Development of a novel mycobiome diagnostic for fungal infection

Abstract

Background: Amplicon-based mycobiome analysis has the potential to identify all fungal species within a sample and hence could provide a valuable diagnostic assay for use in clinical mycology settings. In the last decade, the mycobiome has been increasingly characterised by targeting the internal transcribed spacer (ITS) regions. Although ITS targets give broad coverage and high sensitivity, they fail to provide accurate quantitation as the copy number of ITS regions in fungal genomes is highly variable even within species. To address these issues, this study aimed to develop a novel NGS fungal diagnostic assay using an alternative amplicon target.

Methods: Novel universal primers were designed to amplify a highly diverse single copy and uniformly sized DNA target (Tef1) to enable mycobiome analysis on the Illumina iSeq100 which is a low cost, small footprint and simple to use next-generation sequencing platform. To enable automated analysis and rapid results, a streamlined bioinformatics workflow and sequence database were also developed. Sequencing of mock fungal communities was performed to compare the Tef1 assay and established ITS1-based method. The assay was further evaluated using clinical respiratory samples and the feasibility of using internal spike-in quantitative controls was assessed.

Results: The Tef1 assay successfully identified and quantified Aspergillus, Penicillium, Candida, Cryptococcus, Rhizopus, Fusarium and Lomentospora species from mock communities. The Tef1 assay was also capable of differentiating closely related species such as A. fumigatus and A. fischeri. In addition, it outperformed ITS1 at identifying A. fumigatus and other filamentous pathogens in mixed fungal communities (in the presence or absence of background human DNA). The assay could detect as few as 2 haploid genome equivalents of A. fumigatus from clinical respiratory samples. Lastly, spike-in controls were demonstrated to enable semi-quantitation of A. fumigatus load in clinical respiratory samples using sequencing data.

Conclusions: This study has developed and tested a novel metabarcoding target and found the assay outperforms ITS1 at identifying clinically relevant filamentous fungi. The assay is a promising diagnostic candidate that could provide affordable NGS analysis to clinical mycology laboratories.

Keywords: Amplicon sequencing; Diagnostics; Fungal infection; Mycobiome.

Fig. 1

In silico analyses of ITS1 and TEF1α targets. (A) PrimerTree was used to search primers against fungal sequences within the NCBI nr nucleotide database. The frequency of product lengths are displayed for each target (vertical dotted line indicates median). (B) A curated database of TEF sequences was established. The 8656 sequences cover the five phyla containing known human pathogenic fungi and the number of species covered within these phyla are shown. (C) ITS1 (left) and TEF1α sequences were obtained for over thirty clinically relevant fungal species and alignments performed using Clustal to demonstrate fungal speciation. Phylogenetic trees were generated using maximum-likelihood method and Tamura-Nei model

Fig. 2

Fungal amplification using TEF primers is more efficient and displays reduced species bias compared to ITS1. A. Quantitative PCR efficiency (left) was significantly higher with TEF compared to ITS1 for all three species (P < 0.0001, 2-way ANOVA). Species efficiencies were significantly different from one another for ITS1 (*; P < 0.05, ***; P < 0.001 by 2-way ANOVA Tukey’s multiple comparisons). Relative quantitation of (A) fumigatus, C. albicans and L. prolificans using the ITS1 (middle) and TEF (right) PCR assays shows under-representation of L. prolificans with ITS1. (B) Quantitative PCR of 50,000 haploid genome equivalents of twenty one fungal species shows species quantitation is more variable with ITS1 than TEF. Data is visualised as difference from mean Ct value. Ct values were significantly different between species for TEF and ITS1 (P < 0.001 and P < 0.0001, respectively, by ANOVA. Of the 210 Tukey’s multiple comparisons tests performed, 6 and 81 were significant for TEF and ITS1, respectively (P < 0.05, not indicated)

