for designing oligonucleotides that do not cross-react, we decided to add a corresponding mismatching oligonucleotide with a single base polymorphism for every specific oligonucleotide, in an attempt to enhance data interpretation and the overall specificity of the assay. This array was printed using newly available robotic technology that allowed spotting of a membrane-based array at near microarray density. This study represents an application of full-length barcodes to develop oligonucleotide barcodes, or microcodes (Summerbell et al. 2005), which could reflect the eventual usage of barcode data in laboratory-on-a-chip or hand-held barcoding devices envisioned by the founders of DNA barcoding.

Materials and methods

Design of barcode oligonucleotides

The development of this COX1 DNA array was based on a data set of 358 sequences at the 5′-end of the COX1 gene, for 60 species of Penicillium subgenus Penicillium, including 58 described species and two new species, Penicillium sp. 1 (2 strains, IBT 14452, 20395) and Penicillium sp. 2 (3 strains, IBT 16643, 17769, 22760) (D. Overy et al. in preparation). Two computer programs were used to develop barcode oligonucleotides. SigOli (http://www.lifeintel.com, Zahariev et al. 2009) identifies the location of polymorphisms that are species or clade-specific. Array Designer version 1.1 (Premier Biosoft International) uses the polymorphism location to generate hybridization oligonucleotides of specific melting temperature while minimizing hairpins and dimers. Seifert & Lévesque (2004) described in detail how to perform sequential analyses to obtain oligonucleotides with various level of specificity. In the Penicillium COX1 neighbour-joining (NJ) tree (Seifert et al. 2007), as many as 16 nested clades exist between the first branch and the highest resolution species branch. Separate analyses were run using SigOli to identify putative clade-specific oligonucleotides at each level of resolution down to species. All accepted oligonucleotides, defined as perfect match (PM) oligonucleotides, were subjected to blast searches against both GenBank and the original Penicillium COX1 database, to verify their specificity within the target clade(s). The complete data set and a detailed explanation of the analysis and design steps are available at http://www.lifeintel.com.

When possible, one or more oligonucleotides were designed for each species or supraspecific clade. All oligonucleotides for each target species/clade were designated with a common number, and differentiated by a letter following the number. For example, Pen_COX1_63a and Pen_COX1_63b were both designed for P. crustosum around a polymorphism in base 136, but with slightly different sequence profiles. For most PM oligonucleotides, a mismatch (MM) oligonucleotide from a closely related species was also selected. Each MM oligonucleotide had the same code as the corresponding PM oligonucleotide, followed by ‘M’, e.g. Pen_COX1_63aM and 63 bM (Appendix S4, Supporting information). The hybridization signals generated by MM oligonucleotides were used as references for corresponding PMs in hybridizations with target amplicon(s). Only PMs that produced stronger signals than their corresponding MMs were selected for further testing.

Spotting of DNA array at near-microarray density

A total of 377 oligonucleotides synthesized with 5′-end amine-modification (Sigma-Aldrich) were selected for the COX1 array, including PM and MM oligonucleotides, along with positive control oligonucleotides (Table 1). Positive PCR controls included forward and reverse primers, namely PenF1 (5′-GACAAGAAAGGTGATTTTTATCTTC-3′) and AspR1 (5′-GGTAATGATAATAATAATAATACAGCTG-3′), respectively, were used to amplify the COX1 barcode regions of Penicillium subgenus Penicillium. The short oligonucleotide ST1 (5′-CACGGCGATTTCGCAGTTTA-3′) was used as hybridization control (Tambong et al. 2006; Robideau et al. 2008). Amine-modified oligonucleotides were diluted from 200 µm stock to 40 µm with 0.5 m sodium hydrogen carbonate buffer (4.2 g NaHCO₃, 99 mL HPLC water, 1 mL of 0.004% bromophenol blue, pH 8.4) in a sterile 384-well microplate (Whatman Inc.) and were in a specific pattern to group oligonucleotides designed for the same species/clade.

Table 1.  Oligonucleotides designed in the current study, including perfect-match (PM) oligonucleotides, and positive control oligonucleotides. Oligonucleotide origins are relative to the COX1 sequences represented by Seifert et al. (2007)

