Towards metagenomics-based diagnostics to detect and prevent plant pathogens

Davide Spadaro, Edoardo Piombo

DISAFA and AGROINNOVAUniversità di Torino, Italy

Global trade...

https://research.rabobank.com/far/en/sectors/regional-food-agri/world_fruit_map_2018.html?qsl_reqcnt=1

Global trade... and pests

(2018)

Trade of plant material(seeds, propagation material, wood, food, feed)Global movement of pests and diseases

Plant Health Authorities (EPPO, NPPOs, USDA-APHIS)Safe trade

Rapid diagnosis

Export certification Import inspection Disease surveillance and monitoring Containment and eradication programmes

Disease diagnosis: cross-cutting issue (IPPC, 2016)Diagnosis should be rapid

A range of methods

Untargeted methodsCulturing (in vitro/in vivo)Microscopy (optical MS, electron MS)

Targeted methodsImmunological (ELISA)Molecular (PCR, qPCR)

Classical techniques are labor-intensive, slow andgreatly dependent on the expertise of the operator.

Pros• Faster• More sensitive• More specific• More accurate• Can identify non-culturable pathogens• Can differentiate closely related pathogens• Allows for pathogen quantification (qPCR)

Cons• Limited number of targets in each analysis

Molecular techniques

Direct quantification (droplet digital PCR (ddPCR) Field diagnostics (LAMP)

Multiplexing (microarray) Online application (Integration with IT, LOC)

Directions of molecular techniques

Temperature Day inoculated Day testedday after

inoculum

disease

index

average

Tp (h:min:s)-Crude

extraction method

20°C

16/08/2017 18/08/2017 3 0 00:11:31±00:01:55

16/08/2017 21/08/2017 5 0 00:10:03±00:03:02

16/08/2017 23/08/2017 8 0.67 00:10:56±00:00:10

16/08/2017 25/08/2017 10 1 00:07:06±00:04:25

16/08/2017 28/08/2017 13 4 00:09:02±00:01:32

16/08/2017 30/08/2017 15

16/08/2017 01/09/2017 18

4°C

16/08/2017 18/08/2017 3 0 00:18:52±00:14:48

16/08/2017 21/08/2017 5 0 00:12:13±00:08:21

16/08/2017 23/08/2017 8 0 00:12:11±00:01:25

16/08/2017 25/08/2017 10 1 00:10:28±00:00:23

16/08/2017 28/08/2017 13 0 00:19:50±00:08:54

16/08/2017 30/08/2017 15 1.33 00:10:23±00:02:00

16/08/2017 01/09/2017 18

16/08/2017 03/09/2017 20

Test with nectarines inoculated with M. laxa and M. fructicola

LAMP for Monilinia fructicola and M. laxa

Positive

Negative

LAMP assay

SpecieIsolate

nameTp (min:s) Anneal (°C)

Venturia inaequalis 2.16 B9 29:28±01:40 89.72±0.13

V. inaequalis 2.16 B1 26:67±01:30 89.38±0.03

V. inaequalis 2.16 B6 24:04±00:49 89.08±0.03

V. inaequalis 2.16 B7 26:37±00:52 89.06±0.15

V. inaequalis 2.16 B5 23:48±00:58 89.36±0.39

Validated with:• artificially inoculated apple leaves;• asymptomatic-naturally infected apple leaves.

LAMP for Venturia inaequalis

High-throughput sequencing

Sanger sequencing

NGS: 454-pyrosequencing (early 2000s)Early 2010s: Illumina and IonTorrent (greater throughput, lowercost)

TGS: PacBio (2011) and Oxford Nanopore (2105) longer DNA sequences

NGS and TGS: HTS

Metagenomics or metagenome sequencing

shotgun metagenomics: HTS of all DNAshotgun metatranscriptomics: HTS of all RNA

metabarcoding: amplicon-based HTS

Sampling

DNA extractionPCR

Sequencing

Data filtering and denoising

Sequenceclustering (OTU)

Taxonomicassignment

Interpretation

Sequencing

Cleaning

Assembling

Binning

Assembly-free taxonomicprofiling

Shotgun metagenomicsMetabarcoding

Metagenome sequencing

Gene predictionand annotation

Salter et al. (2015) sequenced with different kitprogressively more diluted cultures ofSalmonella bongori.