Fig. 3

TEF assay outperforms ITS1 when sequencing fungal mock communities. Two representative fungal mock community analyses are shown. Each community contained 5 species, was targeted by TEF and ITS1 in duplicate and sequenced on an Illumina iSeq and MiSeq. For ITS1 samples, iSeq sequencing was performed twice. Bar plots indicate normalised species count data and expected mock community (dashed green box). Dendrograms indicate hierarchical clustering of Bray-Curtis distances and samples are ordered accordingly. A. ITS1 either significantly under-estimates or fails to identify (A) fumigatus when it is dominant (92% genome equivalents) in a community. TEF identifies all species present and at close to the expected relative abundances. (B) ITS1 correctly identifies (C) neoformans when it is dominant but fails to identify filamentous species present at a low level (A. fumigatus and P. rubens). TEF identified all species present and at close to the expected relative abundances

Fig. 4

TEF metabarcoding assay outperforms ITS1 when detecting low level filamentous fungal species in mock communities. (A) Boxplot of percent abundance for species spiked at 2% within mock communities targeted by ITS1(left) or TEF (right). Expected percent abundance is indicated by dashed black line. The red dotted line indicates no identification. Percent abundance was significantly different between species for each target (P < 0.0001 by Kruskal-Wallis test). For each target, percent abundances are grouped by sequencing instrument. For those which were identified, excluding Lomentospora prolificans when targeted by ITS1, species level quantifications were not significantly different between instruments (ns, P < 0.05 by Wilcoxon rank sum test). Wilcoxon rank sum tests were not performed for species with less than 3 data points per human background status and are indicated with ‘ND’. (B) Heatmap representation of identification (ID) rates for species when spiked at 2% within mock communities. ITS1 failed to identify four filamentous fungi in any samples, and another five species had suboptimal ID rates. TEF identified all but one species in all samples tested. Sequencing instrument did not affect ID rates overall

Fig. 5

TEF species detection in mock fungal communities is not significantly hindered by human gDNA background. A. Boxplot of percent abundance for species spiked at 2% within mock communities targeted by TEF. Expected percent abundance is indicated by dashed black line. Dotted red line indicates no identification. Percent abundances are grouped by human DNA background status. Species level quantifications were significantly different when with and without human background for (A) terreus, C. auris and C. albicans (*; P < 0.05 or ***; P < 0.0001 by Wilcoxon rank sum test). For all other species which were identified in a sample, quantifications did not differ significantly (P < 0.05 by Wilcoxon rank sum test). Wilcoxon rank sum tests were not performed for species with less than 3 data points per human background status and are indicated with ‘ND’. (B) Heatmap representation of TEF identification (ID) rates for species when spiked at 2% within mock communities depending on human DNA background status. Human DNA resulted in no change for most species tested

Fig. 6

Assay evaluation using A. fumigatus spiked clinical respiratory samples. A sputum sample (previously negative for Aspergillus by ITS and TEF sequencing) spiked with varying quantities (2, 10 and 25 haploid genome equivalents/µl) of A. fumigatus gDNA. (A) Fungal read counts are shown for each sample in quadruplicate. Samples with less than 500 fungal reads were removed. No template control and unspiked sputum sample did not produce significant fungal read yields. (B) Mean A. fumigatus counts displayed a linear relationship with amount of spiked A. fumigatus (error bars indicated standard error). R2 value for linear relationship is indicated

Fig. 7

Absolute quantification of Aspergillus fumigatus using internal plasmid control PC1. A. Clinical respiratory samples were spiked with four different amounts of A. fumigatus (A. f) genome equivalents (GE) to generate a standard curve ranging from ˜ 3 to 1000 GE per PCR. Read counts from the internal plasmid control (PC) were used to normalize (A) f counts and a positive linear relationship (R2 = 0.89) with spiked GE was observed. (B) Using the linear model generated by the standard curve, A. f GE were estimated in 12 clinical respiratory samples known to be positive for A. f (plus one negative control sample containing no A. f). Estimated A. f GE for test samples were calibrated using the estimated A. f GE value of the negative control sample. The estimated A. f GE values displayed strong correlation (Spearman’s rank-order correlation; r [11] 0.78, p < 0.01) with known GE (calculated by rRNA qPCR)