Oligonucleotide code	Sequence with SNPs (5′–3′)	Oligonucleotide origin	Accession no.	Result
* Accession numbers refer to strains from CBS (Centraalbureau voor Schimmelcultures), and from the Technical University of Denmark Biocentrum collection.
Pen_COX1_1	GGWTACAATCTCAYAGTGGWCC			Group oligonucleotide
Pen_COX1_2a	TTGCATTATTTTCAGGATTAATAGGTACT	P. atramentosum	CBS 291.48	Species oligonucleotide
Pen_COX1_2b	ATTTTCAGGATTAATAGGTACTGCATTT	P. atramentosum	CBS 291.48	Species oligonucleotide
Pen_COX1_2c	TTCAGGATTAATAGGTACTGCATTTTC	P. atramentosum	CBS 291.48	Species oligonucleotide
Pen_COX1_3a	CATTAATTGGAGGATTTGGTAATTTCTTAT	P. bialowiezense	CBS 227.28	Rejected
Pen_COX1_3b	GGTTGAACTTTATATCCACCTTTATCA	P. bialowiezense	CBS 227.28	Rejected
Pen_COX1_3c	TAAGTGGTATAAGTAGTATGTTAGGTTCTA	P. bialowiezense	CBS 227.28	Rejected
Pen_COX1_4a	TAATTCTAGTCTAATCAATACTGAAAATGC	P. atramentosum	CBS 109588	Subspecies oligonucleotide
Pen_COX1_4b	GCACATGCAATTCTGATGATTTTC	P. atramentosum	CBS 109588	Species oligonucleotide
Pen_COX1_4c	CATTAATAGGGGGTTTTGGTAATTTCT	P. atramentosum	CBS 109588	Rejected
Pen_COX1_4d	ATAAGAAATTACCAAAACCCCCTATTAA	P. atramentosum	CBS 109588	Species oligonucleotide
Pen_COX1_5a	CTGCATTTTCAGTATTAATTAGACTAGAATT	P. atramentosum	CBS 291.48	Rejected
Pen_COX1_5b	TGCATTTTCAGTATTAATTAGACTAGAATT	P. atramentosum	CBS 291.48	Species oligonucleotide
Pen_COX1_5c	GCACATGCTATTATGATGATTTTCTTC	P. atramentosum	CBS 291.48	Species oligonucleotide
Pen_COX1_5d	ATGCTATTATGATGATTTTCTTCATGGT	P. atramentosum	CBS 291.48	Group oligonucleotide
Pen_COX1_6a	GTTTGCTTTATTTTCTGGTTTAATAGGTA	P. bialowiezense	CBS 227.28	Rejected
Pen_COX1_6b	GTTGAACTTTATATCCACCTTTATCAGG	P. bialowiezense	CBS 227.28	Rejected
Pen_COX1_6c	TGAGAAGTCCAGGTATTCGTTTAC	P. bialowiezense	CBS 227.28	Rejected
Pen_COX1_6d	AAGTCCAGGTATTCGTTTACACAA	P. bialowiezense	CBS 227.28	Rejected
Pen_COX1_6e	AGGTATTCGTTTACACAAATTAGCTTTA	P. bialowiezense	CBS 227.28	Rejected
Pen_COX1_7a	TGGTTTAATCGGTACAGCATTTTC	P. olsonii	CBS 833.88	Species oligonucleotide
Pen_COX1_7b	ATCGGTACAGCATTTTCTGTTTTAAT	P. olsonii	CBS 833.88	Species oligonucleotide
Pen_COX1_7c	GATTAGAATTAGCAGCACCTGGT	P. olsonii	CBS 833.88	Rejected
Pen_COX1_7d	GAGCTGGTACTGGTTGAACTTTAT	P. olsonii	CBS 833.88	Rejected
Pen_COX1_7e	CTTTATCTGGTATACAATCTCATAGTGG	P. olsonii	CBS 833.88	Species oligonucleotide
Pen_COX1_7f	GTTAGGTTCTATGAATTTCATTACAACTATT	P. olsonii	CBS 833.88	Species oligonucleotide
Pen_COX1_7g	GCAGCACCTGGTGTTCAATATA	P. olsonii	CBS 833.88	Species oligonucleotide
Pen_COX1_8a	ACTAGAATTATCAGGTCCAGGTGT	P. atramentosum	CBS 291.48	Species oligonucleotide
Pen_COX1_8b	ACTGATTATCTGATATATATTGAACACCTG	P. atramentosum	CBS 291.48	Rejected
Pen_COX1_9	TTTCCTAGATTAAATAACATAAGTTTCTGATT	P. atramentosum	CBS 194.88	Rejected
Pen_COX1_10a	TGTTTGCGTTATTTTCTGGTTTAGT	P. formosanum	CBS 211.92	Rejected
Pen_COX1_10b	TAATGTTTGCGTTATTTTCTGGTTTAG	P. formosanum	CBS 211.92	Rejected
Pen_COX1_10c	CAGAAAATAACGCAAACATTAAATATAAAGT	P. formosanum	CBS 211.92	Rejected
Pen_COX1_10d	TTTAATGTTTGCGTTATTTTCTGGTTTA	P. formosanum	CBS 211.92	Rejected
Pen_COX1_10e	GGTGTTCAATATATATCTGACAATCAATTATAT	P. formosanum	CBS 211.92	Rejected
Pen_COX1_10f	AATGCCATATCAGGACCCCC	P. formosanum	CBS 211.92	Rejected
Pen_COX1_10g	CTATTTTCGGTTTACATTTAAGTGGTATAA	P. formosanum	CBS 211.92	Rejected
Pen_COX1_10h	TGAATTTTATAACTACTATTTTAAATATGAGAAGTC	P. formosanum	CBS 211.92	Rejected
Pen_COX1_10i	GCTATGAATTTTATAACTACTATTTTAAATATGAGA	P. formosanum	CBS 211.92	Rejected
Pen_COX1_11a	TTGGTACAATGTATTTAATGTTTGCTTTAT	P. bialowiezense	CBS 227.28	Rejected
Pen_COX1_11b	GATTAGAATTAGCTGCACCTGGT	P. bialowiezense	CBS 227.28	Species oligonucleotide
Pen_COX1_11c	ATTAGCTGCACCTGGTGTACA	P. bialowiezense	CBS 227.28	Species oligonucleotide
Pen_COX1_11d	TAGCTGCACCTGGTGTACAAT	P. bialowiezense	CBS 227.28	Species oligonucleotide
Pen_COX1_11e	ACGCTATTTTAATGATTTTCTTTATGGTTA	P. bialowiezense	CBS 227.28	Species oligonucleotide
Pen_COX1_11f	ATGCCAGCATTAATTGGAGGATT	P. bialowiezense	CBS 227.28	Species oligonucleotide
Pen_COX1_11g	GCCGGTACAGGTTGAACTTTATA	P. bialowiezense	CBS 227.28	Rejected
Pen_COX1_12a	AGCTTTTTCTGTTTTAATTAGATTAGAGTTAG	P. brevicompactum	CBS 110067	Species oligonucleotide
Pen_COX1_12b	TGGTGCAGCTAACTCTAATCTAATT	P. brevicompactum	CBS 110067	Species oligonucleotide
Pen_COX1_12c	CCGTTATTAGTAGGTGGTCCTGA	P. brevicompactum	CBS 110067	Species oligonucleotide
Pen_COX1_12d	AGTAGTATGTTAGGTTCTATGAATTTTATTAC	P. brevicompactum	CBS 110067	Rejected
Pen_COX1_13	TGGTGTTCAATATATATCAGATAATCAAGTAT	P. olsonii	CBS 312.97	Rejected
Pen_COX1_14	CCTGGTGTTCAATATATATCTGATAATCAA	P. olsonii	CBS 833.88	Rejected
Pen_COX1_15a	AGATATTGGAACTTTATATTTAATGTTTGCA	Eu. osmophilum	CBS 462.72	Group oligonucleotide
Pen_COX1_15b	GTTTAGTTGGTACAGCGTTTTCTG	Eu. osmophilum	CBS 462.72	Species oligonucleotide
Pen_COX1_15c	AAAATAAACTAGGTACTAATAATCAAAAACTTATATT	Eu. osmophilum	CBS 462.72	Rejected
Pen_COX1_15d	AGATTTAGCAATTTTTGGTTTACACTTAA	Eu. osmophilum	CBS 462.72	Species oligonucleotide
Pen_COX1_15e	GTAGATTTAGCAATTTTTGGTTTACACT	Eu. osmophilum	CBS 462.72	Rejected
Pen_COX1_16	ATAACTACTATATTAAATATGAGAAGTCCTGG	Penicillium sp. 1	IBT 14452	Rejected
Pen_COX1_17a	TATAATAGTATAATAACAGCGCATGCTATAT	Eu. crustaceum	CBS 581.67	Rejected
Pen_COX1_17b	ACAGCGCATGCTATATTGATGAT	Eu. crustaceum	CBS 581.67	Rejected
Pen_COX1_18a	GTTTGCATTATTCTCTGGTTTAATTGG	P. carneum	CBS 449.78	Group oligonucleotide
Pen_COX1_18b	GGTTTAATTGGTACAGCATTTTCAGT	P. carneum	CBS 449.78	Group oligonucleotide
Pen_COX1_18c	CTCTGGTTTAATTGGTACAGCATTT	P. carneum	CBS 449.78	Group oligonucleotide
Pen_COX1_18d	CAGCATTTTCAGTATTAATTAGATTAGAGTTA	P. carneum	CBS 449.78	Group oligonucleotide
Pen_COX1_19a	GGTTTAGTTGGTACAGCCTTTTC	P. glandicola	CBS 333.48	Species oligonucleotide
Pen_COX1_19b	AATTAGATTAGAATTATCAGGACCAGGT	P. glandicola	CBS 333.48	Species oligonucleotide
Pen_COX1_19c	AATTATCAGGACCAGGTGTTCAATATA	P. glandicola	CBS 333.48	Species oligonucleotide
Pen_COX1_20	AGAGTTATCTGGTCCAGGTGTAC	P. albocoremium	CBS 472.84	Rejected
Pen_COX1_21a	GTTGGAACAGCCTTTTCAGTTTT	P. gladioli	CBS 332.48	Species oligonucleotide
Pen_COX1_21b	TAGTTGGAACAGCCTTTTCAGTT	P. gladioli	CBS 332.48	Species oligonucleotide
Pen_COX1_22a	TTTGCATTATTTTCAGGTTTAGTTGG	P. flavigenum	CBS 419.89	Group oligonucleotide
Pen_COX1_22b	TATTTTCAGGTTTAGTTGGAACAGC	Penicillium sp. 1	IBT 14452	Group oligonucleotide
Pen_COX1_23	GATATTGGAACTTTATACTTAATGTTTGCA	P. persicinum	CBS 111235	Rejected
Pen_COX1_24	AGTATAATAACAGCGCATGCCAT	P. carneum	CBS 449.78	Group oligonucleotide
Pen_COX1_25a	AAGATATCGGTACTTTATACTTAATGTTTG	P. paneum	CBS 465.95	Species oligonucleotide
Pen_COX1_25b	TATCGGTACTTTATACTTAATGTTTGCA	P. paneum	CBS 465.95	Rejected
Pen_COX1_25c	TTTTCATGGTTATGCCAGCATTAATA	P. paneum	CBS 465.95	Species oligonucleotide
Pen_COX1_25d	GTTATGCCAGCATTAATAGGAGGT	P. paneum	CBS 465.95	Rejected
Pen_COX1_26a	ACCCCCTACTAATAATGGTAATAAGAAA	P. glandicola	CBS 498.75	Rejected
Pen_COX1_26b	ATCTGGACCCCCTACTAATAATGG	P. glandicola	CBS 498.75	Rejected
Pen_COX1_27a	AAATGCTAAAGACATAGGAACTTTATATTTAA	P. glandicola	CBS 333.48	Rejected
Pen_COX1_27b	AGACATAGGAACTTTATATTTAATGTTTGC	P. glandicola	CBS 333.48	Species oligonucleotide
Pen_COX1_28a	AGGTGTACAATATATATCTGACAATCAATTA	P. albocoremium	CBS 472.84	Rejected
Pen_COX1_28b	ATTATACTTAATGTTTGCTCTATTTTCTGG	Penicillium sp. 2	IBT 16643	Rejected
Pen_COX1_29a	AAAGATATAGGTACACTATATTTAATGTTTGC	P. vulpinum	CBS 305.63	Species oligonucleotide
Pen_COX1_29b	TTAGATTAGAATTATCAGGTCCAGGTG	P. vulpinum	CBS 305.63	Rejected
Pen_COX1_29c	GAATTATCAGGTCCAGGTGTTCAA	P. vulpinum	CBS 305.63	Rejected
Pen_COX1_29d	CATTATATCCACCATTATCTGGAATACAA	P. vulpinum	CBS 305.63	Rejected
Pen_COX1_29e	ACCATTATCTGGAATACAATCTCACA	P. vulpinum	CBS 305.63	Rejected
Pen_COX1_29f	CCACCATTATCTGGAATACAATCTC	P. vulpinum	CBS 305.63	Rejected