Amplification of 16S

Shotgun metagenomics

Sequencing of negative controls (buffers) is necessary.

High concentration of target DNACertification of DNA free kit

Contamination riskDNA is ubiquitous in commonly used DNA

extraction kits and other laboratory reagents,

varies greatly in composition between different

kits and kit batches.

This contamination critically impacts results from

samples containing a low microbial biomass.

Cell lysis is influenced by the cell wall composition of present microorganisms.Depending on the extraction protocol, differences in estimated abundance of the 25 most abundant bacterial generaWesolowska-Andersen et al. (2014)

Preliminary studies are necessary to establish the best method to extract DNA from target populations

with minimum possible bias.Internal controls (known amounts of different MO).

Choosing the protocol

Comparison of abundance estimations for bacterial genera with MetaHIT and HMP methods

Sampling

DNA extractionPCR

Sequencing

Data filtering and denoising

Sequenceclustering (OTU)

Taxonomicassignment

Interpretation

Sequencing

Cleaning

Assembling

Binning

Assembly-free taxonomicprofiling

Shotgun metagenomicsMetabarcoding

Metagenome sequencing

Gene predictionand annotation

Illumina amplicon correction withDeblur (Amir et al., 2017), DADA2(Callahan et al., 2016) andUNOISE2 (Edgar, 2016) allows toskip OTU construction, obtaininginstead ESVs (Exact SequenceVariants).

ESV: consistent in different experimentsOTU: experiment-dependent

Species-level identification possible.

Pros:• Not much computational power is needed• Not expensive• Generated data are less complex than in shotgun

metagenomics.

Cons:• OTU construction at 97% similarity threshold

underestimates real biological differences.• Even the best primers only detect roughly 50% of

microbial community (Scibetta et al., 2018).• ITS sequencing does not allow precise species

identification in some taxa.

Metabarcoding

Oligotyping focuses on specifichighly variable sites insequences to identify diversityin specific target taxa.

(Eren et al., 2013)

Pros:• Not much computational power is needed• Not expensive• Generated data are less complex than in shotgun

metagenomics.

Cons:• OTU construction at 97% similarity threshold

underestimates real biological differences.• Even the best primers only detect roughly 50% of

microbial community (Scibetta et al., 2018).• ITS sequencing does not allow precise species

identification in some taxa.

Metabarcoding

Pros:• Not much computational power is needed• Not expensive• Generated data are less complex than in shotgun

metagenomics.

Cons:• OTU construction at 97% similarity threshold can

underestimate real biological differences.• Even the best primers only detect roughly 50% of

microbial community (Scibetta et al., 2018).• ITS sequencing does not allow precise species

identification in some taxa.

Results improved with:- more primers for the same locus- degenerate primers- primers for more loci

Metabarcoding

Pros:• Not much computational power is needed• Not expensive• Generated data are less complex than in shotgun

metagenomics.

Cons:• OTU construction at 97% similarity threshold can

cause the missing of real biological differences.• Even the best primers only detect roughly 50% of

microbial community (Scibetta et al., 2018).• ITS sequencing does not allow precise species

identification in some taxa.

Metabarcoding

Many important plant pathogens are Ascomycetes(Alternaria spp., Fusarium spp., Aspergillus spp., Penicillium spp.)

It does not allow identification beyond genus level formany Ascomycetes

ITS:• High sequence diversity• High number of copies per cell• Conserved primer sites• Numerous sequences in the database

Choosing the primers

(Scibetta et al. 2018)

RPB2:• Second largest subunit of

RNA polymerase 2• Most commonly used in

conjunction with others.• Good results in

Ascomycetes at specieslevel (Liu and Hall, 2004)

TEF1-α:• Translational elongation factor

1-alpha• Good for Fusarium and

Trichoderma.• The primers EF1-1018F/EF1-

1620R had an averageamplification success veryclose to that of ITS.

β-tubulin :• Good for important and

variable genera such asPenicillium and Aspergillus.

• Some commonly usedprimers can preferentiallyamplify the paralog tubC,resulting in incongruentanalysis.

MCM7:• Component of the minichromosome maintenance proteins

complex.• Good for Camarops, Lasiosphaeria, Aspergillius,…• Contrastings results on Penicillium.• Availability of degenerate primers for the amplification of

MCM7 in a wide range of most ascomycetes.

Calmodulin:• Good for Aspergillus, Penicillium and

Fusarium.