Pen_COX1_29g	TATCTGGAATACAATCTCACAGTGG	P. vulpinum	CBS 110772	Rejected
Pen_COX1_29h	ATCTGGAATACAATCTCACAGTGG	P. vulpinum	CBS 305.63	Rejected
Pen_COX1_30a	CCTGGACCTGATAACTCTAATCTAAT	P. digitatum	CBS 101026	Rejected
Pen_COX1_30b	TTGAACACCTGGACCTGATAACT	P. expansum	CBS 281.97	Rejected
Pen_COX1_30c	AGTTATCAGGTCCAGGTGTTCA	P. digitatum	CBS 101026	Group oligonucleotide
Pen_COX1_31a	TCCTAGTTTATTTTTATTTATCTTCTCAGCT	P. sclerotigenum	CBS 101033	Rejected
Pen_COX1_31b	CCTAACATACTACTAACACCACTTAAAT	P. sclerotigenum	CBS 101033	Rejected
Pen_COX1_32a	TAATAGTATAATAACAGCACACGCTATATTA	P. aethiopicum	CBS 484.84	Rejected
Pen_COX1_32b	GCACACGCTATATTAATGATTTTCTTC	P. aethiopicum	CBS 484.84	Rejected
Pen_COX1_33	GTGGTCCTAGTGTAGACTTAGCT	P. chrysogenum	CBS 412.69	Group oligonucleotide
Pen_COX1_34	CAGGCATAACCATAAAAAAAATCATCAA	P. carneum	CBS 466.95	Rejected
Pen_COX1_35a	GGTCCAGGTGTACAATATATCTCAG	P. clavigerum	CBS 255.94	Group oligonucleotide
Pen_COX1_35b	TGGTCCAGGTGTACAATATATCTCA	P. clavigerum	CBS 255.94	Group oligonucleotide
Pen_COX1_35c	GGTCCAGGTGTACAATATATCTCAG	P. clavigerum	CBS 255.94	Group oligonucleotide
Pen_COX1_36	CTGATATATACTGTACACCTGGACC	P. nordicum	CBS 112573	Species oligonucleotide
Pen_COX1_37	AATGCTAAAGATATAGGTACATTATACTTAATG	Penicillium sp. 2	IBT 16643	Group oligonucleotide
Pen_COX1_38a	AACGCTGTACCAATTAAACCAGA	P. dipodomyicola	CBS 173.87	Rejected
Pen_COX1_38b	TGGTTTAATTGGTACAGCGTTTTC	P. dipodomyicola	CBS 173.87	Group oligonucleotide
Pen_COX1_38c	TAATTGGTACAGCGTTTTCAGTTTTA	P. dipodomyicola	CBS 173.87	Rejected
Pen_COX1_38d	TGGTACAGCGTTTTCAGTTTTAATT	P. dipodomyicola	CBS 173.87	Group oligonucleotide
Pen_COX1_38e	CTGATAATGGAGGATATAATGTTCAACC	P. dipodomyicola	CBS 173.87	Rejected
Pen_COX1_39a	GTATAATAACAGCACATGCTATCTTGA	P. vulpinum	CBS 110772	Group oligonucleotide
Pen_COX1_39b	GGACCAAGTGTTGATTTAGCTATTTT	P. vulpinum	CBS 110772	Group oligonucleotide
Pen_COX1_40a	AATAATGCAAACATTAAATATAAAGTACCTATATC	P. ulaiense	CBS 262.94	Rejected
Pen_COX1_40b	CTAGTTTATTTTTATTTATATTCTCGGCTACT	P. italicum	CBS 489.84	Rejected
Pen_COX1_41a	ACTGATTATCTGATATATATTGAACACCTG	Eu. crustaceum	CBS 458.72	Rejected
Pen_COX1_41b	GTTCAATATATATCAGATAATCAGTTATATAATAGTATAAT	Eu. crustaceum	CBS 458.72	Rejected
Pen_COX1_42a	AGATATTGGAACTTTATACTTAATGTTTGC	P. aethiopicum	CBS 484.84	Rejected
Pen_COX1_42b	CCAGGAGTACAATATATATCTGATAATCAAT	P. aethiopicum	CBS 484.84	Species oligonucleotide
Pen_COX1_42c	ATATATCTGATAATCAATTATATAATAGTATAATAACAGC	P. aethiopicum	CBS 484.84	Rejected
Pen_COX1_42d	TAACAGCACACGCTATATTAATGATTT	P. aethiopicum	CBS 484.84	Group oligonucleotide
Pen_COX1_42e	AATATTAGTTTCTGATTATTAGTTCCTAGTTTATT	P. aethiopicum	CBS 484.84	Rejected
Pen_COX1_42f	CATTCCCTAGATTAAATAATATTAGTTTCTGAT	P. aethiopicum	CBS 484.84	Rejected
Pen_COX1_42g	ATATTAGTTTCTGATTATTAGTTCCTAGTTTATTT	P. aethiopicum	CBS 484.84	Rejected
Pen_COX1_42h	CACTAGGACCACTATGAGATTGTATT	P. aethiopicum	CBS 484.84	Rejected
Pen_COX1_42i	ATACAATCTCATAGTGGTCCTAGTG	P. aethiopicum	CBS 484.84	Rejected
Pen_COX1_43a	ATATAATAGTATAATAACAGCACACGCTATAT	P. dipodomyis	CBS 170.87	Rejected
Pen_COX1_43b	ATAAAGAAAATCATTAATATAGCGTGTGC	P. dipodomyis	CBS 170.87	Rejected
Pen_COX1_43c	GTTGAACTTTATATCCACCATTATCAGG	P. dipodomyis	CBS 170.87	Rejected
Pen_COX1_44	TATTAATGATTTTCTTTATGGTTATGCCTG	P. chrysogenum	CBS 412.69	Rejected
Pen_COX1_45	GACCTAGTGTAGATTTAGCCATATTTG	P. crustosum	CBS 101025	Rejected
Pen_COX1_46a	AATGCTAAAGATATAGGTACATTATACTTAATG	P. concentricum	CBS 191.88	Rejected
Pen_COX1_46b	AATGTTTGCACTATTCTCTGGTTTAG	P. concentricum	CBS 191.88	Species oligonucleotide
Pen_COX1_46c	TATCAGACAATCAGTTATATAATAGTATAATTACAG	P. concentricum	CBS 191.88	Rejected
Pen_COX1_47a	AAATGCGAAAGATATAGGTACATTATACTT	P. dipodomyicola	CBS 173.87	Species oligonucleotide
Pen_COX1_47b	GATTATTAGTTCCTAGTTTATTCTTATTTATATTCTC	P. dipodomyicola	CBS 173.87	Rejected
Pen_COX1_47c	TTATTAGTTCCTAGTTTATTCTTATTTATATTCTCA	P. dipodomyicola	CBS 173.87	Rejected
Pen_COX1_47d	TCCTAGTTTATTCTTATTTATATTCTCAGCTA	P. dipodomyicola	CBS 173.87	Rejected
Pen_COX1_47e	GTTCCTAGTTTATTCTTATTTATATTCTCAGC	P. dipodomyicola	CBS 173.87	Rejected
Pen_COX1_48	AATGCGAAAGATATTGGTACATTATACT	P. griseofulvum	CBS 485.84	Rejected
Pen_COX1_49	ACCTGGACCTGATAACTCTAATCTA	P. expansum	CBS 281.97	Rejected
Pen_COX1_50a	TTAGTTGGTACAGCGTTTTCAGT	P. expansum	CBS 481.84	Species oligonucleotide
Pen_COX1_50b	CTGGTTTAGTTGGTACAGCGTT	P. expansum	CBS 481.84	Group oligonucleotide
Pen_COX1_50c	TTGGTACAGCGTTTTCAGTTTTAATT	P. expansum	CBS 481.84	Rejected
Pen_COX1_51	CTGATAACGGTGGATATAATGTTCAAC	P. chrysogenum	CBS 412.69	Rejected
Pen_COX1_52a	GTCCAGGTGTTCAATATATATCTGATAAT	P. flavigenum	CBS 419.89	Rejected
Pen_COX1_52b	GCAACTATAGAAAATGGAGCTGGTA	P. flavigenum	CBS 419.89	Rejected
Pen_COX1_53	ATTAGAATTATCTGGTCCAGGTGTAC	P. clavigerum	CBS 255.94	Rejected
Pen_COX1_54a	TAGTTGGAGGTCCTGATATGGC	P. venetum	CBS 405.92	Rejected
Pen_COX1_54b	AGTTGGAGGTCCTGATATGGC	P. venetum	CBS 405.92	Rejected
Pen_COX1_55a	ACAATATATCTCAGACAATCAGTTATACAAT	P. crustosum	CBS 101025	Group oligonucleotide
Pen_COX1_55b	AATATATCTCAGACAATCAGTTATACAATAGT	P. crustosum	CBS 101025	Species oligonucleotide
Pen_COX1_56a	TAATGCCAAAGATATAGGAACTTTATATTTAATAT	P. digitatum	CBS 101026	Species oligonucleotide
Pen_COX1_56b	TTTAATATTTGCATTATTTTCTGGTTTAATTG	P. digitatum	CBS 101026	Species oligonucleotide
Pen_COX1_56c	TATCAGACAATCAGTTATATAATAGTATTATTACAG	P. digitatum	CBS 101026	Rejected
Pen_COX1_56d	CATCAATATAGCATGTGCTGTAATAATAC	P. digitatum	CBS 101026	Species oligonucleotide
Pen_COX1_56e	GCACATGCTATATTGATGATTTTCTTC	P. digitatum	CBS 101026	Rejected
Pen_COX1_57a	AAGATATAGGAACTTTATACTTAATGTTTGC	P. marinum	CBS 109549	Species oligonucleotide
Pen_COX1_57b	CGAAAGATATAGGAACTTTATACTTAATGTTT	P. marinum	CBS 109549	Species oligonucleotide
Pen_COX1_57c	ATATAGGAACTTTATACTTAATGTTTGCATTAT	P. marinum	CBS 109549	Rejected
Pen_COX1_58	GTGTTCAATATATATCTGACAATCAATTATATAATAG	P. expansum	CBS 110402	Rejected
Pen_COX1_59a	ATTAATTGGAGGTTTTGGAAATTTCTTATT	P. nalgiovense	CBS 318.92	Rejected
Pen_COX1_59b	GCATTAATTGGAGGTTTTGGAAATTT	P. nalgiovense	CBS 318.92	Rejected
Pen_COX1_60a	TGATTATCAGATATATATTGTACACCTGGA	P. hordei	CBS 704.68	Rejected
Pen_COX1_60b	CCAGGTGTACAATATATATCTGATAATCA	P. hordei	CBS 704.68	Species oligonucleotide
Pen_COX1_60c	TCTGATAATCAGTTATATAATAGTATAATAACAGC	P. hordei	CBS 704.68	Rejected
Pen_COX1_61	CAAACATTAAGTACAAAGTACCAATATCTT	P. tricolor	CBS 637.93	Species oligonucleotide
Pen_COX1_62a	GGTTGAACATTATATCCACCTTTATCA	P. clavigerum	CBS 189.89	Rejected
Pen_COX1_62b	TGAACATTATATCCACCTTTATCAGGT	P. clavigerum	CBS 189.89	Rejected
Pen_COX1_62c	CTGATAATGGAGGATATAATGTTCAACC	P. venetum	CBS 405.92	Rejected
Pen_COX1_63a	AATATATCTCAGACAATCAGTTATACAATAGT	P. crustosum	CBS 101025	Species oligonucleotide
Pen_COX1_63b	ACAATATATCTCAGACAATCAGTTATACAAT	P. crustosum	CBS 101025	Species oligonucleotide
Pen_COX1_64a	ATAATAGTATAATAACAGCACATGCTATCTT	P. echinulatum	CBS 337.59	Rejected
Pen_COX1_64b	GTATAATAACAGCACATGCTATCTTGA	P. echinulatum	CBS 337.59	species oligonucleotide
Pen_COX1_65a	TATCTGATATATATTGTACACCTGGACC	P. coprobium	CBS 561.9	Rejected
Pen_COX1_65b	TACAGGTTGAACTTTATATCCTCCATT	P. coprobium	CBS 561.9	Rejected
Pen_COX1_66	TATGTAAACGTATACCTGGACTTCTC	P. coprobium	CBS 110761	Rejected
Pen_COX1_67a	ACTTAATGTTTGCTCTATTTTCTGGTT	Penicillium sp. 2	IBT 16643	Species oligonucleotide
Pen_COX1_67b	GCACACGCTATCTTAATGATTTTCT	Penicillium sp. 2	IBT 16643	Rejected
Pen_COX1_67c	TGCATTAATCGGTGGTTTTGGT	Penicillium sp. 2	IBT 16643	Species oligonucleotide
Pen_COX1_67d	TAATCGGTGGTTTTGGTAATTTCTTAT	Penicillium sp. 2	IBT 16643	Species oligonucleotide
Pen_COX1_67e	AATCGGTGGTTTTGGTAATTTCTTAT	Penicillium sp. 2	IBT 16643	Species oligonucleotide
ST1	CACGGCGATTTCGCAGTTTA	Synthetic oligonucleotide		Positve control
ST3	TAAACTGCGAAATCGCCGTG	Synthetic oligonucleotide

The MicroGrid Compact robot (Fig. 1) (BioRobotics) was used to print the array on Immunodyne ABC membrane (PALL Europe Ltd). A membrane holder (L × W × H: 29.50 cm × 26 cm × 0.34 cm, Fig. 1) was fabricated to raise the platform holding the plastic sheet high enough to allow the spotting pins of the robot to reach the membrane. The amount of oligonucleotide buffer spotted onto the membrane by a split pin was controlled by ‘Soft Touch’ parameters within the robot configuration. The robot was mounted with BioRobotics MicroSpot Pins with 100 µm tip diameter (‘2500 pins’, BioRobotics), for spotting oligonucleotides. The robot was then fit with a 6 × 4 Pin Tool Configuration and 4 × 4 Pin Array Size, and programmed to visit the 384-well plate twice, using TAS Application Suite version 2.2.2.6 (BioRobotics). Each 2 × 6 cm² membrane was spotted with two 16 × 24 arrays arranged vertically, including 768 spots with duplicate sets of blank, positive controls and all selected PM and MM oligonucleotides. Bromophenol blue in the spotting solution allowed us to estimate this amount by spot size and intensity of blue spots. To compare spot properties of arrays printed robotically and manually [using solid pin multiblot replicators (V&P Scientific), Fessehaie et al. 2003], we used Image-Pro Plus (version 6.0.0.260, MediaCybernetics, Inc.) to measure spot size/spacing and Genepix Pro version 6.0 (Molecular Devices) to analyse the intensity of the blue spots.