Choosing the primers

Fusarium spp.: TEF

(2016)

Metabarcoding with 454-pyrosequencing

Penicillium spp.: beta-tubulin

Aspergillus spp.: calmodulin

Databases need to contain enough information for each genus of interest

Databases are often not

complete and/or revised.

In-house databases

can be built.

(time-consuming)

Databases

Initiatives such as UNITE, SILVA and UniEuk have generated databases and reference datasets populated with

filtered and third‐party annotated sequences.

Illlumina, but...Use of PacBio for metabarcoding

1.5-kb long fragment covering parts of SSU, ITS and parts of the large ribosomal subunit (LSU).

(2016)

Choice of the sequencing method

Sampling

DNA extractionPCR

Sequencing

Data filtering and denoising

Sequenceclustering (OTU)

Taxonomicassignment

Interpretation

Sequencing

Cleaning

Assembling

Binning

Assembly-free taxonomicprofiling

Shotgun metagenomicsMetabarcoding

Metagenome sequencing

Gene predictionand annotation

Very good for viruses:1 - small genomes2 - low level of sequenceconservation (no universalprimers).

Pros:• Avoids PCR-derived biases.• Allows strain-level resolution for abundant species (eg: viruses).

Cons:• Expensive.• Complex operations required to interpret the data.• Taxonomical assignment difficult when sequencing complex

communities.• Assembling and binning are problematic when many related

strains are present (risk of chimera).• Risk of mistakenly detect pathogens from conserved sequences

originating from non-pathogenic organisms whose genome isnot available.

Shotgun metagenomics

Afshinnekoo et al., 2015:Yersinia pestis andBacillus anthracison the New York subwayRetraction

Pros:• Best base/cost ratio• High quality (99.9% accuracy)

Cons:• Short reads (150-300 bps)

Sequencing methods: Illumina

Pros:• Higher read length (400-450 bp)• More robust species inference

Cons:• More expensive

Sequencing methods: IonTorrent

Pros:• 30-100 kb reads• higher average contig length.• enhancements in binning and genome reconstruction in

shotgun metagenomics (Frank et al., 2016)• Allows for sequencing of the full ITS region and flanking

rRNA small subunit gene (Schlaeppi et al. 2016).

Cons:• Much more expensive

Source: Rhoads and Au (2015)

Sequencing methods: PacBio

Pros:• All the advantages of PacBio• Very fast (minutes)• Small and portable• Operable by a laptop

Cons:• Low accuracy (95%)

Source: https://phys.org/news/2014-02-oxford-nanopore-unveils-portable-genome.html

Sequencing methods: Oxford Nanopore

Databases do not contain enough information for allthe species

Genbank not

revised.

Databases

WGS lacking,

particularly for non

pathogens.

Discovery of new pathogens

Shotgun metagenomics on plant tissues

New pathogen species or strains- > 100 new viruses through HTS- Identification of Calonectria pseudonaviculata, agent of blight on Sarcococca- Duan et al. 2009: WGS of ‘Candidatus Liberibacter asiaticus’ from the psyllid vector- Adams et al. 2011: identified Xanthomonas in a diseased Hedera (ivy)

Causality: relationship between microorganism presence and diseaseKoch’s postulatesAdams et al. 2018: 10 factors to infer causality(including abundance of OTUs/coverage)

Biological characterizationPest risk analysis (PRA)NPPOs: Risk assessment / Risk management

Shotgun metagenomics to improve targeted diagnostics

qPCR or LAMP

Origin of an outbreak

In human health:FDA: HTS as default tracking tool for microbial foodpoisoning outbreak.

In plant pathology:Wheat yellow rust, obligate parasiteField pathogenomics (shotgun metatanscriptomics)4 different lineages, from 219 wheat samples across UK

soilborne

postharvest

foliarseedborne

Surveillance and monitoring

SoilSeed and propagation materialPlant products: fruit, leaves, woodAir

HTS: untargeted approachBaseline surveys airborne

• Metagenome sequencing has great potential for pest survelliance

• It is difficult to distinguish between new pathogens and already presentbut previously unreported organisms.

• EPPO underlines the importance of baseline surveys (knowledge of a pathogenstatus into a territory needed) for correct metagenomic data interpretation.

• Some countries, like Belgium, began projects to detect any virus present incultivated and wild plants of selected botanical families or genera within theirrespective territories.