Figure 1. Photographs of the membrane holder (A, B) and the MicroGrid Compact robot printing the COX1 array (C). The membrane holder was fabricated from two stainless steel sheets screwed together (A) and the Immunodyne ABC sheet was pre-cut to fit the size (B). The edges of the membrane sheet were taped to the membrane holder to lock its position and ensure flatness. This holder was placed on to the original stage to raise the height to the level of the platform that holding the plastic sheet. The Pin Tool and Pin Array were set with a 6 × 4 and a 4 × 4 configuration, respectively.

Ninety-four membranes were printed. They were inactivated in blocking solution [2 × SSC (0.33 m sodium chloride, 0.1 m sodium citrate, pH 7.0), 0.5% (g/mL) skim milk powder, 0.05% (w/w) Tween-20] with agitation for 15 min and then stored in 2× SSC at 4 °C until use.

DNA amplification with DIG labelling

The COX1 regions of all DNA samples were amplified and labelled using a Mastercycler ep gradient S thermocycler (Eppendorf). PCRs were performed in 20-µL reaction mixtures containing 1 µL of total genomic DNA (0.1–10 ng); 2 µL of 10× Titanium. Taq buffer (Clontech Laboratories Inc.); 1.5 µL of 2mm DIG-dNTPs (2 mm of each dATP, dCTP, and dGTP; 1.9 mm dTTP; and 0.1 mm dig-dUTP); 0.5 µL of each of forward (PenF1) and reverse (AspR1) primers (20 µm); 0.4 µL of Titanium (BioCan Scientific) Taq (9:1 ratio); and 14.1 µL of sterile HPLC water. Amplifications were performed with the following thermal profile: 95 °C for 3 min followed by 40 cycles of 95 °C for 1 min, 56 °C for 45 s, and 72 °C for 90 s, followed by a final extension at 72 °C for 10 min. Amplification reactions were stopped and held at 10 °C. Primers PenF1 and AspR1 amplified a fragment of the COX1 of 545 bp. The concentration of PCR products was estimated using a Quant-iT dsDNA High-Sensitivity Assay Kit (Invitrogen Canada Inc.). Synthetic oligonucleotide ST3 (5′-TAAACTGCGAAATCGCCGTG-3′), a reverse complement sequence of ST1, was used as positive control for hybridization. ST3 was 3′-end tailed with DIG-dUTP/dATP using terminal transferase following the manufacturer's instructions (Roche Applied Science).

Sequencing of DIG-labelled amplicons

Each DIG-labelled amplicon was sequenced with PenF1 primer to re-confirm its identity. DIG-labelled COX1 amplicons were first incubated with ExoSAP-IT (USB Corp.) to digest unincorporated primers and inactivate remaining dNTPs present in the PCR product. Each 10 µL sequencing reaction mix contained 1.75 µL of 5× sequencing buffer version 3.1, 0.5 µL of 2.5× BDT Sequencing Mix v3.1, 0.5 µL of PenF1 primer, 6.25 µL of sterile HPLC H₂O and 1 µL of PCR template. Cycle sequencing used the following thermal profile: 95 °C for 3 min followed by 40 cycles of 96 °C for 45 s, 56 °C for 30 s, and 60 °C for 3 min. Amplifications reaction were stopped and held at 10 °C. Sequencing products were subjected to sodium acetate/EDTA/ethanol precipitation and were then resuspended in Hi-Di formamide and sequenced by an Applied Biosystems ABI PRISM 3130xl Genetic Analyser. Sequencing results were compared with the original Penicillium subgenus Penicillium COX1 data set of Seifert et al. (2007).

DNA hybridization

Hybridization of DIG-labelled amplicons to the COX1 array conformed to methods described by Fessehaie et al. (2003) with the following modifications. Membranes were first incubated in pre-hybridization solution [6 × SSC (1 m sodium chloride, 0.1 m sodium citrate), 1% sarcosine, 0.02% sodium dodecyl sulphate (SDS), 1% skim milk] for 1.5 h at 54 °C. Each membrane was then incubated at 54 °C overnight with 10 mL of hybridization buffer [6× SSC (1 m sodium chloride, 0.1 m sodium citrate), 1% sarcosine, 0.02% SDS] containing c. 100 ng of DIG-dUTP-labelled amplicon at 54 °C overnight, followed by two washes with stringency buffer (2× SSC, 0.1% SDS) also at 54 °C. Membranes were then incubated with anti-DIG alkaline phosphatase conjugate (diluted 1:25 000) followed by CDP-star (diluted 1:1000) (both from Roche Diagnostics GmbH) and then sealed in transparent plastic sheets. The chemiluminescent signals were captured in an Alpha Innotech FluoChem darkroom using an HD 16-bit digital camera (Alpha Innotech) and analysed with Genepix Pro version 6.0 software. The average intensity of hybridization signals generated by the same oligonucleotide in the duplicate sets on the same membrane was used for analysis. We stripped and reused membranes up to five times in the current study without detectable degradation in signal intensities using the protocol reported by Fessehaie et al. (2003).

Testing of membrane with environmental samples

Mouldy cheese and plant samples were used to test the COX1 array. Blue and camembert cheese samples, which are fermented with species of subgenus Penicillium, were bought at local grocery stores. Contaminated cheese samples were collected from household refrigerators. Mouldy plant samples, including mouldy apples, oranges, tulip bulbs and cherries, were also collected locally. Total DNA was extracted from mouldy area of plant samples and both spoiled and edible cheese samples with EZNA Fungal DNA extraction kit (Omega Bio-Tek) following the manufacturer's instructions. DNA was visualized after electrophoresis on 1.5% agarose gel, and concentrations were estimated by NanoDrop 1000 (Thermo Fisher Scientific). DIG-labelled PCR amplicons from total genomic DNA were generated with PenF1 and AspR1, and hybridization on the array was carried out as described previously. A positive control consisting of a known template DNA and a negative control with no DNA were included with the hybridization.

Sporulating fungi in cheese and fruit samples were isolated into pure culture by removing conidia to corn meal agar (CMA) with 0.2 mg/g chloramphenicol added to inhibit bacterial growth. Species were identified using morphological characters (Frisvad & Samson 2004) and confirmed by BenA sequencing (Samson et al. 2004). The species isolated into culture were compared with the profile detected by the COX1 array.

Results

Design of the barcode oligonucleotides and printing of the COX1 array

Sequence divergences of COX1 barcodes between and within species of Penicillium subgenus Penicillium were 5.6% and 0.06%, respectively (Seifert et al. 2007). The neighbour-joining tree indicated that six species complexes shared identical COX1 sequences, namely the P. aurantiogriseum complex (7 species), P. camemberti complex (5 species), P. hirsutum complex (3 species), and 3 species pairs, namely P. radicicola and P. tulipae, P. nordicum (5 of 9 strains) and P. verrucosum, P. dipodomyis and P. nalgiovense (1 of 4 strains). Therefore, only clade/group-specific oligonucleotides were possible for these complexes. The remaining 39 species have unique barcode sequences for oligonucleotide design, although the absence of indels made it challenging to design oligonucleotides with sufficient resolution to distinguish closely related species.

SigOli identified from 1 to 30 locations with the potential to provide oligonucleotide(s) for each of the unique sequences in the data set. Array Designer version 1.1 eliminated oligonucleotides with hairpins or dimers, and optimized oligonucleotide sequences for melting temperature (Tm). This resulted in 182 PM oligonucleotides 20–41 bases long, with an average G:C content of 31.9% (range 16–55%) and an average theoretical Tm of 54.9 °C (range, 54.0–55.7 °C). Oligonucleotide codes, sequences and origins are shown in Table 1. Among the selected PM oligonucleotides, approximately 20% had only 1 single nucleotide polymorphism (SNP) within 3 bases from either end of the sequence to enable distinguishing the target clade/species from other closely related species. We expected these oligonucleotides to lack specificity, and cross-reactions were anticipated.

Each PM oligonucleotide was evaluated by blast searches against the Penicillium subgenus Penicillium COX1 data set to confirm the expected specificity (Appendix S2, Supporting information). One hundred oligonucleotides were selected to target species, or infraspecies haplotypes, representing 27 species, including P. atramentosum, P. olsonii, P. formosanum, P. brevicompactum, P. glandicola, P. gladioli, P. paneum, P. vulpinum, P. sclerotigenum, P. carneum, P. nordicum (some strains), P. aethiopicum, P. concentricum, P. dipodomyicola, P. griseofulvum, P. expansum (3 of 5 strains), P. chrysogenum, P. clavigerum, P. crustosum, P. digitatum, P. marinum, P. nalgiovense (3 of 5 strains), P. hordei, P. tricolor, P. coprobium (4 of 5 strains), Eupenicillium osmophilum and Penicillium sp. 2 (Table 1). The remaining 82 oligonucleotides were designed for clades containing two species or more.

No group-specific oligonucleotides could be designed for six of the species complexes that shared identical barcode sequences. The exception was the P. hirsutum complex, for which one complex-specific oligonucleotide was selected (Pen_COX1_28a). However, because of polymorphisms within P. aurantiogriseum, P. nalgiovense and P. nordicum, it was possible to design infraspecies-specific oligonucleotides. Pen_COX1_10 (h and i) were designed for three strains of P. aurantiogriseum (CBS 792.95, 110327, 110329), Pen_COX1_59 (a and b) were designed for three strains of P. nalgiovense (CBS 318.92, 112438, 109610), and Pen_COX1_36 was designed for four strains of P. nordicum (CBS 110770, 110771, 112565, 112573). Oligonucleotide Pen_COX1_20 was designed for a large group of 27 species, comprising the entire section Viridicata as defined by Frisvad & Samson (2004), except P. atramentosum. Some other group oligonucleotides, such as Pen_COX1_16 (targeting eight species in Section Chrysogena, plus Eu. crustaceum) and Pen_COX1_35a-c (targeting six species, including P. clavigerum, P. venetum, P. crustosum, P. discolor, P. echinulatum and P. solitum), were also designed for large clades. Oligonucleotide Pen_COX1_1 contains three ambiguity codes and was designed as a universal probe for this subgenus.

Properties of the spots after printing the membranes were visible due to the Bromophenol blue in solution and were compared between arrays printed robotically and manually. Spots printed with split pins were approximately one-fifth the size of those made manually, and robotically printed arrays had greater consistency of spot size, spacing, and intensity, with significantly smaller standard errors (Table 2).

Table 2.  Comparison of size, spacing and colour intensity of blue spots on arrays printed by robot compared with the manual V384 solid pin replicator method (n = 500)

	MicroGrid compact robot	VP384 solid pin multi blot replicator
	Mean	Mean	S.E.
Spot diameter (µm)	226.6	1230.3	2.5
Spot spacing (µm)	803.7	1986.7	15.7
Colour intensity of blue spots	14439.9	1488.2	168.0

In pilot experiments, hybridization conditions were optimized for consistency and reproducibility. These tests showed that, in comparison with macroarrays, the COX1 array functioned with similar prehybridization and hybridization conditions (6× SSC, 54 °C) but required higher stringency conditions in post-hybridization washes (2× SSC instead of 6× SSC, 54 °C) to deliver comparable specificity. Further increasing the stringency of the washing buffers from 2× SSC to 1× SSC or 0.5× SSC resulted in significant reduction of the chemiluminescent signal and subsequently affected the accuracy of signal intensity analysis. PCR amplicons from P. cyclopium (CBS 110337) were diluted to 20 ng, 10 ng, 5 ng, and 1 ng per mL hybridization buffer for hybridization reactions. An amplicon concentration of 10 ng/mL resulted in the best balance between specificity and signal intensity and was used for the other species.

Validation of oligonucleotides using DNA from pure cultures

Because of 100% sequence homology among multiple strains and several closely related species, 70 COX1 amplicons from strains representing all available barcode groups were hybridized to the array, to test the predicted in silico hybridization pattern (Appendix S2). Sequencing results of DIG-labelled COX1 amplicons of these samples reconfirmed their identity by blast and pairwise alignment.

Hybridization with DNA from pure cultures showed that most PM oligonucleotides displayed much stronger signals when hybridizing with target amplicons, than corresponding MM oligonucleotides. The pairwise ratios of the signal intensities were affected by oligonucleotide length, number of SNPs, SNP position, the mismatch-pair type and the base-pair type that flanked the SNP (Fig. 2). Shorter PM oligonucleotides with central SNPs had better discriminatory potential between the PM and MM oligonucleotides. Pyrimidine–pyrimidine mismatch pair and purine–purine base pair flanking a single SNP seemed less stable than other types. This resulted in a higher log ratio of signal intensity between the PM and the corresponding MM oligonucleotide (Fig. 2b).

Figure 2. (A) Scatter plot of hybridization signal [log of (signal + 1)] for each oligonucleotide pair: y-axis for perfect match (PM) oligonucleotides, x-axis is for corresponding mismatch (MM) oligonucleotides. Most of these pairs are above the 45-degree line, showing pairwise log ratio PM to MM signal greater than 0 and good differential discriminatory potential between the PM and MM oligonucleotides. (B) Average pairwise log ratio of hybridization signal intensity (PM to MM) for different oligonucleotide properties: including oligonucleotide length (1), number of SNPs per oligonucleotide (2), for oligonucleotides with 1 SNP, location of SNP from the centre of the oligonucleotide (3), the pyrimidine/purine (py/pµ) mismatch type (4) and nucleotide-pair type that flanking the SNP (5) between the PM and MM oligonucleotides.

Seventy-six oligonucleotides had the high specificity predicted, with minimal cross-reaction (54 oligonucleotides designated as group 1 in Fig. 3) or strong signal intensity compared to a few undesired low signal cross-reactions (22 oligonucleotides in group 2, Fig. 3). All oligonucleotides in group 1 had less than five false positive signals, whereas those in group 2 had five to 13 false positives but gave a stronger signal with the target species. Of the 76 oligonucleotides, 52 were species-specific and 24 were group-specific (Table 1, Fig. 3). The 52 species-specific oligonucleotides can identify species or haplotypes of 20 Penicillium species, including Eu. osmophilum, P. aethiopicum, P. atramentosum, P. brevicompactum, P. concentricum, P. crustosum, P. digitatum, P. dipodomyicola, P. expansum, P. formosanum, P. gladioli, P. glandicola, P. hordei, P. marinum, P. nordicum, P. olsonii, P. paneum, P. tricolor, P. vulpinum and Penicillium sp. 2. Twelve oligonucleotides only hybridized with their target species/clade and had no cross-reactions. The remainder displayed background signals with some nontarget amplicons, although of much fainter intensity in comparison with those from target amplicons. Within the 24 group-specific oligonucleotides, 15 were designed for groups containing 2–4 species, such as Pen_COX1_38 (b and d), which was specific for two sister species (P. dipodomyicola, P. griseofulvum), while the other eight were for large clades.

Figure 3. Hybridization patterns of DIG-labelled PCR amplicons of subgenus Penicillium species to the COX1 oligonucleotide array on nylon membranes. Chemiluminescent signals were captured by 16-bit digital camera in a darkroom and analysed with Genepix Pro version 6.0 software. Intensities are indicated by the following symbols: < 1000, blank; 2001 to 4000, ; 4001 to 10000, ; 10001 to 20000, ; 20001 to 30001, ; above 30001, . Signal symbols with a grey background indicate the expected positive reactions, while signal symbols lacking grey background indicate unexpected cross hybridization. All selected oligonucleotides in group 1 (54 oligonucleotides) have fewer than five false positive signals, and those in group 2 (22 oligonucleotides) have five to 13 false positives. All oligonucleotides with more than 13 false positives or which did not hybridize to any target species were rejected.

One hundred and six PM oligonucleotides were rejected for at least one of the following reasons: (i) no hybridization was detected for the target species [six oligonucleotides, including Pen_COX1_10 (a, b, c, d, and g), and 48]; (ii) more than 13 false positives occurred (76 oligonucleotides); and (iii) 5–13 false positive signals occurred, but the maximum target signal was equal to or less than the false positives (29 oligonucleotides). The rejected oligonucleotides are listed in Appendix S3, Supporting Information. Of these, 29 had only one SNP within three bases from either end, which would allow less stringent hybridization to occur.

Oligonucleotides with two SNPs delivered good specificity in general; 17 such oligonucleotides were designed. Four oligonucleotides with two SNPs were rejected, including Pen_COX1_11a, 15e, 32a and 47d. The latter two did not perform well, perhaps because one SNP was located at 5′-end of each sequence, and the other SNP was located within 3 bases of the 3′-end.