Monitor background population

Source: Tremblay et al. 2018

Monitoring of pathogens: IonTorrent

Air samplesSoil samplesInsect samples

ESV for surveillance of pathogens in seed lots of forest trees

Source: Franic et al. (2019)

Illumina Miseq

Exact Sequence Variants

• Air microbiome is frequently understudied.• 4-11% of fine particle mass is made up of fungal spores.• Spores can move over 500 km (rusts).• Air is important for transmission of many diseases.• Spore identification is time-consiming.• Forecasting models capable of predicting the probability

of infection of a crop (frequently measure spores).

Airborne pathogens

52 forecasting models for Magnaporthe oryzae.

Brown spot (Cochliobolus miyabeanus), rice blast (Magnaporthe oryzae)

and bakanae (Fusarium fujikuroi) are airborne.

Rice pathogens

Traditional monitoring: quantification of airborne conidia under a microscope from

spore trap samples.

Difficult, especially if insects, dust or other materials are captured by the spore trap.

UNTARGETED APPROACHES

Metabarcoding of the ITS rDNA- (ITS1)-Illumina platform

Ky02F (5’- TAGAGGAAGTAAAAGTCGTAA-3’) and ITS1 Wobble (5’- CWGYGTTCTTCATCGATG-3’)

PE: 2X300 bp

TARGETED APPROACHES

qPCR for the detection of Magnaporthe oryzae

qPCR for the detection of Cochliobolus miyabeanus

Rice paddy

Spore trap: aerial mycobiome

Daily sampling: 73 days (25 June – 12 September)

Cochliobolus

Magnaporthe

Fusarium

Rice aerial mycobiome

Alpha-diversity of rice aerial mycobiome

GenusRelative abundance

(Average over season)

Cladosporium 20.1771

Alternaria 9.209575

Myrothecium 7.312185

Ascomycota 7.026058

Epicoccum 5.786639

Basidiomycota 5.292312

Sporidiobolales 5.154338

Sordariomycetes 4.44805

Davidiella 4.298913

Russulaceae 2.72734

Leptosphaerulina 2.332917

Entylomatales 1.851611

Magnaporthe 1.573834

Auriculibuller 1.515907

Pleosporales 1.356646

Sporobolomyces 0.975467

Tremellaceae 0.880249

Lewia 0.672838

Cochliobolus 0.568411

Uromyces 0.543934

Hyphodontia 0.526904

Fusarium 0.516212

Polyporales 0.501581

Schizophyllum 0.496519

Coriolopsis 0.487519

Aspergillus 0.460448

Sclerotiniaceae 0.450546

The difficult collection of the airborne particulate, the low amount of

organisms collected in the samples and the problematic DNA extraction

are challenging steps for airborne-mycobiota studies.

Method for DNA extraction from sticky tape.

Oligotyping was used to analyze the most interesting OTUs.

Strong correlation between the output of the qPCR and the Magnaporthe oligotypes, which were identified as M. oryzae and M. grisea, demonstrated the reliability of this analysis even with fungi.

qPCR confirmation

M. oryzae –Rice blast symptoms on leaves

Date 19.7.16 26.7.16 2.8.16 10.8.16 17.8.16 23.8.16 30.8.16

Disease index 3.00 3.75 4.00 3.25 6.50 8.25 9.00

The cells of C. miyabeanus were not correlated with any of the oligotypes of Cochliobolus that were identified as different species.

The combination with the TaqMan assay could solve the limit of detection of oligotyping.

Conclusions

High potential of HTS for diagnostics.

Metabarcoding, due to the low cost, has the greatest array of applications in diagnostics. Choice of the best ITS primers able to amplify most Ascomycotes. Choice of primers based on the target genus.

Shotgun metagenomics has important advantages, but it is more expensive and it requires particular conditions to be applicable (low biodiversity, low number of related strains, mostly even organism distribution).

It remains the only HTS-based technique to detect viruses, due to small genomes and low level of sequence conservation.

Both strategies require: extraction protocols planned to not introduce biases, sequencing of negative controls and complete and trustworthy databases.

Acknowledgements

Silvia Valente, Betta Meloni, Simona Prencipe, Edoardo Piombo

Samir DrobyMichael Wisniewski

Leonardo Schena

Neil BoonhamIan Adams

Ilario Ferrocino