Testing of COX1 array with environmental samples

Genomic DNA was isolated from a few fermented or moulded cheese samples and some plant samples with visible growth of species of Penicillium. In Fig. 4, the top rows represent samples of brie cheese (fermented with P. camemberti and sometimes also P. caseifulvum) and blue cheese (fermented with P. roqueforti). No species-specific oligonucleotides for P. camemberti, P. caseifulvum or P. roqueforti could be derived from this COX1 data set. The former two species are within a large clade targeted by Pen_COX1_20, and the presence of P. roqueforti can be inferred from other positive reactions, as discussed below. For the brie cheese samples, very few oligonucleotides displayed positive signals. Blue and stilton cheese samples reacted positively with oligonucleotides Pen_COX1_18 (a–d) and 24, designed for the three species of the P. roqueforti complex, i.e. P. carneum, P. paneum, and P. roqueforti. These oligonucleotides also reacted with Pen_COX1_34, designed for P. carneum, but with a much weaker intensity. No reaction was detected with oligonucleotide Pen_COX1_25 c, designed for P. paneum. These results indicated that no P. paneum was present in these sample, and because the signal intensity from the P. carneum-specific oligonucleotide was weaker than that of Pen_COX1_18 (a–d) and 24, we infer that P. roqueforti is present in these samples. This was confirmed by isolation of this species from the cheeses, with the identification confirmed by morphology and BenA sequencing.

Figure 4. Hybridization patterns of DIG-labelled COX1 amplicons of DNA extracted from cheese and plant samples to 71 selected oligonucleotides on the Penicillium COX1 array. Intensities of chemiluminescent signals are indicated by the following symbols: < 1000, blank; 2001 to 4000, ; 4001 to 10000, ; 10001 to 20000, ; 20001 to 30001, ; above 30001, .

The goat, Oka, havarti and parmesan cheese samples lacked visible mould and served as negative controls. However, in these samples, a few Penicillium species were detected. Oligonucleotide Pen_COX1_18 (a–d) and 24 displayed positive signals for parmesan and goat cheese, but there were no reactions for oligonucleotide Pen_COX1_34 and 25 c, specific for P. carneum and P. paneum. The inferred presence of P. roqueforti in these cheese samples was confirmed by morphological and BenA sequencing of cultures isolated from the parmesan sample. These cheese samples came from a shop that packaged cheese on site, and we suspect that the contamination of the non-blue cheeses originated with P. roqueforti from blue cheeses handled in the same facility. A weak signal for P. glandicola, a species usually associated with Quercus, was detected in goat cheese, but the fungus was not isolated. In apparently uncolonized areas of mouldy cheddar cheese samples collected from household refrigerator, only very weak signals for P. tricolor and P. gladioli were detected. These cultures isolated from these samples were P. commune.

The bottom rows of Fig. 4 represent samples of genomic DNA isolated from visibly mouldy plants. Penicillium expansum was detected using the COX1 array in the mouldy apple and cherry samples; this was confirmed by morphological and BenA identification of isolated cultures. Strong signals for P. digitatum were found in the mouldy orange samples, but P. italicum was isolated from the mouldy area on the orange peel. Both species are well-known post-harvest spoilage agents of Citrus, and it is likely that both were present on the sample; no oligonucleotides for P. italicum were on the array. Penicillium tulipae was isolated from the mouldy tulip bulbs, but again no species-specific oligonucleotides were available. Weak signals for P. olsonii and P. tricolor were detected in mouldy areas from both apple and tulip samples, and from mouldy cherries, respectively, but these two species were not isolated by culturing and we suspect that these were false-positive reactions.

Discussion

We developed a high-density membrane-based COX1 oligonucleotide array, potentially a useful tool for detection of species of Penicillium subgenus Penicillium in ecological surveys, epidemiological studies, and cases of food spoilage. The array was validated using genomic DNA isolated from 70 DNA extracts selected from of 358 strains representing the 60 species in this group, including two new species. The array behaved as predicted in silico for some oligonucleotides, but did not behave as expected for other oligonucleotides.

Validation tests using environmental samples showed that this COX1 array can successfully detect the apple rot fungus P. expansum and the citrus mould P. digitatum, and can be useful for distinguishing P. roqueforti from the other green penicillia that contaminate cheese. DNA hybridization results correlated with microbial isolations from the same samples. However, because the interspecies divergence among COX1 barcodes is only 5.6%, and in many cases, sequences of sister species differ by only a single nucleotide, it was difficult to find taxon-specific probes for many of the species. Thus, the array has limited utility for several clades, including the brie cheese complex, and is unable to identify some of the critical spoilage species in cheese, such as P. commune.

Generally, more species-specific oligonucleotides were found for species on long branches in the COX1 gene tree, such as P. bialowiezense, P. olsonii, and P. atramentosum. A few group oligonucleotides [Pen_COX1_18 (a–d) and 24] were found for the well-structured section Roquefortii, including P. roqueforti, P. carneum and P. paneum, with only the latter two species having species-specific oligonucleotides identified (namely, Pen_COX1_34 and Pen_COX1_25c). However, as shown with the cheese samples tested with the array, the presence of P. roqueforti can be inferred by the combination of positive reactions for group oligonucleotides, and negative reactions for the two species-specific oligonucleotides, as well as by differences in the signal intensity of these reactions. Unfortunately, some oligonucleotides detected groups with two or more included species that lacked species-specific oligonucleotides, and inferred detection of individual species was then not possible. For example, the group oligonucleotide Pen_COX1_45 was designed for four species including P. crustosum, P. echinulatum, P. solitum and P. discolor, but only the former two had species-specific oligonucleotides. Therefore, P. solitum and P. discolor could not be distinguished with this array.

Many oligonucleotides were rejected because of unexpected cross-reactions. We questioned whether sequencing errors in the original study may have led to the design of faulty oligonucleotides, leading to false-negative/false-positive results. All DIG-labelled amplicons were sequenced to reconfirm their identity, eliminating sequencing errors as a possible reason for false negatives or false positives. Instead, the limited number of differential nucleotides among closely related Penicillium taxa dictated that no candidate oligonucleotides could be identified for many species. Many probes designed for this study had mismatches with nontarget species close to the end of the oligonucleotide, such that the discriminatory ability was less than expected (Table 1). Array Designer rejected some oligonucleotides with central mismatches because of hairpins, dimers and unsuitable thermodynamic characters for the selected hybridization conditions.

Probes ranged from 20 to 41 bp, with the optimal length and best specificity being 20 to 35 bp, and mostly 20 to 30 bp. Oligonucleotides longer than 35 bp were AT-rich and often had low specificity. When there are only one or two polymorphisms, cross-reactivity increases as oligonucleotide length increases, because the ratio of mismatched-to- matched base pairs decreases. Probe length also affects how many bases are actually available for hybridization (Dorris et al. 2003) and therefore affects signal intensity (Lievens et al. 2006).

A few well-documented studies indicate that the number, type and position of mismatches play a more important role in the thermodynamics of DNA hybridization, but that each factor may affect hybridization specificity in different ways in different sequences (Urakawa et al. 2002, 2003; Lievens et al. 2006; Pozhitkov et al. 2006). These studies designed single or double mismatch variants for each PM oligonucleotide in order to fully assess and explore the discriminatory power of the array. Theoretically, a single mismatch near the middle of an oligonucleotide should allow two species to be distinguished (Kawasaki et al. 1993), whereas oligonucleotides with mismatches at the extreme 5′-end or 3′-end (especially at the 5′-end) have the least discriminatory potential (Lievens et al. 2006). As for a short oligonucleotide with a single SNP, the location of the SNP may affect the stability of an MM duplex more than the MM type of the SNP (Urakawa et al. 2002). Our results clearly showed that shorter (16 to 25 mer) oligonucleotides with more than one SNP had higher differential potential between the PM and MM oligonucleotides, and this potential decreased dramatically when an SNP localized (especially by more than 10 bases) away from the centre of the sequence (Fig. 2B).

By comparing the normalized signal intensity profiles between perfect-match and mismatched oligonucleotide-target duplexes, Lievens et al. (2006) suggested only one SNP within the sequence of an oligonucleotide can affect its discriminatory potential, and that centre mismatches may actually provide less selectivity than substitutions towards the 5′- or 3′-ends. They also suggested that higher specificity should be expected from oligonucleotides with SNP(s) located on the 3′-half of the sequence. Pozhitkov et al. (2006) noted that it is difficult to predict the actual performance of particular probes in DNA microarrays in silico when hybridizing to rRNA targets. They suggested (i) that pyrimidine–pyrimidine mismatch pairs are more stable than purine–purine mismatches; (ii) that the stability of MM duplexes increases as follows: G–A < G–G < C–A < A–G < A–A < C–U < C–C < A–C < G–U < T–C < T–U < T–G; and (iii) that the nucleotides flanking SNPs seem to affect duplex stability as follows: purine-MM-Purine < purine-MM-pyrimidine < pyrmidine-MM-purine < pyrimidine-MM-pyrimidine. The Gibbs free energy and other thermodynamic parameters calculated for the secondary structure of rRNA sequences did not accurately predict signal intensity of DNA-rRNA duplexes (Pozhitkov et al. 2006).

Some of these rules may or may not be universally applicable, such as for both DNA–rRNA and DNA–DNA duplexes, but these studies suggest that in addition to the number of SNPs and their location, the actual sequence of an oligonucleotide affects how selective it can be. Indeed, some oligonucleotides in our study, with more than one mismatch in the middle of the sequence, still cross-reacted with amplicons from closely related species, which should not have occurred in our highly stringent hybridization conditions (e.g. Pen_COX1_4b and Pen_COX1_18d). This emphasizes that the specificity of an oligonucleotide can be affected by several interacting factors, and these effects need to be considered when developing oligonucleotide arrays. For these reasons, we designed multiple oligonucleotides for each targeted species/clade whenever possible, to allow us to compare the specificity of oligonucleotides harbouring SNPs at with different locations. For instance, Pen_COX1_7a–g were designed for P. olsonii. Oligonucleotides 7b, c, d and f have SNPs located near the middle of the sequence, whereas 7b, e and g have SNPs towards the 5′-end. In our experiments, Pen_COX1_7 c and Pen_COX1_7 d did not deliver satisfactory specificity and were rejected. Instead, 7e with one T–T mismatch eight bases from the 5′-end had more discriminatory potential and was retained. Similarly, Pen_COX1_50 a and Pen_COX1_50 b have a single mismatch near the 3′-end flanked with pyrimidine–purine and pyrimidine–pyrimindine), respectively. Both oligonucleotides reacted selectively but weakly with their target species, P. expansum. Oligonucleotides with SNPs near either terminus normally have low discriminatory potential; however, as mismatch duplexes were unstable under the hybridization condition, both oligonucleotides were considered worth retaining. Our results also demonstrate that in silico modelling of DNA hybridization, with controlled and simplified virtual hybridization conditions, may be inadequate to represent more complicated and less restrictive experimental settings. For example, current data showed that duplexes with pyrimidine–pyrimidine mismatch pair are less stable than other mismatch types, and the signal intensity of a mismatch with purine–purine base pair flanking the SNP was significantly lower than other types (Fig. 2b). These results are inconsistent with the conclusions made by Pozhitkov et al. (2006). Such inconsistency may be especially apparent when the array is used for environmental samples.

Nevertheless, a few simple rules can be suggested for designing oligonucleotides with greater specificity: (i) when possible, select probes with more than one SNP located close to the centre of the sequence; (ii) avoid using probes containing less destabilizing mismatch types, such as T:G mismatches, using antisense probes instead; (iii) probes ranging from 25–35 bases may have the best balance between specificity and sensitivity.

DNA arrays based on gene regions with low sequence divergence often cannot distinguish among closely related species by simply setting a threshold for signal intensity, because false positives occur frequently. Similarly, a barcoding gene lacking indels such as COX1 may not be amenable to adaptation to oligonucleotide-based, automated identification schemes. Engelmann et al. (2008) proposed a matrix algebra approach to estimate DNA amount present in a sample and correct for cross-reaction signal. They assumed that the signal intensity matrix Y (oligonucleotides × experiments) and the DNA concentration matrix X (target amplicons × experiments) have a linear correlation Y = A · X, and that the affinity matrix A (oligonucleotides × target amplicons) therefore can be calculated. If the affinity matrix is known, the DNA concentration matrix can be predicted from the probe intensity matrix if oligonucleotides and target amplicons used in all experiments are the same. This approach was useful even where both cross-reactions and binding properties between the probe and the target were considered, when using threshold values for species differentiation (Engelmann et al. 2008). Oligonucleotides rejected in our study seemed to perform better than those presented in the matrix algebra paper (Engelmann et al. 2008). Among the 106 rejected oligonucleotides, more than 75% generated stronger signals with amplicons from the target species, although significant cross-reactions were widely detected. For example, oligonucleotides Pen_COX1_15e, 45, 54a, 54b and 60a, all cross-hybridized with more than 20 nontarget amplicons, but the strongest PM signal was significantly more intense than the highest false-positive signals. As such, it may be possible to apply the general linear model to our data set for distinguishing closely related Penicillium species in environmental samples. However, this simplified model used only one immobilized DNA strand for each target species, and did not adapt well to our study with multiple oligonucleotides being designed for each species/clade. If the convergence of the matrix is close to or equal to zero, the affinity matrix cannot be accurately calculated. Nevertheless, adaptations of this model may significantly improve the discriminatory ability of DNA arrays for more complicated experimental scenarios.

Overall, the automated nylon membrane array printing procedure at near microarray density described here could accommodate many more oligonucleotides and be adapted or employed for similar applications to improve detection efficiency. We were also satisfied with the performance of the in-house computer program SigOli (Zahariev et al. 2009), which successfully identified multiple potential locations for barcode oligonucleotide design at most resolution levels (15 out of 16 in total) for the Penicillium subgenus Penicillium data set. This allowed us to design species/subspecies-specific oligonucleotides for 27 Penicillium species out of the 39 with unique COX1 sequences. The 76 accepted oligonucleotides performed well with natural samples, and with the help of the matrix algebra theory (Engelmann et al. 2008), the remaining 102 rejected oligonucleotides may also be useful for detection of species in this subgenus.

It is worth emphasizing that irrespective of what DNA marker is chosen as the barcode for the fungal kingdom, an oligonucleotide barcode array with high specificity must be based on a sequence data set with sufficient variation to allow the design of multiple oligonucleotides per species or clade. The presence of indels or introns in this region may or may not be favourable for full-length barcode identification, but such sites provide obvious polymorphism locations for designing oligonucleotides that are less likely to cross-react with unintended targets. Tambong et al. (2006) and Robideau et al. (2008) developed their oligonucleotide arrays using ITS data sets. Probes designed from polymorphisms in indels sites ensured high specificity of both arrays. Very few cross-reactions were detected when these arrays were validated with environmental samples. Therefore, to further improve the discriminatory ability of the DNA array for Penicillium species, we are considering a second generation array, including oligonucleotides from both COX1 and β-tubulin. At least one more highly variable gene will be required to design species-specific assays for all closely related species. We are also considering modifying the immobilized oligonucleotides at the SNP positions with locked nucleic acids (LNAs) to improve PM duplex stability and shorten the oligonucleotides, therefore enhancing the specificity of selected oligonucleotides (Chou et al. 2005; You et al. 2006).

Acknowledgements

This research was supported through funding to the Canadian Barcode of Life Network from Genome Canada (through the Ontario Genomics Institute), NSERC and other sponsors listed at http://www.BOLNET.ca. We thank Anne Johnson and Nicole Désaulniers, Agriculture and Agri-Food Canada, for helping us print the COX1 array using the MicroGrid Compact robot and for work on pilot experiments. We are also grateful to Mr Jeremy deWaard and Gerry Louis-Seize for providing the DNA samples extracted from pure cultures of Penicillium subgenus Penicillium used in our experiments.

Conflict of interest statement

The authors have no conflict of interest to declare and note that the funders of this research had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.